CN102804233B

CN102804233B - Paper discriminating gear and paper method of discrimination

Info

Publication number: CN102804233B
Application number: CN201080065507.5A
Authority: CN
Inventors: 鹈饲和岁; 横田政宪
Original assignee: Glory Ltd
Current assignee: Glory Ltd
Priority date: 2010-03-17
Filing date: 2010-03-17
Publication date: 2015-08-12
Anticipated expiration: 2030-03-17
Also published as: CN102804233A; WO2011114447A1

Abstract

A kind of paper discriminating gear, use photographed images and the study image of proper paper, and based on the edge direction of the fringe region detected from each study image and this fringe region, generate each edge direction by direction dictionary image, by to become differentiate object paper photographed images and input picture and contrast by direction dictionary image, thus extract the remaining fringe region only existed in the input image, detect the density of the remaining fringe region in the surface area of regulation, and the complete damage of carrying out paper based on the density detected differentiates and kind differentiates.

Description

Paper discriminating gear and paper method of discrimination

Technical field

The present invention relates to paper discriminating gear and paper method of discrimination that the complete damage of the paper of bank note etc. is differentiated, even in particular to there is multifarious situation on the pattern or quality of paper under, also high precision can be carried out and most suitable complete paper discriminating gear and the paper method of discrimination damaging differentiation and kind differentiation.

Background technology

In the past, known use imageing sensor etc. and differentiate the stained bank note producing spot, fold, folding line etc. and the bank note discriminating gear not producing above-mentioned stained bank note.In addition, below, the so stained bank note of generation being recited as " damage certificate ", being recited as not producing above-mentioned stained bank note " complete certificate ".

This bank note discriminating gear is generally (following by the photographed images of the complete certificate obtained using imageing sensor etc., be recited as " benchmark image ") (following with the photographed images becoming the bank note differentiating object, be recited as " differentiation object images ") compare and analyze, thus differentiate.

Further, when carrying out such comparison and analyzing, the various methods at the edge using image etc. are proposed.Here, edge refers to the position of the deep or light marked change of image, and the concentration vergence direction at this edge is called edge direction.

Such as, in patent documentation 1, disclose following technology: based on benchmark image, stainedly on pattern will easily seem that significant region pre-determines as defect detection subject area, the deviation at the edge in this defect detection subject area is analyzed, thus has carried out damaging differentiation.

In addition, stainedly on pattern easily seem that significant region refers to, edge do not concentrate region, namely deep or light simple area of the pattern with low uncertainty.

In addition, in patent documentation 2, disclose following technology: edge direction, benchmark image and differentiation object images in units of pixel, and the deviation of the edge direction detected being compared by this is analyzed, thus carried out damaging differentiation.

Prior art document

Patent documentation

Patent documentation 1: Japanese Unexamined Patent Publication 2006-302109 publication

Patent documentation 2: patent No. 3207931 publication

Summary of the invention

The problem that invention will solve

But, when employing the technology of above-mentioned patent documentation 1, owing to carrying out the comparison only for being condensed to the stained significant region that easily seems, so there is the complete problem points of damaging differentiation can not carried out when producing stained beyond this region.

In addition, when employing the technology of above-mentioned patent documentation 2, due to benchmark image and differentiation object images in units of pixel, so when creating printing deviation at multiple printing, even if such deviation is trickle, be also determined as the possibility of damaging certificate high.

Can say this point, this situation being determined as and damaging certificate is carry out the high differentiation of precision, but then, also becomes and will not have the bill handling of obstacle when circulating as damage certificate, thus imappropriate.

Thus, even also can carry out high precision under how realizing there is multifarious situation on the pattern or quality of bank note and most suitable complete damage differentiate bank note discriminating gear or bank note method of discrimination become large problem.Whether diversity alleged here refers to, such as, be that edge is many and be stainedly difficult to the seem diversity of pattern of significant pattern or the variation etc. of the press quality of multiple printing.

In addition, this problem is not limited to bank note, is the same problem occurred there is multifarious situation on the pattern or quality of the paper of exchange ticket, bank's booklet etc. under.In addition, be the problem occurred too in differentiating using this diversity as the kind of the paper of prerequisite.

The present invention completes in order to the problem points eliminating above-mentioned conventional art, its object is to, a kind of paper discriminating gear and paper method of discrimination are provided, even under its pattern at paper or quality exist multifarious situation, also high precision can be carried out and most suitable complete damage differentiates and kind differentiates.

For solving the means of problem

In order to solve above-mentioned problem and realize object, the present invention is a kind of paper discriminating gear, based on paper photographed images and differentiate paper, it is characterized in that, comprise: by direction dictionary image generating member, when from just when the photographed images of paper and study image in detect in marginal position and this marginal position edge direction, by multiple by the dictionary image of direction what prepare the edge direction scope of each regulation, this marginal position is distributed by the corresponding position of direction dictionary image described in corresponding with this edge direction, thus generate described by direction dictionary image respectively, and judgement part, by becoming the photographed images of paper and input picture that differentiate object and describedly contrasting by direction dictionary image, thus differentiate complete damage and the kind of this paper.

In addition, in above-mentioned invention, the invention is characterized in, also comprise: by direction input picture generating unit, when detecting the edge direction in marginal position and this marginal position from described input picture, by being separated this marginal position based on this edge direction to each described edge direction scope, thus generate by direction input picture; And remaining edge extracting parts, by by described in relevant with same described edge direction scope by direction input picture and described by the dictionary image of direction, only described by direction input picture in the described marginal position that exists, whole described edge direction scope is overlapped, thus extract remaining fringe region, described judgement part, based on the described remaining fringe region extracted by described remaining edge extracting parts, differentiates the complete damage becoming the paper of described differentiation object.

In addition, in above-mentioned invention, the invention is characterized in, described when belonging in the scope of regulation from the boundary line of adjacent two described edge direction scopes by the described edge direction of direction dictionary image generating member in the described marginal position detected, distribute this marginal position to described in these two edge direction scopes are corresponding respectively by the corresponding position of direction dictionary image.

In addition, in above-mentioned invention, the invention is characterized in, also comprise: Density Detection parts, detect the density of the described remaining fringe region extracted by described remaining edge extracting parts, described judgement part, based on the described density detected by described Density Detection parts, differentiates the complete damage becoming the paper of described differentiation object.

In addition, in above-mentioned invention, the invention is characterized in, the described described study image using regulation number by direction dictionary image generating member, and will the pixel of the described marginal position of more than stated number be assigned with as efficient frontier region.

In addition, in above-mentioned invention, the invention is characterized in, described by direction dictionary image generating member to each described expansion process of being undertaken described marginal position or described efficient frontier region being expanded to neighboring pixel by direction dictionary image.

In addition, in above-mentioned invention, the invention is characterized in, describedly to be undertaken the described expansion process preferential with the described described edge direction scope corresponding by direction dictionary image by direction dictionary image generating member.

In addition, in above-mentioned invention, the invention is characterized in, also comprise: periphery edge counter block, the number being assigned with the pixel of described marginal position comprised in the nearby scope of the described regulation by each pixel in the dictionary image of direction is counted; And normalized set parts, calculate the statistic of by described periphery edge counter block, the described number of each pixel counts will be added up multiple described study image, described by direction dictionary image generating member when going out described statistic by described normalized set component computes, by setting the described pixel value by the pixel in the dictionary image of direction based on this statistic, thus generate described by direction dictionary image.

