CN101014978A - Image checking device, image checking method, and image checking program - Google Patents

Abstract

A correlation value image is created from an input image and a template image. The correlation value image is divided into positive and negative correlation value images depending on whether the pixel value is above a threshold or not, and the template image is divided into positive and negative template images depending on whether the pixel value is above a threshold or not. Positive and negative division correlation images are created by combining the positive and negative correlation value images and the positive and negative template images and used for check and judgment. For this input image and the template image, polar coordinate transformed input and template images are used.

Description

Image is checked device, image checking method and image check program

Technical field

Thereby the present invention relates to check device, image checking method and image check program, thereby particularly can be by carrying out checking the image that improves the rate of checking and checking device, image checking method and image check program of input picture and template image efficiently by between a plurality of template images of the input picture of checking object thing and registered in advance, the feature of image being compared the image that image is checked.

Background technology

In the past, known will be by CCD (Charge Coupled Device, charge-coupled device (CCD)) camera etc. is taken the currency of gathering and the input picture that obtains and the template image of registered in advance are checked, and the image that the true and false of this currency is judged is checked device.

For example, disclose a kind of image in the patent documentation 1 and checked technology, the template image of the input picture of coin and coin compared calculate correlation, in the part in the general image that becomes the contrast object more than the regulation area, surpassed at this correlation under the situation of threshold value, the coin of input picture has been judged to be real coin.

Patent documentation 1: the spy opens the 2003-187289 communique

Summary of the invention

But, use under the situation of the conventional art, there is following problem, promptly check owing to only using the correlation that surpasses threshold value to carry out image, so because the image when the checking of input picture and template image taken place departs from or the noise of accompanying image conversion etc., correlation is got under the situation of low relatively value, although the coin of input is real coin for example, the result also is judged to be and counterfeits coins, and is difficult to improve the rate of checking that image is checked.

In addition, this problem points is not the problem that only takes place in the image of coin is checked, for example, the image of bank note check or FA (Factory Automation, factory automation) in waiting article or the image of product check in the problem of generation too.

The present invention finishes in view of above-mentioned problem (problem points), and its purpose is to provide a kind of image that can improve the article beyond currency and the currency to check precision and the image that improves the rate of checking that image checks is checked device, image checking method and image check program.

In order to solve above-mentioned problem and to reach purpose, the inventive images of technical scheme 1 is checked device and is compared by the feature to image between a plurality of template images of the input picture of checking object thing and registered in advance, thereby image is checked, it is characterized in that, this image is checked device and is comprised: correlation separation of images parts, whether generate the correlation image by described input picture and described template image, be more than the threshold value and with this correlation separation of images to be positive correlation value image and negative correlation value image according to pixel value; Whether the template image separating component is more than the threshold value described template image to be separated into positive template image and negative norm plate image according to pixel value; Positive and negative separation associated picture generates parts, and the combination by described positive correlation value image and described negative correlation value image and described positive template image and described negative norm plate image generates a plurality of positive and negative separation associated pictures; And check judging part, use described positive and negative separation associated picture to check judgement.

In addition, the inventive images of technical scheme 2 is checked device in the invention of technical scheme 1, it is characterized in that, described positive and negative separation associated picture generates parts and generates positive feature regional images and negative feature regional images, described positive feature regional images with to long-pending calculating of the described positive correlation value image of each pixel and described positive template image and the value that obtains as pixel value, described negative feature regional images with to long-pending calculating of the described negative correlation value image of each pixel and described positive template image and the value that obtains as pixel value.

In addition, the inventive images of technical scheme 3 is checked device in the invention of technical scheme 1, it is characterized in that, described positive and negative separation associated picture generates parts and generates positive background area image and negative background area image, described positive background area image with to long-pending calculating of the described positive correlation value image of each pixel and described negative norm plate image and the value that obtains as pixel value, described negative background area image with to long-pending calculating of the described negative correlation value image of each pixel and described negative norm plate image and the value that obtains as pixel value.

In addition, the inventive images of technical scheme 4 is checked device in the invention of technical scheme 1, it is characterized in that, described positive and negative separation associated picture generates parts and generates positive feature regional images, negative feature regional images, positive background area image and negative background area image, described positive feature regional images with to long-pending calculating of the described positive correlation value image of each pixel and described positive template image and the value that obtains as pixel value, described negative feature regional images with to long-pending calculating of the described negative correlation value image of each pixel and described positive template image and the value that obtains as pixel value, described positive background area image with to long-pending calculating of the described positive correlation value image of each pixel and described negative norm plate image and the value that obtains as pixel value, described negative background area image with to long-pending calculating of the described negative correlation value image of each pixel and described negative norm plate image and the value that obtains as pixel value.

In addition, the inventive images of technical scheme 5 is checked device in the invention of technical scheme 2,3 or 4, it is characterized in that, described positive and negative separation associated picture generates parts this surrounding pixel of gazing at the respective pixel of pixel correspondence in the positive region image of gazing at pixel and the described positive correlation value image generation of use in the negative region image that uses described negative correlation value image generation is compared, gaze at greater than this at the pixel value of at least one this surrounding pixel under the situation of pixel value of pixel, carry out this is gazed at the expansion process that pixel moves to this respective pixel.

In addition, the inventive images of technical scheme 6 is checked device in the invention of technical scheme 1, it is characterized in that, described input picture and described template image are to handle the edge image that has carried out image transformation by the edge extracting that uses the edge extracting operator.

In addition, the inventive images of technical scheme 7 is checked device in the invention of technical scheme 6, it is characterized in that, described edge image is that the edge strength at the edge that will extract carries out standardization (normalize) and the standardized images that obtains.

In addition, the inventive images of technical scheme 8 is checked device in the invention of technical scheme 1, it is characterized in that, described template image is each the individual image for described checking object thing to be averaged and the average image that obtains.

In addition, the inventive images of technical scheme 9 is checked device in the invention of technical scheme 1, it is characterized in that, described correlation image be with the correlation with each pixel of described input picture or described template image carry out standardization and the standardization correlation that obtains as the image of pixel value.

In addition, the inventive images of technical scheme 10 is checked device in the invention of technical scheme 1, it is characterized in that, described check that judging part carries out to described positive and negative separation associated picture that piece is cut apart and the summation of calculating the pixel value in each piece as the piece value, thereby check judgement by whole described positive and negative separation associated pictures are carried out that the value of checking is calculated by the long-pending Calais mutually of this piece value and weighting coefficient.