In addition, the present invention is a kind of paper method of discrimination, based on paper photographed images and differentiate paper, it is characterized in that, comprise: by direction dictionary Computer image genration step, when from just when the photographed images of paper and study image in detect in marginal position and this marginal position edge direction, by multiple by the dictionary image of direction what prepare the edge direction scope of each regulation, this marginal position is distributed by the corresponding position of direction dictionary image described in corresponding with this edge direction, thus generate described by direction dictionary image respectively, and discriminating step, by becoming the photographed images of paper and input picture that differentiate object and describedly contrasting by direction dictionary image, thus differentiate complete damage and the kind of this paper.

Invention effect

According to the present invention, due to when from just when the photographed images of paper and study image in detect in marginal position and this marginal position edge direction, by multiple by the dictionary image of direction what prepare the edge direction scope of each regulation, distribute by the corresponding position of direction dictionary image the marginal position detected to corresponding with the edge direction detected, thus generate respectively by direction dictionary image, by to become differentiate object paper photographed images and input picture and contrast by direction dictionary image, thus differentiate complete damage and the kind of this paper, therefore, play following effect: even there is multifarious situation on the pattern or quality of paper under, also high precision can be carried out and most suitable complete damage differentiation and kind differentiation.

In addition, according to the present invention, due to when detecting the edge direction in marginal position and this marginal position from input picture, by being separated this marginal position based on the edge direction detected to each edge direction scope, thus generate by direction input picture, by pressing direction input picture by relevant with same edge direction scope and press in the dictionary image of direction, the marginal position only existed in by direction input picture, whole edge direction scope is overlapped, thus extract remaining fringe region, and the complete damage becoming the paper differentiating object is differentiated based on the remaining fringe region extracted, therefore, play following effect: can more clearly extract remaining fringe region from multiple edge direction, even the pattern of pattern complexity and stained part is difficult to seem significant paper, also high precision can be carried out and most suitable complete damage differentiation and kind differentiation.

In addition, according to the present invention, due to the edge direction in the marginal position detected belong in the scope of regulation from the boundary line of adjacent two edge direction scopes, distribute by the corresponding position of direction dictionary image the marginal position detected to these two edge direction scopes are corresponding respectively, therefore, following effect is played: the dictionary image being good at resisting rotating deviation that can be created on the rotating deviation uptake adding regulation in the edge direction detected.

In addition, according to the present invention, due to the density of the remaining fringe region that Detection and Extraction go out, and the complete damage becoming the paper differentiating object is differentiated based on the density detected, therefore, play following effect: can carry out by limit is have width to a certain degree to differentiate with the complete damage of the stained part of height, can prevent the mistake of the fold of linearly shape extension on paper face etc. from differentiating.

In addition, according to the present invention, owing to using the study image of regulation number, and will the pixel of the marginal position of more than stated number be assigned with as efficient frontier region, therefore, following effect is played: the dictionary image that can generate the printing deviation considered between certificate.

In addition, according to the present invention, due to each described expansion process of being undertaken marginal position or efficient frontier region being expanded to neighboring pixel by direction dictionary image, therefore, play following effect: the extractability that can not reduce remaining fringe region, just can generate the dictionary image being good at resisting position deviation.In addition, compared with there is not according to direction the situation of dictionary image, the effect that can increase swell increment is also played.

In addition, according to the present invention, owing to carrying out, by expansion process preferential for the edge direction scope corresponding with by direction dictionary image, therefore, playing following effect: the dictionary image expanded in the direction that can be created on the detection error that edge occurs.

In addition, according to the present invention, owing to counting the number being assigned with the pixel of marginal position comprised in the nearby scope of the regulation by each pixel in the dictionary image of direction, calculate the statistic of will the number of each pixel counts be added up multiple study image, when calculating statistic, by the pixel value set by the pixel in the dictionary image of direction based on this statistic, thus generate by direction dictionary image, therefore, play following effect: the dictionary image corresponding with the characteristic of the periphery of each pixel (diversity etc. of pattern) can be generated.

Accompanying drawing explanation

Fig. 1 is the figure of the summary representing paper method of discrimination of the present invention.

Fig. 2 is the block scheme of the structure of the bank note discriminating gear representing embodiment 1.

Fig. 3 is the figure of the edge direction check processing carried out for illustration of edge direction test section.

Fig. 4 is the figure of the efficient frontier determination processing of carrying out for illustration of efficient frontier detection unit.

Fig. 5 is the figure of the expansion process of carrying out for illustration of expansion process portion.

Fig. 6 is the figure of the variation representing expansion process.

Fig. 7 is the figure of the remaining edge extracting process carried out for illustration of remaining edge extracting portion.

Fig. 8 is the figure of the differentiation process that the remaining edge analysis process carried out for illustration of remaining edge analysis portion and judegment part carry out.

Fig. 9 is the process flow diagram representing the treatment step that the bank note discriminating gear of embodiment 1 performs.

Figure 10 is the process flow diagram of the treatment step representing the dictionary Computer image genration process that the bank note discriminating gear of embodiment 1 performs.

Figure 11 is the complete process flow diagram damaging the treatment step differentiating process representing that the bank note discriminating gear of embodiment 1 performs.

Figure 12 is the figure of the summary of the dictionary Computer image genration of the bank note discriminating gear representing embodiment 2.

Figure 13 is the block scheme of the structure of the bank note discriminating gear representing embodiment 2.

Figure 14 is the process flow diagram of the treatment step representing the dictionary Computer image genration process that the bank note discriminating gear of embodiment 2 performs.

Figure 15 is the complete process flow diagram damaging the treatment step differentiating process representing that the bank note discriminating gear of embodiment 2 performs.

Embodiment

Below, the preferred embodiment of the paper method of discrimination that present invention will be described in detail with reference to the accompanying.In addition, below, after use Fig. 1 illustrates the summary of paper method of discrimination of the present invention, the embodiment of the bank note discriminating gear of Fig. 2 ~ Figure 15 application paper method of discrimination of the present invention is used.In addition, below, the situation of mainly having carried out damaging differentiation in paper differentiates is described.

In addition, below, bank note discriminating gear is processed the photographed images of proper paper and study image and by direction generate being recited as " dictionary image " by direction dictionary image.

First, use Fig. 1 that the summary of paper method of discrimination of the present invention is described.Fig. 1 is the figure of the summary representing paper method of discrimination of the present invention.

As shown in the figure, the feature of paper method of discrimination of the present invention is, has detected marginal point and its edge direction of certificate image, and has generated the dictionary image of this each edge direction.Here, suppose that marginal point is the pixel at the edge corresponded in image.

In addition, being further characterized in that, when differentiating the comparing and analyze of object images and dictionary image, obtaining the difference of differentiation object images and dictionary image (below, be recited as " remaining marginal point "), and use the density of this remaining marginal point in given size region and carried out damaging and differentiate.

Below, paper method of discrimination of the present invention is illustrated.In addition, paper method of discrimination of the present invention can roughly be divided into, and complete certificate image 1 is damaged as the complete of input data the stage differentiated as the stage of the dictionary Computer image genration of input data and by differentiation object images 5.

In paper method of discrimination of the present invention, first, as the stage of dictionary Computer image genration, whole marginal point and its edge direction (with reference to Fig. 1 (1)) are detected to the complete certificate image 1 of multiple (=N opens).Further, the marginal point of the complete certificate image opened based on this N and edge direction, generate the dictionary image (with reference to Fig. 1 (2)) of each edge direction.

Such as, as shown in Fig. 1 (2), to suppose edge direction by 360 degree 60 degree for scale is divided into 6 directions.Further, all directions in 6 directions divided are distributed to the address in unique A ~ F direction.In addition, below, the division in this direction is recited as " quantization ", number of partitions is recited as " quantization number ".

And, in paper method of discrimination of the present invention, prepare the plane (following, be recited as " direction plane ") corresponding with this quantization, to the respective coordinates of all directions plane corresponding to the edge direction of each marginal point with the complete certificate image 1 detected, store the situation detecting marginal point.Then, the dictionary image of coordinate as efficient frontier point of the storage number of times using having more than stated number is generated.

Such as, as shown in Fig. 1 (2), prepare to correspond to direction plane A, the direction plane B corresponding to B direction in A direction, the direction plane C corresponding to C direction, the direction plane D corresponding to D direction, the direction plane E corresponding to E direction, direction plane F corresponding to F direction, each marginal point is stored all directions plane A ~ F by each edge direction, generates the dictionary image in each direction.In addition, Fig. 3 ~ Fig. 5 is used to describe the details of this dictionary Computer image genration later.

Here, also relative direction can be stored on same direction plane.This is because detect that near each marginal point the situation of the marginal point of the edge direction of 180 degree of symmetries is more.Such as, the marginal point in A direction and D direction can be stored in direction plane A, in direction plane B, store the marginal point in B direction and E direction, in direction plane C, store the marginal point in C direction and F direction.Now, relative to 6 of quantization number, direction plane number becomes 3.

In addition, in the stage of this dictionary Computer image genration, in paper method of discrimination of the present invention, each pixel as marginal point is carried out to the expansion of ormal weight.Such as, in Fig. 1 (2), represent that the pixel Px of part surrounded by circle o is registered to the example of dictionary image after being expanded to range of expansion D0.In addition, Fig. 5 is used to describe the details of the expansion of this ormal weight later.

In addition, in the technology (such as, patent documentation 2) that make use of edge direction in the past, the method employing the dictionary image generating each edge direction as paper method of discrimination of the present invention is not had.Therefore, dictionary image is the panel data that a piece of paper is opened, and needs marginal point and its edge direction to be defined as one.

In contrast, in paper method of discrimination of the present invention, owing to generating dictionary image by each edge direction, so such as to the marginal point of the edge direction had near quantized border, can process across 2 edge directions.That is, to an image of dictionary image, there are 2 edge directions.Therefore, it is possible to the complete damage carrying out absorbing the rotating deviation occurred when the printing of paper and when scanning etc. differentiates.In addition, below, the degree of the deviation from the quantized border of carrying out processing across 2 edge directions is recited as " rotating deviation uptake ".

In addition, the expansion about above-mentioned ormal weight is also identical, expands pixel value or edge direction that a marginal point has by expanding, thus the complete damage can carrying out the diversity or various position and rotating deviation absorbing pattern differentiates.In addition, be not originally the pixel of marginal point when being included in the range of expansion (that is, in " ormal weight ") of different direction planes in the expansion by this ormal weight, this pixel can have the edge direction corresponding with all directions plane.Fig. 5 is used to describe this point later.In addition dictionary image coordinate, owing to being judged by efficient frontier point, will there is the coordinate of the marginal point amount of more than stated number as efficient frontier point from multiple complete certificate image, so can be made to have multiple edge direction.Therefore, it is possible to the complete damage of the diversity or multiple rotation and position deviation of carrying out the pattern absorbing paper differentiates.

Then, damage in the stage differentiated complete, in paper method of discrimination of the present invention, detect that the differentiation object images 5 of marginal point and its edge direction and each dictionary image compare ((3) with reference to Fig. 1) to identical with the situation of complete damage image 1.Further, use and to be compared by this and the density in the given size region of remaining marginal point that extracts has been carried out damaging and differentiated ((4) with reference to Fig. 1).In addition, Fig. 8 is used to describe the details of this point later.

In addition, in the technology (such as, patent documentation 1) that make use of remaining marginal point in the past, employ the threshold value whether exceeding regulation according to the pixel count detected as remaining marginal point and carried out damage and sentence method for distinguishing.Therefore, the fold etc. that paper face extends using linearity is differentiated also more as stained situation.

In contrast, in paper method of discrimination of the present invention, owing to using the density of the remaining marginal point in given size region, so only can differentiate to have width and highly stained, can prevent the mistake of above-mentioned fold etc. from differentiating.

Thus, in paper method of discrimination of the present invention, generate the dictionary image that edge point carries out each edge direction of the expansion of ormal weight, use the density in the given size region of remaining marginal point to differentiate to have carried out damaging.Therefore, even there is multifarious situation on the pattern or quality of paper under, also high precision can be carried out and most suitable complete damage differentiation.

Below, the embodiment of the bank note discriminating gear applying the paper method of discrimination of the present invention using Fig. 1 to illustrate is described in detail.In addition, below, generation has been carried out the embodiment of the bank note discriminating gear of the dictionary image of the expansion of ormal weight as embodiment 1 to each marginal point, periphery edge number based on each marginal point is generated the embodiment 2 of the bank note discriminating gear of dictionary image as embodiment 2, is described respectively.In addition, in each following embodiment, the situation of bank note discriminating gear by the image of image line sensor (following, to be recited as " line sensor ") shooting bank note is described.

Embodiment 1

Fig. 2 is the block scheme of the structure of the bank note discriminating gear 10 representing embodiment 1.In addition, in same figure, only represent for illustration of the structural element needed for the feature of bank note discriminating gear 10, eliminate the record of general structural element.

As shown in Figure 2, bank note discriminating gear 10 comprises line sensor portion 11, control part 12, storage part 13.In addition, control part 12 comprises configuration part 12a, view data obtaining section 12b, edge direction test section 12c, efficient frontier detection unit 12d, expansion process portion 12e, remaining edge extracting portion 12f, remaining edge analysis portion 12g, judegment part 12h further.In addition, storage part 13 stores set information 13a, bank of filters 13b, by direction plane marginal point amount 13c, by direction dictionary image 13d, discrimination standard information 13e.

Line sensor portion 11 is the sensors through light or reflected light accepting bank note that freely not shown connecting gear transmits, by being formed with linearity configuration by multiple optical receiving sensor.In addition, line sensor portion 11 carries out the process of the view data obtaining section 12b data obtained being outputted to control part 12 in the lump.

Control part 12 is the handling parts be handled as follows: the view data generating bank note based on the output from line sensor portion 11, and to the view data detected edge points generated and its direction, when view data is the complete certificate image of multiple (=N opens), dictionary image is generated based on the marginal point detected and its direction, and be that when differentiating object images, complete damage of carrying out bank note to differentiating object images and dictionary image to compare and analyze differentiates in view data.

Configuration part 12a is the handling part be handled as follows: the various parameters being extracted in the quantization number, rotating deviation uptake etc. comprised in the set information 13a of storage part 13, carries out generating with dictionary and complete damage differentiates relevant initial setting.In addition, configuration part 12a carries out the process various parameters extracted being outputted to edge direction test section 12c in the lump.

View data obtaining section 12b is the handling part be handled as follows: to the synthesis of each bank note from the output in line sensor portion 11, generate the view data of all bank note.In addition, view data obtaining section 12b is when generating dictionary image, and unification generates the complete certificate view data of N tensor.

In addition, view data obtaining section 12b carries out the process view data of generation being outputted to edge direction test section 12c in the lump.

Edge direction test section 12c is the handling part be handled as follows: view data smoothingization exported view data obtaining section 12b, detects the whole marginal point in the view data of smoothing and its direction.

In addition, in the detection in this smoothing, marginal point and its direction, edge direction test section 12c is used in the bank of filters 13b of storage part 13 the various wave filters prestored.Fig. 3 is used to describe this point later.

In addition, edge direction test section 12c is handled as follows in the lump: when generating dictionary image, the marginal point detected and its direction are outputted to efficient frontier detection unit 12d, when the complete damage of carrying out bank note differentiates, the marginal point detected and its direction are outputted to remaining edge extracting portion 12f.