In addition, the inventive images of technical scheme 11 is checked device in the invention of technical scheme 1, it is characterized in that, describedly checks judging part calculates described weighting coefficient by linear discriminant analysis value.

In addition, the inventive images of technical scheme 12 is checked device in the invention of technical scheme 1, it is characterized in that, described checking object thing is a currency.

In addition, the inventive images checking method of technical scheme 13 compares by the feature to image between a plurality of template images of the input picture of checking object thing and registered in advance, thereby image is checked, it is characterized in that, this image checking method comprises: correlation separation of images step, whether generate the correlation image by described input picture and described template image, be more than the threshold value and with this correlation separation of images to be positive correlation value image and negative correlation value image according to pixel value; Whether the template image separating step is more than the threshold value described template image to be separated into positive template image and negative norm plate image according to pixel value; Positive and negative separation associated picture generates step, and the combination by described positive correlation value image and described negative correlation value image and described positive template image and described negative norm plate image generates a plurality of positive and negative separation associated pictures; And check determination step, use described positive and negative separation associated picture to check judgement.

In addition, the inventive images check program of technical scheme 14 compares by the feature to image between a plurality of template images of the input picture of checking object thing and registered in advance, thereby image is checked, it is characterized in that, this image check program makes computing machine carry out following steps: correlation separation of images step, whether generate the correlation image by described input picture and described template image, be more than the threshold value and with this correlation separation of images to be positive correlation value image and negative correlation value image according to pixel value; Whether the template image separating step is more than the threshold value described template image to be separated into positive template image and negative norm plate image according to pixel value; Positive and negative separation associated picture generates step, and the combination by described positive correlation value image and described negative correlation value image and described positive template image and described negative norm plate image generates a plurality of positive and negative separation associated pictures; And check determination step, use described positive and negative separation associated picture to check judgement.

In addition, the inventive images of technical scheme 15 is checked device, compare by feature between a plurality of template images of the input picture of circular object and registered in advance image, thereby image is checked, it is characterized in that, this image is checked device and is comprised: polar coordinate transform image production part spare, described input picture and described template image are being carried out on the basis of polar coordinate transform, generating ρ-θ input picture and ρ-θ template image after rotation with two images departs from correction; Whether correlation separation of images parts generate the correlation image by described ρ-θ input picture and described ρ-θ template image, be more than the threshold value and with this correlation separation of images to be positive correlation value image and negative correlation value image according to pixel value; Whether the template image separating component is more than the threshold value described ρ-θ template image to be separated into positive template image and negative norm plate image according to pixel value; Positive and negative separation associated picture generates parts, and the combination by described positive correlation value image and described negative correlation value image and described positive template image and described negative norm plate image generates a plurality of positive and negative separation associated pictures; And check judging part, use described positive and negative separation associated picture to check judgement.

In addition, the inventive images of technical scheme 16 is checked device in the invention of technical scheme 15, it is characterized in that, described positive and negative separation associated picture generates parts and generates positive feature regional images and negative feature regional images, described positive feature regional images with to long-pending calculating of the described positive correlation value image of each pixel and described positive template image and the value that obtains as pixel value, described negative feature regional images with to long-pending calculating of the described negative correlation value image of each pixel and described positive template image and the value that obtains as pixel value.

In addition, the inventive images of technical scheme 17 is checked device in the invention of technical scheme 15, it is characterized in that, described positive and negative separation associated picture generates parts and generates positive background area image and negative background area image, described positive background area image with to long-pending calculating of the described positive correlation value image of each pixel and described negative norm plate image and the value that obtains as pixel value, described negative background area image with to long-pending calculating of the described negative correlation value image of each pixel and described negative norm plate image and the value that obtains as pixel value.

In addition, the inventive images of technical scheme 18 is checked device in the invention of technical scheme 15, it is characterized in that, described positive and negative separation associated picture generates parts and generates positive feature regional images, negative feature regional images, positive background area image and negative background area image, described positive feature regional images with to long-pending calculating of the described positive correlation value image of each pixel and described positive template image and the value that obtains as pixel value, described negative feature regional images with to long-pending calculating of the described negative correlation value image of each pixel and described positive template image and the value that obtains as pixel value, described positive background area image with to long-pending calculating of the described positive correlation value image of each pixel and described negative norm plate image and the value that obtains as pixel value, described negative background area image with to long-pending calculating of the described negative correlation value image of each pixel and described negative norm plate image and the value that obtains as pixel value.

In addition, the inventive images of technical scheme 19 is checked device in the invention of technical scheme 16,17 or 18, it is characterized in that, described positive and negative separation associated picture generates parts this surrounding pixel of gazing at the respective pixel of pixel correspondence in the positive region image of gazing at pixel and the described positive correlation value image generation of use in the negative region image that uses described negative correlation value image generation is compared, gaze at greater than this at the pixel value of at least one this surrounding pixel under the situation of pixel value of pixel, carry out this is gazed at the expansion process that pixel moves to this respective pixel.

In addition, the inventive images of technical scheme 20 is checked device in the invention of technical scheme 1, it is characterized in that described ρ-θ input picture and described ρ-θ template image are to handle the edge image that has carried out image transformation by the edge extracting that uses the edge extracting operator.

In addition, the inventive images of technical scheme 21 is checked device in the invention of technical scheme 20, it is characterized in that, described edge image is that the edge strength at the edge that will extract carries out standardization and the standardized images that obtains.

In addition, the inventive images of technical scheme 22 is checked device in the invention of technical scheme 1, it is characterized in that, described template image is each the individual image for described checking object thing to be averaged and the average image that obtains.

In addition, the inventive images of technical scheme 23 is checked device in the invention of technical scheme 1, it is characterized in that, described correlation image be with the correlation with each pixel of described ρ-θ input picture and described ρ-θ template image carry out standardization and the standardization correlation that obtains as the image of pixel value.

In addition, the inventive images of technical scheme 24 is checked device in the invention of technical scheme 1, it is characterized in that, described check that judging part carries out to described positive and negative separation associated picture that piece is cut apart and the summation of calculating the pixel value in each piece as the piece value, thereby check judgement by whole described positive and negative separation associated pictures are calculated the value of checking with the long-pending Calais mutually of this piece value and weighting coefficient.

In addition, the inventive images of technical scheme 25 is checked device in the invention of technical scheme 1, it is characterized in that, describedly checks judging part calculates described weighting coefficient by linear discriminant analysis value.

In addition, the inventive images of technical scheme 26 is checked device in the invention of technical scheme 1, it is characterized in that described ρ-θ input picture or described ρ-θ template image are parallel to be moved described polar coordinate transform image production part spare by making, thereby the rotation of proofreading and correct two images departs from.

In addition, the inventive images of technical scheme 27 is checked device in the invention of technical scheme 1, it is characterized in that, described circular object is a coin.

In addition, the inventive images checking method of technical scheme 28 compares by the feature to image between a plurality of template images of the input picture of circular object and registered in advance, thereby image is checked, it is characterized in that, this image checking method comprises: the polar coordinate transform image generates step, described input picture and described template image are being carried out on the basis of polar coordinate transform, generating ρ-θ input picture and ρ-θ template image after rotation with two images departs from correction; Whether correlation separation of images step generates the correlation image by described ρ-θ input picture and described ρ-θ template image, be more than the threshold value and with this correlation separation of images to be positive correlation value image and negative correlation value image according to pixel value; Whether the template image separating step is more than the threshold value described ρ-θ template image to be separated into positive template image and negative norm plate image according to pixel value; Positive and negative separation associated picture generates step, and the combination by described positive correlation value image and described negative correlation value image and described positive template image and described negative norm plate image generates a plurality of positive and negative separation associated pictures; And check determination step, use described positive and negative separation associated picture to check judgement.

In addition, the inventive images check program of technical scheme 29 compares by the feature to image between a plurality of template images of the input picture of circular object and registered in advance, thereby image is checked, it is characterized in that, this image check program makes computing machine carry out following steps: the polar coordinate transform image generates step, described input picture and described template image are being carried out on the basis of polar coordinate transform, generating ρ-θ input picture and ρ-θ template image after rotation with two images departs from correction; Whether correlation separation of images step generates the correlation image by described ρ-θ input picture and described ρ-θ template image, be more than the threshold value and with this correlation separation of images to be positive correlation value image and negative correlation value image according to pixel value; Whether the template image separating step is more than the threshold value described ρ-θ template image to be separated into positive template image and negative norm plate image according to pixel value; Positive and negative separation associated picture generates step, and the combination by described positive correlation value image and described negative correlation value image and described positive template image and described negative norm plate image generates a plurality of positive and negative separation associated pictures; And check determination step, use described positive and negative separation associated picture to check judgement.

Invention according to technical scheme 1, owing to generate the correlation image by input picture and template image, whether according to pixel value is more than the threshold value and with this correlation separation of images to be positive correlation value image and negative correlation value image, whether according to pixel value is more than the threshold value template image to be separated into positive template image and negative norm plate image, combination by positive correlation value image and negative correlation value image and positive template image and negative norm plate image, generate a plurality of positive and negative separation associated pictures, use positive and negative separation associated picture to check judgement, so not only use the relevant high part of input picture and template image but also use the correlation of relevant low part, while not only uses the characteristic of template image but also uses background portion to assign to carry out image and check, thereby play and to carry out the high image of precision and check, and can improve the effect of the rate of checking of image.

In addition, invention according to technical scheme 2, generate positive feature regional images and negative feature regional images owing to constitute, positive feature regional images with value that long-pending calculating of the positive correlation value image of each pixel and positive template image obtained as pixel value, negative feature regional images with value that long-pending calculating of the negative correlation value image of each pixel and positive template image obtained as pixel value, so the area image of feature occurred and checked at the area image that feature does not appear in the part that feature should occur by using in the part that feature should occur, thereby play and to carry out high-precision image and check, and can improve the effect of the rate of checking of image.

In addition, invention according to technical scheme 3, generate positive background area image and negative background area image owing to constitute, positive background area image with value that long-pending calculating of the positive correlation value image of each pixel and negative norm plate image obtained as pixel value, negative background area image with value that long-pending calculating of the negative correlation value image of each pixel and negative norm plate image obtained as pixel value, so have the area image of background and do not exist the area image of background to check in the part that should have background in the part that should have background by using, thereby play and to carry out high-precision image and check, and can improve the effect of the rate of checking of image.

In addition, invention according to technical scheme 4, generate positive feature regional images owing to constitute, negative feature regional images, positive background area image and negative background area image, positive feature regional images with value that long-pending calculating of the positive correlation value image of each pixel and positive template image obtained as pixel value, negative feature regional images with value that long-pending calculating of the negative correlation value image of each pixel and positive template image obtained as pixel value, positive background area image with value that long-pending calculating of the positive correlation value image of each pixel and negative norm plate image obtained as pixel value, negative background area image with value that long-pending calculating of the negative correlation value image of each pixel and negative norm plate image obtained as pixel value, so by using the area image that has occurred feature in the part that feature should occur, the area image that does not occur feature in the part that feature should occur, there is the area image of background and do not exist the area image of background to check in the part that should have background in the part that should have background, thereby play and to carry out high-precision image and check, and can improve the effect of the rate of checking of image.

In addition, invention according to technical scheme 5, owing to this surrounding pixel of gazing at the respective pixel of pixel correspondence that constitutes in the positive region image of gazing at pixel and the generation of use positive correlation value image in the negative region image that uses the generation of negative correlation value image compares, gaze at greater than this at the pixel value of at least one this surrounding pixel under the situation of pixel value of pixel, carry out this is gazed at the expansion process that pixel moves to this respective pixel, thereby carry out high-precision image and check so play the influence that to get rid of the isolated point of following correlation value calculation, and can improve the effect of the rate of checking of image.

In addition, invention according to technical scheme 6, owing to constitute input picture and template image is to handle the edge image that has carried out image transformation by the edge extracting that uses the edge extracting operator, so play the characteristic of each image that extracts by contrast, check thereby can carry out high-precision image, and can improve the effect of the rate of checking of image.

In addition, invention according to technical scheme 7, owing to constituting edge image is that the edge strength at the edge that will extract carries out the standardized images that standardization obtains, thereby carry out high-precision image and check so play other influence of individual difference that to get rid of the checking object thing, and can improve the effect of the rate of checking of image.

In addition, invention according to technical scheme 8, owing to constituting template image is that each the individual image for described checking object thing is averaged the average image that obtains, even so play the individuality that has the checking object thing under the situation of intrinsic pattern, also can carry out high-precision image and check, and can improve the effect of the rate of checking of image.

In addition, invention according to technical scheme 9, owing to constituting the correlation image is to carry out standardization correlation that standardization the obtains image as pixel value with the correlation with each pixel of input picture or template image, carry out high-precision image and check so play the deviation that can suppress correlation, and can improve the effect of the rate of checking of image.