Here, use Fig. 3 further describes the edge direction check processing that edge direction test section 12c carries out.Fig. 3 is the figure of the edge direction check processing carried out for illustration of edge direction test section 12c.

In addition, below, represent the edge direction check processing employing wave filter, but when this wave filter is applied to image, the pixel on the image overlap the center pixel of wave filter, is below recited as " concerned pixel ".

As shown in Figure 3, edge direction test section 12c carries out the complete certificate image 1 that view data obtaining section 12b exports or the smoothing ((1) with reference to Fig. 3) differentiating object images 5.In this smoothing, such as can use the coefficient near centered by concerned pixel 3 × 3 be all 1 smoothing wave filter f1.

Further, by carrying out this smoothing, certificate image 1 can have been removed or differentiated the noise component in object images 5.Therefore, also can the application (such as, 3 times) of repeatedly smoothing wave filter f1, till obtaining best noise removing result.Or, also can increase the size (such as, near 7 × 7) of smoothing wave filter f1, thus only apply 1 time.In addition, even when not applying smoothing wave filter f1, also can detected edge points and edge direction.

Then, edge direction test section 12c has detected certificate image 1 or has differentiated the marginal point ((2) with reference to Fig. 3) of object images 5.Here, in the detection of this marginal point, such as, Laplace operator (laplacian) wave filter f2 can be used.That is, the zero crossing detected by application laplacian filter f2 has been become certificate image 1 or has differentiated the marginal point of object images 5.

Then, edge direction test section 12c such as uses Prewitt wave filter f3 and f4, has detected the edge direction ((3) with reference to Fig. 3) of certificate image 1 or differentiation object images 5.Here, the white arrow in comprise in rectangle r 5 black picture elements as shown in Figure 3 represents each edge direction detected.Then, the edge direction ((4) with reference to Fig. 3) in each marginal point is detected.

In addition, in the explanation employing Fig. 3, illustrate the example using laplacian filter f2, Prewitt wave filter f3 and f4 and detected edge points and its direction, but also can use other wave filter of Sobel wave filter etc., also various wave filter suitably can be combined.In addition, also content is not particularly limited about the size of various wave filter and coefficient value.

Turn back to the explanation of Fig. 2, efficient frontier detection unit 12d is described.Efficient frontier detection unit 12d judges that each pixel of direction plane is whether as the handling part of efficient frontier point, use each pixel of direction plane is stored undertaken this by direction plane marginal point amount 13c and judge.

In addition, efficient frontier detection unit 12d carries out the process all directions plane saving efficient frontier point being outputted to expansion process portion 12e in the lump.

Here, use Fig. 4 further describes the efficient frontier determination processing that efficient frontier detection unit 12d carries out.Fig. 4 is the figure of the efficient frontier determination processing of carrying out for illustration of efficient frontier detection unit 12d.In addition, (A) of Fig. 4 has represented the marginal point gauge number process in the arbitrary marginal point P1 of certificate image, (B) of Fig. 4 represents the marginal point gauge number process in arbitrary marginal point P2 in the same manner, and (C) of Fig. 4 represents the efficient frontier determination processing in this marginal point P1 and P2.

As shown in (A) of Fig. 4, when detecting that the edge direction of marginal point P1 is A direction by edge direction test section 12c, the marginal point amount of the pixel P1 of the co-located in direction plane A is increased counting by efficient frontier detection unit 12d.

In addition, as shown in (B) of Fig. 4, when detecting that the border of the edge direction of marginal point P2 from A direction and F direction belongs within rotating deviation uptake, the marginal point amount of the pixel P2 of the co-located in the pixel P2 of the co-located in direction plane A and direction plane F is increased counting by efficient frontier detection unit 12d.

That is, as shown in (B) of Fig. 4, when the edge direction of marginal point belongs within rotating deviation uptake from border, the marginal point amount of the pixel of the co-located of both direction plane is increased counting by efficient frontier detection unit 12d.

Then, efficient frontier detection unit 12d carries out the marginal point gauge number process of this each marginal point to the whole complete certificate image obtained when dictionary Computer image genration, and using the pixel of marginal point the amount more than decision threshold each direction plane being obtained to regulation as efficient frontier point, each direction plane is preserved.

Specifically, as shown in (C) of Fig. 4, when having read certificate number and be " 200 " and decision threshold being all " 5 " to any pixel, the marginal point amount of " 200 " of more than decision threshold is obtained, so be saved as the efficient frontier point in direction plane A due to the pixel P1 in direction plane A.

In addition, similarly, the marginal point amount of " 180 " of more than decision threshold is also obtained due to the pixel P2 in direction plane A, so be saved as the efficient frontier point in direction plane A.

But, only obtain the marginal point amount of " 3 " of not enough decision threshold due to the pixel P2 in direction plane F, so be not saved as the efficient frontier point in direction plane F.

Turn back to the explanation of Fig. 2, expansion process portion 12e is described.Expansion process portion 12e is the handling part be handled as follows: the expansion whole efficient frontier points of all directions plane being carried out ormal weight.In addition, expansion process portion 12e carries out having carried out all directions plane after this expansion process and to be stored into as the dictionary image in each direction the process of storage part 13 in the lump.

Here, use Fig. 5 further describes the expansion process that expansion process portion 12e carries out.Fig. 5 is the figure of the expansion process of carrying out for illustration of expansion process portion 12e.In addition, (A) of Fig. 5 illustrates the example substantially of expansion process, and (B) of Fig. 5 illustrates from the step of expansion process to dictionary Computer image genration, and (C) of Fig. 5 illustrates the key diagram of the pixel about the same position comprised in multiple range of expansion.

As shown in (A) of Fig. 5, in the coordinate system be made up of m axle and n axle, suppose " m " to be " i " and " n " is that the pixel of " j " is as concerned pixel.In addition, below, when representing the position of pixel using pixel as " P ", and to record like this as " P(i, j) ".

Here, expansion process refers to that the pixel value by the nearby pixel of closing on this concerned pixel is transformed to the process of the pixel value of concerned pixel.In addition, in the conversion of this pixel value, wave filter can be used.

Such as, when concerned pixel is expanded to nearby 8 pixels of closing on (following, to be recited as " 1 pixel expansion "), P(i-1, j-1), P(i-1, j), P(i-1, j+1), P(i, j-1), P(i, j+1), P(i+1, j-1), P(i+1, j), P(i+1, j+1) each pixel value is transformed to concerned pixel P(i, j) pixel value (black picture element of (A) with reference to Fig. 5)).

The fundamental purpose of this expansion process is, absorbs the position deviation caused by expansion or the rotating deviation (with reference to Fig. 1) of pixel coverage.

In addition, here, exemplify and describe 1 pixel expansion, but this 1 pixel expansion can be repeated by the swell increment of specified.Such as, when 3 pixels are appointed as the swell increment of regulation, 1 pixel expansion is repeated 3 times.In addition, when having carried out this 3 pixel expansion, the pixel value of final concerned pixel is to each pixel expansion near 7 × 7 centered by concerned pixel.

In addition, as shown in (B) of Fig. 5, the whole efficient frontier points of expansion process portion 12e to all directions plane carry out above-mentioned expansion process ((B-1) with reference to Fig. 5).Then, after the expansion process of carrying out whole efficient frontier point, all directions plane is stored into (B-2) of storage part 13(with reference to Fig. 5 as the dictionary image in each direction).

Here, the pixel of the same position comprised in the range of expansion of multiple directions plane is described.As shown in (C) of Fig. 5, such as, suppose the pixel P3 of assigned position be included in the arbitrary range of expansion D1 of the direction plane A finishing expansion process, the arbitrary range of expansion D2 of direction plane F both in.