In addition, invention according to technical scheme 10, since constitute check that judging part carries out to positive and negative separation associated picture that piece is cut apart and the summation of calculating the pixel value in each piece as the piece value, thereby check judgement by whole positive and negative separation associated pictures being carried out the value of checking is calculated by the long-pending Calais mutually of this piece value and weighting coefficient, so play the weight that to adjust the zone that occurs feature easily and be difficult to occur the weight in the zone of feature, simultaneously by simplifying calculation procedure, can carry out efficiently image and check, and can improve the effect of the rate of checking of image.

In addition, invention according to technical scheme 11, owing to constitute and check judging part calculates described weighting coefficient by linear discriminant analysis value, and owing to the suitable weight coefficient that can access based on learning sample, so play and to carry out efficiently image and check, and can improve the effect of the rate of checking of image.

In addition, according to the invention of technical scheme 12,, can carry out efficiently image about checking of currency and check so play, and can improve the effect of the rate of checking of image because to constitute the checking object thing be currency.

In addition, invention according to technical scheme 13, owing to constitute by input picture and template image and generate the correlation image, whether according to pixel value is more than the threshold value and with this correlation separation of images to be positive correlation value image and negative correlation value image, whether according to pixel value is more than the threshold value template image to be separated into positive template image and negative norm plate image, combination by positive correlation value image and negative correlation value image and positive template image and negative norm plate image, generate a plurality of positive and negative separation associated pictures, use this positive and negative separation associated picture to check judgement, so not only use the relevant high part of input picture and template image but also use the correlation of relevant low part, while not only uses the characteristic of template image but also uses background portion to assign to carry out image and check, thereby play and to carry out the high image of precision and check, and can improve the effect of the rate of checking of image.

In addition, invention according to technical scheme 14, owing to constitute by input picture and template image and generate the correlation image, whether according to pixel value is more than the threshold value and with this correlation separation of images to be positive correlation value image and negative correlation value image, whether according to pixel value is more than the threshold value template image to be separated into positive template image and negative norm plate image, combination by positive correlation value image and negative correlation value image and positive template image and negative norm plate image, generate a plurality of positive and negative separation associated pictures, use this positive and negative separation associated picture to check judgement, so not only use the relevant high part of input picture and template image but also use the correlation of relevant low part, while not only uses the characteristic of template image but also uses background portion to assign to carry out image and check, and can carry out the effect that the high image of precision is checked thereby play.

In addition, invention according to technical scheme 15, owing to constitute input picture and template image carried out on the basis of polar coordinate transform, generation departs from ρ-θ input picture and ρ-θ template image after the correction with the rotation of two images, generate the correlation image by ρ-θ input picture and ρ-θ template image, whether according to pixel value is more than the threshold value and with this correlation separation of images to be positive correlation value image and negative correlation value image, whether according to pixel value is more than the threshold value ρ-θ template image to be separated into positive template image and negative norm plate image, combination by positive correlation value image and negative correlation value image and positive template image and negative norm plate image, generate a plurality of positive and negative separation associated pictures, use this positive and negative separation associated picture to check judgement, so not only use the relevant high part of input picture and template image but also use the correlation of relevant low part, while not only uses the characteristic of template image but also uses background portion to assign to carry out image and check, and can carry out the effect that the high image of precision is checked thereby play.

In addition, invention according to technical scheme 16, generate positive feature regional images and negative feature regional images owing to constitute, positive feature regional images with value that long-pending calculating of the positive correlation value image of each pixel and positive template image obtained as pixel value, negative feature regional images with value that long-pending calculating of the negative correlation value image of each pixel and positive template image obtained as pixel value, so the area image of feature occurred and carry out image at the area image that feature does not appear in the part that feature should occur and check by using, can carry out the effect that high-precision image is checked thereby play in the part that feature should occur.

In addition, invention according to technical scheme 17, generate positive background area image and negative background area image owing to constitute, positive background area image with value that long-pending calculating of the positive correlation value image of each pixel and negative norm plate image obtained as pixel value, negative background area image with value that long-pending calculating of the negative correlation value image of each pixel and negative norm plate image obtained as pixel value, so have the area image of background and do not exist the area image of background to check in the part that should have background by using, can carry out the effect that high-precision image is checked thereby play in the part that should have background.

In addition, invention according to technical scheme 18, generate positive feature regional images owing to constitute, negative feature regional images, positive background area image and negative background area image, positive feature regional images with value that long-pending calculating of the positive correlation value image of each pixel and positive template image obtained as pixel value, negative feature regional images with value that long-pending calculating of the negative correlation value image of each pixel and positive template image obtained as pixel value, positive background area image with value that long-pending calculating of the positive correlation value image of each pixel and negative norm plate image obtained as pixel value, negative background area image with value that long-pending calculating of the negative correlation value image of each pixel and negative norm plate image obtained as pixel value, so by using the area image that has occurred feature in the part that feature should occur, the area image that does not occur feature in the part that feature should occur, there is the area image of background and do not exist the area image of background to check in the part that should have background, can carry out the effect that high-precision image is checked thereby play in the part that should have background.

In addition, invention according to technical scheme 19, owing to this surrounding pixel of gazing at the respective pixel of pixel correspondence that constitutes in the positive region image of gazing at pixel and the generation of use positive correlation value image in the negative region image that uses the generation of negative correlation value image compares, gaze at greater than this at the pixel value of at least one this surrounding pixel under the situation of pixel value of pixel, carry out this is gazed at the expansion process that pixel moves to this respective pixel, so thereby play the influence that to get rid of the isolated point of following correlation value calculation and carry out the effect that high-precision image is checked.

In addition, invention according to technical scheme 20, owing to constitute ρ-θ input picture and ρ-θ template image is to handle the edge image that has carried out image transformation by the edge extracting that uses the edge extracting operator, so play the characteristic of each image that extracts by contrast, thereby can carry out the effect that high-precision image is checked.

In addition, invention according to technical scheme 21, owing to constituting edge image is that the edge strength at the edge that will extract carries out the standardized images that standardization obtains, so thereby play other influence of individual difference that can get rid of the checking object thing and carry out the effect that high-precision image is checked.

In addition, invention according to technical scheme 22, owing to constituting template image is that each the individual image for described checking object thing is averaged the average image that obtains, even so play the individuality that has the checking object thing under the situation of intrinsic pattern, also can carry out the effect that high-precision image is checked.

In addition, invention according to technical scheme 23, owing to constituting the correlation image is to carry out standardization correlation that standardization the obtains image as pixel value with the correlation with each pixel of ρ-θ input picture and ρ-θ template image, carries out the effect that high-precision image is checked so play the deviation that can suppress correlation.

In addition, invention according to technical scheme 24, since constitute check that judging part carries out to positive and negative separation associated picture that piece is cut apart and the summation of calculating the pixel value in each piece as the piece value, thereby, whole positive and negative separation associated pictures check judgement by being calculated the value of checking with the long-pending Calais mutually of this piece value and weighting coefficient, so play the weight that to adjust the zone that occurs feature easily and be difficult to occur the weight in the zone of feature, simultaneously by simplifying calculation procedure, can carry out the effect that image is efficiently checked.

In addition, invention according to technical scheme 25, owing to constitute and check judging part and calculate the value of described weighting coefficient, and owing to can access suitable weight coefficient, can carry out the effect that image is efficiently checked so play based on learning sample by linear discriminant analysis.

In addition, invention according to technical scheme 26, because ρ-θ input picture or ρ-θ template image is parallel to be moved by making to constitute polar coordinate transform image production part spare, thereby the rotation of proofreading and correct two images departs from, so thereby play and can cut down the calculated amount of following this correction and carry out the effect that high efficiency image is checked.

In addition, according to the invention of technical scheme 27, be coin owing to constitute circular object, can carry out the effect that high-precision image is checked so play about checking of currency.

In addition, invention according to technical scheme 28, owing to constitute input picture and template image carried out on the basis of polar coordinate transform, generation departs from ρ-θ input picture and ρ-θ template image after the correction with the rotation of two images, generate the correlation image by ρ-θ input picture and ρ-θ template image, whether according to pixel value is more than the threshold value and with this correlation separation of images to be positive correlation value image and negative correlation value image, whether according to pixel value is more than the threshold value ρ-θ template image to be separated into positive template image and negative norm plate image, combination by positive correlation value image and negative correlation value image and positive template image and negative norm plate image, generate a plurality of positive and negative separation associated pictures, use positive and negative separation associated picture to check judgement, so not only use the relevant high part of input picture and template image but also use the correlation of relevant low part, while not only uses the characteristic of template image but also uses background portion to assign to carry out image and check, and can carry out the effect that the high image of precision is checked thereby play.

In addition, invention according to technical scheme 29, owing to constitute input picture and template image carried out on the basis of polar coordinate transform, generation departs from ρ-θ input picture and ρ-θ template image after the correction with the rotation of two images, generate the correlation image by ρ-θ input picture and ρ-θ template image, whether according to pixel value is more than the threshold value and with this correlation separation of images to be positive correlation value image and negative correlation value image, whether according to pixel value is more than the threshold value ρ-θ template image to be separated into positive template image and negative norm plate image, combination by positive correlation value image and negative correlation value image and positive template image and negative norm plate image, generate a plurality of positive and negative separation associated pictures, use positive and negative separation associated picture to check judgement, so not only use the relevant high part of input picture and template image but also use the correlation of relevant low part, while not only uses the characteristic of template image but also uses background portion to assign to carry out image and check, and can carry out the effect that the high image of precision is checked thereby play.

Description of drawings

Fig. 1 is the functional-block diagram that the image of expression embodiment 1 is checked the structure of device.

Fig. 2 is used to illustrate that image shown in Figure 1 cuts out the key diagram of the processing summary of part.

Fig. 3 is the key diagram that is used for illustrating the Sobel operator that uses in edge extracting part shown in Figure 1.

Fig. 4 is the key diagram that is used to illustrate edge extracting shown in Figure 1 processing summary partly.

Fig. 5 is the key diagram that is used to illustrate the processing summary that the coupling of matching treatment part shown in Figure 1 is judged.

Fig. 6 is the relevant key diagram of judging the image of part input of positive and negative separation that is used to illustrate to embodiment 1.

Fig. 7 is used to illustrate the characteristic area that the positive and negative separation associated picture generating portion by embodiment 1 generates and the key diagram of background area.

Fig. 8 is the process flow diagram of the positive and negative separating treatment of standardization correlation image of embodiment 1.

Fig. 9 is the process flow diagram of the positive and negative separating treatment of template image of embodiment 1.

Figure 10 is the process flow diagram of treatment step of the positive and negative separation associated picture generating portion of expression embodiment 1.

Figure 11 is used to illustrate that the image corresponding with each zone shown in Figure 7 generates the key diagram of step.

Figure 12 is the key diagram that is used for illustrating the pattern mask (imagemask) that uses in the expansion process part of embodiment 1.

Figure 13 is the process flow diagram of the expansion process treatment step partly of expression embodiment 1.

Figure 14 is the key diagram that is used to illustrate the image that the expansion process by embodiment 1 partly generates.

Figure 15 is the key diagram that is used for illustrating that the piece of the image that the value of the checking calculating section at embodiment 1 uses is cut apart.

Figure 16 is the key diagram of modification that is used to illustrate the expansion process of embodiment 1.

Figure 17 is the key diagram of the pattern mask that is used for illustrating that modification shown in Figure 16 is used.

Figure 18 is the process flow diagram of the treatment step of the expansion process part in the expression modification shown in Figure 16.

Figure 19 is the functional-block diagram that the image of expression embodiment 2 is checked the structure of device.

Figure 20 is used to illustrate that image shown in Figure 19 cuts out the key diagram of the processing summary of part.

Figure 21 is the key diagram that is used for illustrating the Sobel operator that uses in edge extracting part shown in Figure 19.

Figure 22 is the key diagram that is used to illustrate edge extracting shown in Figure 19 processing summary partly.

Figure 23 is the key diagram of processing summary that is used to illustrate the polar coordinate transform of embodiment 2.

Figure 24 is the key diagram that is used to illustrate the processing summary of rotation angle test section shown in Figure 19.

Figure 25 is the key diagram that is used to illustrate each area image of embodiment 2.

Figure 26 is the process flow diagram of the positive and negative separating treatment of standardization correlation image of embodiment 2.

Figure 27 is the process flow diagram of the positive and negative separating treatment of template image of embodiment 2.

Figure 28 is the process flow diagram of treatment step of the positive and negative separation associated picture generating portion of expression embodiment 2.

Figure 29 is used to illustrate that the image corresponding with each zone shown in Figure 25 generates the key diagram of step.