Now, not by direction but pixel P3 when regarding a complete certificate image as has 2 edge directions (scope with the angle θ shown in figure) in A direction, F direction.Therefore, it is possible to be judged to differentiate the pixel P3 of object images in A direction, which direction in F direction, its edge direction all exists in complete certificate image.

In addition, in the explanation using Fig. 5, describe pixel near whole centered by concerned pixel are closed on and carry out the situation of expansion process.But, also can carry out expansion process to pixel near the closing on of the direction corresponding with edge direction.

Therefore, below, use Fig. 6 that the variation of expansion process is now described.Fig. 6 is the figure of the variation representing expansion process.In addition, in the following description, also " m " is " i " and the pixel (with reference to Fig. 5) that " n " is " j " is described as concerned pixel.In addition, represent that the situation of the position of pixel is also identical with the situation of the explanation of above-mentioned Fig. 5.

As shown in (A) of Fig. 6, when concerned pixel has the edge direction in A or D direction, only can carry out expansion process to pixel near corresponding with this A or D direction closing on.Specifically, to P(i-2, j), P(i-2, j+1), P(i-2, j+2), P(i-1, j), P(i-1, j+1), P(i-1, j+2), P(i, j-1), P(i, j+1), P(i+1, j-2), P(i+1, j-1), P(i+1, j), P(i+2, j-2), P(i+2, j-1), P(i+2, j) etc. each pixel carry out expansion process.

In addition, as shown in (B) of Fig. 6, when concerned pixel has the edge direction in B or E direction, can only to P(i-1, the j-2 corresponding with this B or E direction), P(i-1, j-1), P(i-1, j), P(i-1, j+1), P(i-1, j+2), P(i, j-2), P(i, j-1), P(i, j+1), P(i, j+2), P(i+1, j-2), P(i+1, j-1), P(i+1, j), P(i+1, j+1), P(i+1, j+2) etc. each pixel carry out expansion process.

In addition, as shown in (C) of Fig. 6, when concerned pixel has the edge direction in C or F direction, can only to P(i-2, the j-2 corresponding with this C or F direction), P(i-2, j-1), P(i-2, j), P(i-1, j-2), P(i-1, j-1), P(i-1, j), P(i, j-1), P(i, j+1), P(i+1, j), P(i+1, j+1), P(i+1, j+2), P(i+2, j), P(i+2, j+1), P(i+2, j+2) etc. each pixel carry out expansion process.

Turn back to the explanation of Fig. 2, remaining edge extracting portion 12f is described.Remaining edge extracting portion 12f is the handling part be handled as follows: when the complete damage of carrying out bank note differentiates, obtain the difference differentiating object images and dictionary image, generate and be extracted only at the difference image differentiating the remaining marginal point existed in object images.

In addition, remaining edge extracting portion 12f carries out the process difference image of generation being outputted to remaining edge analysis portion 12g in the lump.

Here, use Fig. 7 further describes the remaining edge extracting process that remaining edge extracting portion 12f carries out.Fig. 7 is the figure of the remaining edge extracting process carried out for illustration of remaining edge extracting portion 12f.

As shown in Figure 7, each marginal point of differentiation object images 5 that edge direction test section 12c detects by remaining edge extracting portion 12f stores ((1) with reference to Fig. 7) all directions plane by each edge direction.In addition, differentiating that each marginal point of object images 5 comprises the marginal point outside the efficient frontier point forming dictionary image.

Then, remaining edge extracting portion 12f from storage part 13 store by extract the dictionary image 13d of direction as the bank note differentiating object by direction dictionary image ((2) of reference Fig. 7).

Then, remaining edge extracting portion 12f obtains and deducted each difference after each dictionary image corresponding with this all directions plane ((3) with reference to Fig. 7) from all directions plane storing each marginal point differentiating object images 5.

Then, remaining edge extracting portion 12f synthesizes this each difference and generates difference image 5 ' ((4) with reference to Fig. 7).Thus, in difference image 5 ', extract and only differentiating the remaining marginal point (comprising the marginal point outside effective marginal point) ((4) with reference to Fig. 7) existed in object images 5.

In addition, with in figure, the part of being surrounded by dotted line R1 or R2 is represented as remaining marginal point.

Turn back to the explanation of Fig. 2, remaining edge analysis portion 12g is described.Remaining edge analysis portion 12g is the handling part be handled as follows: remove the isolated point in the remaining marginal point extracted at remaining edge extracting portion 12f, generates the density image detected by the density of remaining marginal point in the size of regulation remained.

In addition, in the removing of this isolated point or the generation of density image, remaining edge analysis portion 12g is used in the bank of filters 13b of storage part 13 the various wave filters prestored.Fig. 8 is used to describe this point later.

In addition, remaining edge analysis portion 12g carries out the process density image of generation being outputted to judegment part 12h in the lump.

Judegment part 12h is the handling part be handled as follows: based on the density image using the threshold value of regulation to carry out 2 values, the final complete damage differentiated as the bank note differentiating object.In addition, about the threshold value of regulation, judegment part 12h is with reference to the discrimination standard information 13e of storage part 13.

Here, use Fig. 8 to further describe remaining edge analysis process that remaining edge analysis portion 12g carries out and the differentiation process that judegment part 12h carries out.Fig. 8 is the figure of the differentiation process that the remaining edge analysis process carried out for illustration of remaining edge analysis portion 12g and judegment part 12h carry out.

As shown in Figure 8, in the remaining marginal point of the difference image 5 ' that remaining edge analysis portion 12g exports at remaining edge extracting portion 12f, remove as isolated point from remaining marginal point ((1) with reference to Fig. 8) isolated around.Such as, in same figure, represent in the remaining marginal point of difference image 5 ', remove the remaining marginal point surrounded by dotted line R2 as isolated point, remain the situation of the remaining marginal point surrounded by dotted line R1 as final remaining marginal point.

Here, in the removing of this isolated point, isolated point removing wave filter f5(can be used with reference to (1) of Fig. 8).This isolated point removing wave filter f5 pixel near 8 of concerned pixel is all white, concerned pixel is transformed to white.

Further, remaining edge analysis portion 12g is to (2) of difference image 5 ' Application density detection wave filter f6 and f7(with reference to Fig. 8), the density of remaining marginal portion is detected as pixel value.In addition, this Density Detection wave filter f6 and f7 is the wave filter of the size of the regulation be made up of the wave filter element of L1 × L2.

Here, preferably the size of regulation is set to the stained minimum dimension wanting when complete damage differentiates to detect.In addition, filter coefficient all can be set to 1, but when considering the closeness of remaining marginal point, also can increase the coefficient near wave filter central portion.

In addition, in order to detect the density of the remaining marginal portion of growing crosswise, the Density Detection wave filter f6 that grows crosswise being applied to difference image 5 ', in order to detect the density of the remaining marginal portion of lengthwise, lengthwise Density Detection wave filter f7 being applied to difference image 5 '.

Such as, shown in (2-a) of Fig. 8 is situation difference image 5 ' being applied to the Density Detection wave filter f6 that grows crosswise.Now, the density of the remaining marginal portion of growing crosswise as the part of being surrounded by dotted line Ry each pixel value and represent.

In addition, as shown in (2-b) of Fig. 8, when applying lengthwise Density Detection wave filter f7 to difference image 5 ', the density of the remaining marginal portion of lengthwise as the part of being surrounded by dotted line Rt each pixel value and represent.

Further, the density image 5 of the application result having synthesized Density Detection wave filter f6 and f7 " is outputted to judegment part 12h by remaining edge analysis portion 12g.