Figure 30 is the key diagram that is used for illustrating the pattern mask that uses in the expansion process part of embodiment 2.

Figure 31 is the process flow diagram of the expansion process treatment step partly of expression embodiment 2.

Figure 32 is the key diagram that is used to illustrate the image that the expansion process by embodiment 2 partly generates.

Figure 33 is the key diagram that is used for illustrating that the piece of the image that the value of the checking calculating section of embodiment 2 uses is cut apart.

Figure 34 is the key diagram of modification that is used to illustrate the expansion process of embodiment 2.

Figure 35 is the key diagram that is used for illustrating the pattern mask that uses in modification shown in Figure 34.

Figure 36 is the process flow diagram of the treatment step of the expansion process part in the expression modification shown in Figure 34.

Symbol description

1 image is checked device

10 image importations

11 input pictures

20 images cut out part

21 horizontal direction projections

22 vertical direction projections

23 cut out image (reverse side)

24 cut out image (front)

30 edge extracting parts

30a Sobel operator (being used for calculated level direction edge)

30b Sobel operator (being used to calculate the vertical direction edge)

31 edge extracting images

32 edge standardized images (reverse side)

33 edge standardized images (front)

40 matching treatment parts

50 registered images storage areas

51 template images

51a t+ image

51b t-image

The relevant judgement of 100 positive and negative separation part

110 standardization correlation value calculation parts

111 standardization correlation images

111a r+ image

111b r-image

120 positive and negative separation associated picture generating portions

121 A+ area images

122 A-area images

123 B+ area images

124 B-area images

130 expansion process parts

130 a positive region pattern masks

130 b negative region pattern masks

130 c input picture masks

130 d template image masks

The 131 A+ area images that expanded

The 132 A-area images that expanded

The 133 B+ area images that expanded

The 134 B-area images that expanded

The 135 standardization correlation images that expanded

The 135 a r+ image that expanded

The 135 b r-image that expanded

140 value of checking calculating sections

Cut apart (A+ zone) for 141

Cut apart (A-zone) for 142

Cut apart (B+ zone) for 143

Cut apart (B-zone) for 144

201 images are checked device

210 image importations

211 input pictures

220 images cut out part

221 horizontal direction projections

222 vertical direction projections

223 cut out image

230 edge extracting parts

230a Sobel operator (being used for calculated level direction edge)

230b Sobel operator (being used to calculate the vertical direction edge)

231 edge extracting images

232 edge standardized images

233 coordinate transform to the utmost edge standardized images

240 matching treatment parts

240a polar coordinate transform part

240b rotation angle test section

The positive and negative judgement part of 240c

250 registered images storage areas

251 template images

251a t+ image

251b t-image

The relevant judgement of 300 positive and negative separation part

310 standardization correlation value calculation parts

311 standardization correlation images

311a r+ image

311b r-image

320 positive and negative separation associated picture generating portions

321 A+ area images

322 A-area images

33 B+ area images

324 B-area images

330 expansion process parts

330a positive region pattern mask

330b negative region pattern mask

330c input picture mask

330d template image mask

The 331 A+ area images that expanded

The 332 A-area images that expanded

The 333 B+ area images that expanded

The 334 B-area images that expanded

The 335 standardization correlation images that expanded

The 335a r+ image that expanded

The 335b r-image that expanded

340 value of checking calculating sections

Embodiment

Below, the image that present invention will be described in detail with reference to the accompanying is checked the embodiment 1～2 of device, image checking method and image check program.In addition, explanation uses the image of orthogonal coordinate system to check in embodiment 1, and explanation uses the image of polar coordinate system to check in embodiment 2.

Embodiment 1

Fig. 1 is the functional-block diagram that the image of expression embodiment 1 is checked the structure of device.As shown in the drawing, this image is checked device 1 and comprised: image importation 10, image cut out part 20, edge extracting part 30, matching treatment part 40, registered images storage area 50, the relevant judgement of positive and negative separation part 100, and this positive and negative separation is relevant judges that part 100 comprises: standardization correlation value calculation part 110, positive and negative separation associated picture generating portion 120, expansion process part 130, the value of checking calculating section 140.

Image importation 10 is the importations that are used for becoming in the input picture capture apparatus of coin of checking object, the image of input is outputed to image cut out part 20.Specifically, handle input picture image importation 10 as the aggregate of the pixel of stated number.For example, input picture is identified as the have 256 gray shade scales gray level image of concentration value of (gradation level), outputs to image as the rectangular image of prescribed level and cut out part.

Image cuts out part 20 and obtains this rectangular image from image importation 10, only cut out with the external square area of coin image in image, the image that cuts out is outputed to edge extracting part 30.

Fig. 2 is used to illustrate that this image cuts out the key diagram of the processing summary of part 20.As shown in the drawing, image cuts out part 20 and scans input picture of obtaining from image importation 10 11 and the concentration value that adds up whole pixels in the horizontal direction, generates horizontal direction projection 21.In addition, scan input picture 11 in vertical direction, and generate vertical direction projection 22 with same step.Then, image cuts out part 20 horizontal scan direction projections 21 and vertical direction projection 22, calculates the rising coordinate and the decline coordinate of the concentration value of accumulative total.Then, shown in four dotted lines of this figure, each the coordinate institute area surrounded that calculates is cut out as cutting out image 23, and this is cut out image 23 output to edge extracting part 30.

Turn back to the explanation of Fig. 1, edge extracting part 30 is described.Edge extracting part 30 cuts out part 20 from image and obtains and cut out image 23, and calculates and cut out the concentration change (edge strength) of image 23 to avoid other influence of individual difference based on brightness that cuts out image 23 or color (hue) etc.In addition, carry out the standardization of edge strength for the deviation of the edge strength that suppresses to calculate.Specifically, thus by using the edge extracting of Sobel operator to handle edge calculation intensity, with the result of calculation standardization to cutting out image 23.In addition, in embodiment 1, suppose to use the Sobel operator, carry out edge extracting but also can use the Roberts operator to wait.

Fig. 3 is the key diagram that is used to illustrate the Sobel operator.As shown in the drawing, edge extracting part 30 usage level direction edge calculations are carried out the calculating of edge strength with 30a and vertical direction edge calculations with two Sobel operators of 30b.Specifically, for each Sobel operator of the whole picture element scans that cut out image 23 (30a and 30b), obtain horizontal direction edge calculations Gx and vertical direction edge calculations Gy as a result as a result.Then, the edge strength (G) in calculating each pixel is afterwards with this edge strength standardization (E).