Further, judegment part 12h uses the threshold value of regulation to density image 5 " carries out 2 values ((3) with reference to Fig. 8).The threshold value of this regulation is based on the above-mentioned size of regulation of Density Detection wave filter f6 and f7, the pixel value of the reference density of regulation.In addition, the reference density of regulation, if damaging certificate is then count relative to the ratio of the aggregate value of filter coefficient in the remaining edge that prediction detects in the size of regulation.

Here, the concrete example representing the threshold value of this regulation.Such as, " L1 " that suppose Density Detection wave filter f6 and f7 as shown in (2) of Fig. 8 is " 15 ", " L2 " be " 51 ", filter coefficient is all " 1 ".In addition, suppose that the reference density specified is " 0.05 ".Now, the multiplication value of the aggregate value " 765 (=15 × 51 × 1) " and the reference density " 0.05 " of regulation that the threshold value of regulation can be set to filter coefficient is

Therefore, when the threshold value of regulation is set to " 38 ", the part of the black surrounded by dotted line R shown in (3) of Fig. 8 is the part that pixel value all exceeds " 38 ".Further, the bank note as differentiation object, when there is the part of the black surrounded by this dotted line R, is determined as damage certificate by judegment part 12h.

In addition, in the explanation employing same figure, illustrate the example that Density Detection wave filter f6 or f7 is the wave filter of rectangle, but carrying out compared with stained shape or size, when focusing on the complete damage differentiation of closeness, also can use circular Gaussian filter.

Turn back to the explanation of Fig. 2, storage part 13 is described.Storage part 13 is the storage parts be made up of the memory device such as hard disk drive or nonvolatile memory, stores set information 13a, bank of filters 13b, by direction plane marginal point amount 13c, by direction dictionary image 13d, discrimination standard information 13e.

Set information 13a is such as the information relevant with initial setting such as quantization number, rotating deviation uptake.Bank of filters 13b is the group having pre-defined the various wave filters used in edge direction check processing (with reference to Fig. 3) or remaining edge analysis process (with reference to Fig. 8), by edge direction test section 12c and the reference of 12g institute of remaining edge analysis portion.

Be the marginal point amount of each marginal point in all directions plane of correspondence that N has opened certificate image by direction plane marginal point amount 13c, registered by efficient frontier detection unit 12d and upgrade.By the set that direction dictionary image 13d is by direction dictionary image, registered by expansion process portion 12e by quantized each direction.In addition, when comparing with differentiation object images, by the reference of 12f institute of remaining edge extracting portion.Discrimination standard information 13e is the information relevant with final complete threshold value of damaging the regulation in differentiating, by the reference of judegment part 12h institute.

In addition, set information 13a, bank of filters 13b, discrimination standard information 13e are predefined definition information, but also can suitably change in the application of bank note discriminating gear 10.

Then, use Fig. 9 that the treatment step of the process that the bank note discriminating gear 10 of embodiment 1 performs is described.Fig. 9 is the process flow diagram representing the treatment step that the bank note discriminating gear 10 of embodiment 1 performs.

As shown in Figure 9, configuration part 12a carries out the initial setting (step S101) of quantization number, rotating deviation uptake etc.Then, when generating dictionary image, view data obtaining section 12b obtains the complete certificate image (step S102) that N opens bank note.

Further, bank note discriminating gear 10 carries out N having been opened certificate image generates the dictionary image of each edge direction dictionary Computer image genration process (step S103) as input data.In addition, Figure 10 is used to describe the treatment step of this dictionary Computer image genration process later.

Further, when the complete damage of carrying out bank note differentiates, view data obtaining section 12b obtains the image (step S104) as the bank note differentiating object.

Further, bank note discriminating gear 10 carries out differentiating that the complete damage of object images as input data differentiates process (step S105).In addition, Figure 11 is used to describe this complete treatment step damaging differentiation process later.

Further, bank note discriminating gear 10 determines whether do not have the next bank note (step S106) differentiating object, when not having the bank note of next differentiation object (step S106: yes), ends process.On the other hand, when not meeting the decision condition of step S106 (step S106: no), bank note discriminating gear 10 repeat step S104 after process.

In addition, in the application of reality, also separately can carry out the stage (part of being surrounded by dotted line F1 with reference to Fig. 9) of dictionary Computer image genration and complete stage (part of being surrounded by dotted line F2 with reference to Fig. 9) of damaging differentiation.

Then, use Figure 10 that the detailed treatment step of the dictionary Computer image genration process (the step S103 with reference to Fig. 9) shown in Fig. 9 is described.Figure 10 is the process flow diagram of the treatment step representing the dictionary Computer image genration process that the bank note discriminating gear 10 of embodiment 1 performs.In addition, alleged here dictionary Computer image genration process is corresponding to the process of having opened from N till the dictionary image of each edge direction of certificate Computer image genration.

As shown in Figure 10, edge direction test section 12c has opened certificate image smoothingization (step S201) to N, has opened certificate image detect each marginal point and its direction (step S202) to N.

Then, efficient frontier detection unit 12d has opened each marginal point of certificate image to N, and the marginal point amount (storage part 13 by direction plane marginal point amount 13c) in all directions plane of correspondence is carried out increase counting (step S203).Then, the pixel of marginal point the amount more than decision threshold obtaining regulation in all directions plane is preserved (step S204) as efficient frontier point by efficient frontier detection unit 12d in all directions plane.

Then, each efficient frontier point preserved in all directions plane is expanded (step S205) with the swell increment of regulation by expansion process portion 12e.Then, all directions plane as (step S206) after being stored in storage part 13 by direction dictionary image 13d, ends process by expansion process portion 12e.

Then, the complete detailed treatment step damaging differentiation process (the step S105 with reference to Fig. 9) shown in Figure 11 key diagram 9 is used.Figure 11 is the complete process flow diagram damaging the treatment step differentiating process representing that the bank note discriminating gear 10 of embodiment 1 performs.In addition, here alleged complete damage differentiate process correspond to differentiate object images and dictionary image compare and analyze and differentiate as the complete damage of the bank note differentiating object till process.

As shown in figure 11, edge direction test section 12c, to differentiation object images smoothingization (step S301), detects each marginal point and its direction (step S302) of differentiating object images.

Then, remaining edge extracting portion 12f will differentiate that all directions plane of each marginal point to correspondence of object images stores (step S303).Then, remaining edge extracting portion 12f obtains all directions plane and each difference (step S304) by each dictionary image corresponding to direction dictionary image 13d, and synthesizes each difference and extract remaining marginal point (step S305).

Then, remaining edge analysis portion 12g (step S306) after eliminating the isolated point in remaining marginal point, analyzes the density (step S307) of remaining marginal point.

Further, judegment part 12h judges whether the density of the remaining marginal point that remaining edge analysis portion 12g analyzes is less than the threshold value (step S308) of regulation.In addition, about the threshold value reference discrimination standard information 13e of regulation.

Further, when density is less than the threshold value of regulation (step S308: yes), the bank note as differentiation object has been determined as certificate (step S309) by judegment part 12h, and ends process.

On the other hand, when not meeting the decision condition of step S308 (step S308: no), the bank note as differentiation object is judged to damage certificate (step S310) by judegment part 12h, and ends process.

As mentioned above, in embodiment 1, following formation bank note discriminating gear: the initial setting of quantization etc. is carried out in configuration part, view data obtaining section generates multiple complete certificate view data when the generation of dictionary image, and differentiation object image data is generated when the complete damage of bank note differentiates, edge direction test section detects marginal point and its direction of the view data generated, efficient frontier detection unit is determined to be the efficient frontier point of the structural element of dictionary image, expansion process portion carries out the expansion of ormal weight to efficient frontier point, remaining edge extracting portion extracts remaining marginal point from differentiation object images, the density of remaining marginal point is detected in remaining edge analysis portion, the complete damage that judegment part carries out bank note based on the density of the remaining marginal point detected differentiates.Therefore, even there is multifarious situation on the pattern or quality of paper under, also high precision can be carried out and most suitable complete damage differentiation.