[equation 1]

G＝|G _x|+|G _y| ...(1)

$E=c\×\frac{G}{\mathrm{\ΣG}}...\left(2\right)$

Shown in formula (1), the edge strength in each pixel (G) is represented as the horizontal direction edge calculations absolute value of Gx and the vertical direction edge calculations absolute value sum of Gy as a result as a result.In addition, shown in formula (2), the standardization edge strength (E) in each pixel be to each coin kind set the constant c of value of regulation and edge strength (G) long-pending divided by whole results of the summation of the edge strength (G) of pixels.

Like this, by carrying out the standardization of edge strength, can be suppressed at the new coin at easy extraction edge and be difficult to extract the deviation that edge strength takes place between the coin circulation at edge, new and old irrelevant so can carry out checking of various coins accurately with coin.

Fig. 4 is the key diagram that is used to illustrate the summary of the edge extracting processing of being undertaken by edge extracting part 30 (image transformation processing).As shown in the drawing, cut out image 23 by using the edge strength computing of Sobel operator to be edge extracting image 31 by image transformation.Then, edge extracting image 31 is transformed to edge standardized images 32 by the edge strength standardization of using formula (1) and formula (2).Edge extracting part 30 outputs to matching treatment part 40 with this edge standardized images 32.

Each pixel value of edge extracting image 31 shown in this figure is for example got 0～255 value, and 0 gets corresponding to black gray-scale value, and 255 get corresponding to white gray-scale value.In the edge extracting image 31 of this figure, white part is the marginal portion that extracts, and black part is a background parts.In addition, each pixel value of edge standardized images 32 is for example got 0～255 value, and 0 gets corresponding to black gray-scale value, and 255 corresponding to white gray-scale value.In addition, in the edge of this figure standardized images, white part is equivalent to the marginal portion, and black part is equivalent to background parts, and this and edge extracting image 31 are same.

Return the explanation of Fig. 1, matching treatment part 40 is described.Matching treatment part 40 obtains edge standardized images 32 from edge extracting part 30, carry out with registered images storage area 50 in the collation process of template image of storage.Specifically, template image is rotated the angle of regulation at every turn, obtain the template image of each rotation angle and the consistent degree (M) of edge standardized images 32 and be maximum rotation angle (φ).This unanimity degree (M) passes through

[equation 2]

$M\left(\mathrm{\φ}\right)\underset{x}{\mathrm{\Σ}}\underset{y}{\mathrm{\Σ}}{t}_{\mathrm{\φ}}(x,y).s(x,y)...\left(3\right)$

Calculate.

Shown in formula (3), the consistent degree M (φ) of each rotation angle (φ) be whole pixels rotation angle φ template image each pixel concentration value t φ (x, y) and the concentration value s of each pixel of edge standardized images 32 (x, long-pending summation y).

Fig. 5 is the key diagram that is used to illustrate the summary that the coupling of matching treatment part 40 is judged.As shown in the drawing, the curve of the value of M (φ) for having peaked chevron in a certain rotation angle.Matching treatment part 40 obtains the value of this M (φ) for the φ of maximum (apex portion of chevron), with the template image anglec of rotation φ that is stored in the registered images storage area 50.Then, edge standardized images 32 and the template image that rotated are outputed to positive and negative separation is relevant judges part 100.

Fig. 6 is used to illustrate output to from matching treatment part 40 that positive and negative separation is relevant judges the standardization of the edge image of part 100 and the key diagram of the image examples of rotary template image.In the figure, the surface image of representing 10 Japanese yen coins is transfused to the image examples that image is checked the situation in the device 1.That is, implement the standardization of above-mentioned edge and generate edge standardized images 33, the template image rotation is generated the template image 51 that has rotated by above-mentioned matching treatment to cutting out image 24.In the explanation afterwards, use this edge standardized images (front) 33 to replace edge standardized images (reverse side) 32.

In addition, the rotation angle of matching treatment part 40 by rotary template image degree of agreeing maximum is illustrated, but also rotary template image and rotate the rotation angle of edge standardized images 32 degree of agreeing maximums not.

Turn back to the explanation of Fig. 1, registered images storage area 50 is described.Registered images storage area 50 is stored a plurality of template images corresponding with the various coins of registered in advance, and provides these template images to matching treatment part 40.The deviation that causes for the individual difference that such template image is suppressed coin and use the coin image of same kind is synthetic a plurality of and the average image that obtain.By using this average image, make wait in year each coin intrinsic relief pattern part and template image counterpart correlation for and the relevant correlation of the average image (mean value), so be difficult to produce influence when checking.That is, even can prevent real coin but be judged to be situations about counterfeiting coins owing to make a year difference.

Such template image has been implemented after the edge standardization with input picture equally in order to check with the input picture of implementing the edge standardization, is registered in the registered images storage area 50.In addition, a plurality of front and the average image of the reverse side images that applied the edge standardization to each metal species of registration in the registered images storage area 50.

Positive and negative separation is relevant judges that part 100 obtains edge standardized images 33 shown in Figure 6 (below be called ' input picture 33 ') and rotary template image 51 (below be called ' template image 51 ') from matching treatment part 40, by these images being checked thereby whether the coin that carries out input picture 33 is the judgement of checking of real coin, and export this result of determination.

The correlation of each pixel of the correspondence of standardization correlation value calculation part 110 calculating input images 33 and template image 51 generates standardization correlation image with this correlation standardization.Specifically, for coordinate figure be (x, each pixel y), use input picture 33 concentration value s (x, y) and the concentration value t of template image 51 (x, y), by

[equation 3]

$r(x,y)=\frac{(t(x,y)-\frac{\mathrm{\Σt}}{n})(s(x,y)-\frac{\mathrm{\Σs}}{n})}{\sqrt{\{{\mathrm{\Σt}}^2-\frac{{\left(\mathrm{\Σt}\right)}^2}{n}\}.\{{\mathrm{\Σs}}^2-\frac{{\left(\mathrm{\Σs}\right)}^2}{n}\}}}...\left(4\right)$

Calculate each pixel standardization correlation r (x, y).In addition, (x y) for example gets-1.0～+ 1.0 value to the standardization correlation r of each pixel shown in the formula (4).In addition, the n remarked pixel number in the formula (4).

And whether standardization correlation value calculation part 110 is to be separated into positive standardization correlation image (r+ image) and negative standardization correlation image (r-image) more than 0 according to the pixel value of this standardization correlation image.In addition, about template image 51, whether be the threshold value (T of regulation according to each pixel value _t) the above and template image (t-image) that is separated into positive template image (t+ image) and bears.