In addition, in the above embodiments 1, describe the situation that bank note discriminating gear generates the dictionary image each marginal point having been carried out to the expansion of ormal weight, but, bank note discriminating gear also can generate dictionary image based on the periphery edge number of each pixel here.Therefore, below, the embodiment periphery edge number based on each pixel being generated the bank note discriminating gear of dictionary image is described as embodiment 2.

Embodiment 2

Figure 12 is the figure of the summary of the dictionary Computer image genration of the bank note discriminating gear representing embodiment 2.As shown in figure 12, the principal character of the bank note discriminating gear of embodiment 2 is, to each pixel counts periphery edge number of all directions plane, and based on this periphery edge number statistic and generate the dictionary image in each direction.

In addition, alleged here periphery edge refers to the marginal point comprised in the nearby scope of the regulation of a certain pixel.

Specifically, the bank note discriminating gear of embodiment 2 has opened each of certificate image to the input data of dictionary image and N, by each pixel counts periphery edge number ((1) with reference to Figure 12) in all directions plane.Here, in the counting of periphery edge number, can use such as in 9 × 9 pixels coefficient be all the counting wave filter (not shown) of 1.

In addition, represent the periphery edge number of each pixel in (1) of Figure 12, opened to the 1st ~ the N all directions plane that certificate image distinguishes direction plane group 1 ~ N that corresponding periphery edge counts, carried out the example stored by each pixel.This example the content of the dictionary Computer image genration process of the bank note discriminating gear of non-limiting examples 2, but be described as prerequisite following.

Further, the bank note discriminating gear of embodiment 2 calculates to the periphery edge number of counting the statistic ((2) with reference to Figure 12) that N opens (direction plane group 1 ~ N).

Further, the bank note discriminating gear of embodiment 2, based on the statistic calculated, determines benchmark pixel value to each pixel in all directions plane, and all directions plane of this benchmark pixel value to dictionary image is stored ((3) with reference to Figure 12).In addition, Figure 13 is used to describe the decision of this benchmark pixel value later.

Then, use Figure 13 that the structure of the bank note discriminating gear 10a of embodiment 2 is described.Figure 13 is the block scheme of the structure of the bank note discriminating gear 10a representing embodiment 2.In addition, in fig. 13, give identical label for the structural element identical with the bank note discriminating gear 10 of the embodiment 1 shown in Fig. 2, below, omit the explanation of the structural element repeated with the above embodiments 1.

As shown in figure 13, except also comprise the point of periphery edge count section 12i in control part 12 except, the difference of the bank note discriminating gear 10a of embodiment 2 and the bank note discriminating gear 10 of the above embodiments 1 is also, replaces efficient frontier detection unit 12d and expansion process portion 12e(with reference to Fig. 2) and possess the point of statistical treatment portion 12j.In addition, be with the difference of the bank note discriminating gear 10 of the above embodiments 1, in storage part 13, replace the point possessing periphery edge counting plane 13f by direction plane marginal point amount 13c.

Periphery edge count section 12i is the handling part be handled as follows: the number each pixel being counted in the nearby scope of regulation to the marginal point that edge direction test section 12c detects.

In addition, periphery edge count section 12i is also the handling part be handled as follows: by the periphery edge number of each pixel of all directions plane of counting, be stored into periphery edge counting plane 13f when dictionary Computer image genration, output to remaining edge extracting portion 12f when complete damage differentiates.

Statistical treatment portion 12j is the handling part be handled as follows: to the periphery edge number of each pixel of all directions plane that periphery edge count section 12i counts, and uses plane 13f with reference to periphery edge counting, calculates the statistic that N has opened certificate image.This statistic comprises maximal value, mean value, variance yields, standard deviation etc.

In addition, statistical treatment portion 12j is also the handling part be handled as follows: based on the statistical value calculated, and determines the benchmark pixel value of each pixel of all directions plane.In addition, about the decision of this benchmark pixel value, the various value and parameter that comprise in statistic can be combined in.

Here, the maximal value comprised in statistic is set to V, mean value is set to μ, standard deviation is set to σ, parameter is set to α.Such as, the maximal value of Corpus--based Method amount, can be set to benchmark pixel value " V+ α ".In addition, also can the mean value of Corpus--based Method amount and standard deviation, benchmark pixel value is set to " μ+α σ ".In addition, according to the discrimination precision etc. of the pattern of bank note or requirement, the calculating formula of this benchmark pixel value can suitably be changed.

In addition, the all directions plane of the benchmark pixel value of each pixel of all directions plane of decision to dictionary image stores by statistical treatment portion 12j, and all directions plane after storing is stored into storage part 13 as by direction dictionary image 13d.

Then, use Figure 14 that the treatment step of the dictionary Computer image genration process that the bank note discriminating gear 10a of embodiment 2 performs is described.Figure 14 is the process flow diagram of the treatment step representing the dictionary Computer image genration process that the bank note discriminating gear 10a of embodiment 2 performs.

As shown in figure 14, edge direction test section 12c has opened certificate image smoothingization (step S401) to N, has opened certificate image detect each marginal point and its direction (step S402) to N.

Then, periphery edge count section 12i has opened certificate image to N, by the number (periphery edge number) (step S403) of the marginal point in the nearby scope of each pixel counts regulation of all directions plane of direction plane group 1 ~ N.

Further, statistical treatment portion 12j, to the periphery edge number of each pixel that periphery edge count section 12i counts, calculates the statistic (step S404) that N opens (direction plane group 1 ~ N).Further, statistical treatment portion 12j, based on the statistic calculated, determines benchmark pixel value (step S405) to each pixel of all directions plane.

Further, the benchmark pixel value of each pixel that determines as (step S406) after being stored into storage part 13 by direction dictionary image 13d, ends process by statistical treatment portion 12j.

Then, complete that the bank note discriminating gear 10a of embodiment 2 performs damages the treatment step differentiating process to use Figure 15 to illustrate.Figure 15 is the complete process flow diagram damaging the treatment step differentiating process representing that the bank note discriminating gear 10a of embodiment 2 performs.

As shown in figure 15, edge direction test section 12c, to differentiation object images smoothingization (step S501), detects each marginal point and its direction (step S502) of differentiating object images.

Then, each marginal point, to differentiation object images, is stored into corresponding all directions plane (step S503) by periphery edge count section 12i.Then, each pixel of periphery edge count section 12i to all directions plane carries out the counting of periphery edge number, and count value is set to the pixel value (step S504) of each pixel of all directions plane.

And, remaining edge extracting portion 12f by each pixel compare all directions plane with by each dictionary image (step S505) corresponding to direction dictionary image 13d, and the pixel of the pixel value exceeding benchmark pixel value is set to the remaining marginal point (step S506) in all directions plane.Further, remaining edge extracting portion 12f synthesizes the remaining marginal point (step S507) of all directions plane.

Then, after the isolated point of remaining edge analysis portion 12g in the remaining marginal point of removing (step S508), the density (step S509) of remaining marginal point is analyzed.

Further, judegment part 12h judges whether the density of the remaining marginal point that remaining edge analysis portion 12g analyzes is less than the threshold value (step S510) of regulation.In addition, about the threshold value of regulation, with reference to discrimination standard information 13e.

Further, when density is less than the threshold value of regulation (step S510: yes), the bank note as differentiation object has been determined as certificate (step S511) by judegment part 12h, and ends process.

On the other hand, when not meeting the decision condition of step S510 (step S510: no), the bank note as differentiation object is judged to damage certificate (step S512) by judegment part 12h, and ends process.

As mentioned above, in example 2, due to count based on periphery edge count section, the statistic of the periphery edge number of each pixel of all directions plane, statistical treatment portion determines the benchmark pixel value of each pixel, so the element of the such macroscopic view of statistic can also be added while element based on the such microcosmic of the periphery edge number of each pixel, even there is multifarious situation on the pattern or quality of paper under, also high precision can be carried out and most suitable complete damage differentiation.

In addition, in each above-mentioned embodiment, mainly describe the complete situation of damaging differentiation of carrying out bank note, but be not particularly limited as the complete paper damaging the object differentiated.Such as, also can be exchange ticket or bank's booklet etc.

In addition, in each above-mentioned embodiment, mainly describe the complete situation of damaging differentiation of carrying out paper, but the kind that also method of the present invention can be applied to paper differentiates.Now, if such as bank note, then each cash kind of pressing bank note generates dictionary image, carries out differentiating comparing (that is, cash kind for several times) of object images and dictionary image to whole cash kind.Further, can judge that the cash kind of the dictionary image when the density of remaining marginal point becomes below setting is as the cash kind differentiating object images.

Utilizability in industry

As mentioned above, even paper discriminating gear of the present invention and paper method of discrimination exist on the pattern or quality of paper also want to carry out high precision in multifarious situation and most suitable complete damage differentiates and kind differentiates when useful, be specially adapted to the device differentiating the paper that the circulation of bank note etc. is high.

Label declaration

1 complete certificate image

5 differentiate object images

10 bank note discriminating gears

10a bank note discriminating gear

11 line sensor portions

12 control parts

12a configuration part

12b view data obtaining section

12c edge direction test section

12d efficient frontier detection unit

12e expansion process portion

The remaining edge extracting portion of 12f

The remaining edge analysis portion of 12g

12h judegment part

12i periphery edge count section

12j statistical treatment portion

13 storage parts

13a set information

13b bank of filters

13c presses direction plane marginal point amount

13d is by direction dictionary image sets

13e discrimination standard information

13f periphery edge counting plane

Claims

1. a bank note judgement system, based on bank note photographed images and differentiate bank note, it is characterized in that, comprising:

By direction dictionary image generating member, when the edge direction as concentration vergence direction in the marginal position detecting the position of the deep or light marked change as image from the photographed images and study image of complete certificate bank note and this marginal position, by multiple by direction plane what prepare the edge direction scope of each regulation, this marginal position is registered in by direction plane described in the scope belonging to this edge direction, generates described by direction dictionary image respectively;

Memory unit, store by described by direction dictionary image generating member generate described in by direction dictionary image;

Line sensor, is arranged on transfer path, obtains and is differentiated the photographed images of bank note by what transmit;

By direction input picture generating unit, when detecting the edge direction in marginal position and this marginal position from the banknote image obtained by described line sensor, by being separated this marginal position based on this edge direction to each described edge direction scope, generate by direction input picture;

Remaining marginal point extracting parts, from described in relevant with same described edge direction scope by deduct the input picture of direction corresponding described in by the difference processing of direction dictionary image, extract remaining marginal point;

Remaining edge analysis parts, remove the isolated point in the remaining marginal point extracted at remaining edge extracting parts, thus generate by the density image of the density of remaining marginal point in the size of regulation remained; And

Judgement part, based on the described density image generated by described remaining edge analysis parts, differentiates the complete damage of bank note.

2. bank note judgement system as claimed in claim 1, is characterized in that,

Described when belonging in the scope of regulation from the boundary line of adjacent two described edge direction scopes by the described edge direction of direction dictionary image generating member in the described marginal position detected, register this marginal position to described in these two edge direction scopes are corresponding respectively by the corresponding position of direction plane.

3. bank note judgement system as claimed in claim 1, is characterized in that,

The described described study image using regulation number by direction dictionary image generating member, and will the pixel of the described marginal position of more than stated number be have registered as efficient frontier region.

4. bank note judgement system as claimed in claim 3, is characterized in that,

Described by direction dictionary image generating member to each described expansion process of being undertaken described marginal position or described efficient frontier region being expanded to neighboring pixel by direction dictionary image.

5. bank note judgement system as claimed in claim 4, is characterized in that,

Describedly to be undertaken the described expansion process preferential with the described described edge direction scope corresponding by direction dictionary image by direction dictionary image generating member.

6. bank note judgement system as claimed in claim 1, is characterized in that, also comprise:

Periphery edge counter block, counts the number being assigned with the pixel of described marginal position comprised in the nearby scope of the described regulation by each pixel in direction plane; And

Normalized set parts, calculate the statistic of the described number counted by described periphery edge counter block being added up multiple described study image,

Described by direction dictionary image generating member when going out described statistic by described normalized set component computes, by based on this statistic, by in the dictionary image of direction, benchmark pixel value is set to each pixel described, generate described by direction dictionary image.

7. a bank note discriminating gear, based on bank note photographed images and differentiate bank note, it is characterized in that, comprising:

Directivity dictionaries store parts, when the edge direction as concentration vergence direction in addition in the marginal position detecting the position of the deep or light marked change as image from the photographed images and study image of complete certificate bank note and this marginal position, by multiple by direction plane what prepare the edge direction scope of regulation, this marginal position is registered in the scope belonging to this edge direction corresponding described in by direction plane stores;

By direction input picture generating unit, when detecting the edge direction in marginal position and this marginal position from the banknote image obtained by described line sensor, by being separated this marginal position based on this edge direction to each described edge direction scope, thus generate by direction input picture;

Remaining marginal point extracting parts, about only described by direction input picture in the described marginal position that exists, to whole described edge direction scope carry out from described in relevant with same described edge direction scope by deduct the input picture of direction corresponding described in by the difference processing of direction dictionary image, thus extract remaining marginal point;

Remaining edge analysis parts, remove the isolated point in the remaining marginal point extracted at described remaining edge extracting parts, thus generate by the density image of the density of remaining marginal point in the size of regulation remained; And

8. a bank note method of discrimination, based on bank note photographed images and differentiate bank note, it is characterized in that, comprising:

By direction dictionary Computer image genration step, when the edge direction as concentration vergence direction in the marginal position detecting the position of the deep or light marked change as image from the photographed images and study image of complete certificate bank note and this marginal position, by multiple by direction plane what prepare the edge direction scope of each regulation, this marginal position is registered in corresponding with this edge direction described in by the corresponding position of direction dictionary image, generate described by direction dictionary image respectively;

Storing step, store described generation by direction dictionary image;

Input picture obtains step, obtains by the photographed images differentiating bank note by the line sensor arranged on transfer path;

By direction input picture generation step, when from when being obtained in banknote image that step obtains the edge direction detected in marginal position and this marginal position by described input picture, by being separated this marginal position based on this edge direction to each described edge direction scope, thus generate by direction input picture;

Remaining marginal point extraction step, by carry out by from described in relevant with same described edge direction scope by deduct in the input picture of direction corresponding described in by the difference processing of direction dictionary image, extract remaining marginal point;

Density image generation step, removes the isolated point in the remaining marginal point extracted in described remaining edge extracting step, thus generates by the density image of the density of remaining marginal point in the size of regulation remained; And

Discriminating step, based on the described density image generated by described density image generation step, differentiates the complete damage of bank note.

9. bank note method of discrimination as claimed in claim 8, is characterized in that,

Described by direction dictionary Computer image genration step to each described expansion process of being undertaken described marginal position or efficient frontier region being expanded to neighboring pixel by direction dictionary image.

10. bank note method of discrimination as claimed in claim 9, is characterized in that,

Describedly to be undertaken the described expansion process preferential with the described described edge direction scope corresponding by direction dictionary image by direction dictionary Computer image genration step.