In addition, the pixel value of r+ image is for example got 0.0～1.0 value, and the pixel value of r-image is got for example 0.0～1.0 value by the absolute value of getting each pixel value.In addition, the pixel value of t+ image and t-image is for example got 0 or 1 two-value.That is, t+ image and t-image have the effect as the employed pattern mask of image transformation of each standardization correlation image.

Here, the meaning of each image is described, r+ graphical representation is becoming the pixel that relevant (similar) arranged between the image of checking object, if very strong being correlated with arranged, then this pixel is got big value.In addition, r-graphical representation is becoming the pixel that does not have relevant (dissmilarity) between the image of checking object, if very strong negative correlation is arranged, then this pixel is got big value.And, the marginal portion of t+ graphical representation template image, 1 value is got in the marginal portion, and background parts is got 0 value.In addition, the background parts of t-graphical representation template image (not being the part at edge), background parts is got 1 value, and 0 value is got in the marginal portion.

Positive and negative separation associated picture generating portion 120 generates positive and negative separation associated picture by the combination of r+ image, r-image, t+ image and t-image that standardization correlation value calculation part 110 generates.Specifically, generate the A+ area image, generate the A-area image, generate the B+ area image, generate the B-area image from r-image and t-image from r+ image and t-image from r-image and t+ image from r+ image and t+ image.

The meaning of each area image is described here.Fig. 7 is the key diagram that is used to illustrate these 4 zones.As shown in the drawing, the A+ area image is the area image with the r+ image and the t+ doubling of the image, and expression has relevantly with the marginal portion, promptly the situation that the edge has appearred in edge should occur, corresponding to the positive feature regional images in the technical scheme.The A-area image is the area image with the r-image and the t+ doubling of the image, and expression does not have relevantly with the marginal portion, promptly the situation that the edge does not appear in edge should occur, corresponding to the negative feature regional images in the technical scheme.The B+ area image is the area image with the r+ image and the t-doubling of the image, and expression has relevantly with background parts, promptly the situation that the edge does not appear in edge should not occur, corresponding to the positive background area image in the technical scheme.The B-area image is the area image with the r-image and the t-doubling of the image, and expression does not have relevantly with background parts, promptly the situation that the edge does not appear in edge should not occur, corresponding to the negative background area image in the technical scheme.

Return the explanation of Fig. 1, expansion process part 130 is described.The pattern mask of expansion process part 130 use regulations moves the pixel of A-area image to the A+ area image, the pixel with the B-area image moves to the B+ area image simultaneously.Carrying out this expansion process is owing to be the isolated point that negative correlation has appearred having in noise-like ground in the standardization correlation.That is, by carrying out this expansion process, the influence that can suppress this isolated point feeds through to the result of determination of the value of checking.

The value of checking calculating section 140 for example is divided on the horizontal direction 4 respectively with A+ area image, A-area image, B+ area image and B-area image, totally 16 pieces of 4 on the vertical direction, by

[equation 4]

$Z=\underset{j=0}{\overset{3}{\mathrm{\Σ}}}\underset{i=0}{\overset{3}{\mathrm{\Σ}}}(a_{\mathrm{ij}}{A}_{\mathrm{ij}}^{+}+b_{\mathrm{ij}}{A}_{\mathrm{ij}}^{-}+c_{\mathrm{ij}}{B}_{\mathrm{ij}}^{+}+d_{\mathrm{ij}}{B}_{\mathrm{ij}}^{-})...\left(5\right)$

Formula (5) is asked the value of checking (Z).Here, coefficient a _Ij, b _Ij, c _IjAnd d _IjUse learning sample to ask optimum solution by linear discriminatory analysis.In addition, as the A of the piece value of each area image ⁺ _Ij, A ^- _Ij, B ⁺ _IjAnd B ^- _IjThe summation of representing the pixel value in each piece.

And if this value of checking (Z) is more than the threshold value, then to check the coin that is judged to be input picture 33 be real coin to the value of checking calculating section 140, and if less than threshold value the value of checking calculating section 140 check to be judged to be and counterfeit coins, export this result of determination then.

After, further specify the relevant processing of judging part 100 of positive and negative separation shown in Figure 1.At first, the positive and negative separating treatment of standardization correlation of using Fig. 8 and Figure 11 description standard correlation value calculation part 110 to carry out.Fig. 8 is the process flow diagram of the positive and negative separating treatment of standardization correlation, and Figure 11 is used for illustrating the relevant key diagram of judging the image generation step of part 100 of positive and negative separation.

As shown in figure 11, standardization correlation value calculation part 110 at first generates standardization correlation image 111 from input picture 33 and template image 51.Then, carry out the positive and negative separating treatment of standardization correlation with the standardization correlation image 111 that generates as input, this standardization correlation image 111 is separated into as the r+ image 111a of positive correlation image with as the r-image 111b of the correlation image of bearing.

As shown in Figure 8, in the positive and negative separating treatment of standardization correlation, at first the starting point pixel to standardization correlation image 111 moves (step S501).This starting point pixel for example is x=0, the pixel of y=0.Then, use formula (4) calculate this pixel standardization correlation r (x, y) (step S502) is if the r (x that calculates, y) be (step S503 " affirms ") more than 0, then with the pixel value (step S504) of this pixel value as the same coordinate of r+ image 111a.On the other hand, if (x is y) less than 0 (step S503 " negates " r that calculates), then with the absolute value of the pixel value of this pixel pixel value (step S505) as the same coordinate of r-image 111b.

Then, under the situation of also whole pixels of standardization correlation image 111 not being finished positive and negative separating treatment (step S506 " negates "), move to the next one and gaze at pixel (step S507), repeat the following processing of step S502.On the other hand, under the situation of positive and negative separating treatment that whole pixels are through with (step S506 " affirms "), end process.By the positive and negative separating treatment of this standardization correlation, r+ image 111a and r-image 111b are generated as the image with the pixel of getting 0.0～1.0 pixel value.In addition, in embodiment 1, illustrated that the pixel value of the pixel of r-image 111b is got 0.0～1.0 pixel value, but this pixel value also can be got-1.0～0.0 value.

Then, the positive and negative separating treatment of template image of using Fig. 9 and Figure 11 description standard correlation value calculation part 110 to carry out.Fig. 9 is the process flow diagram of the positive and negative separating treatment of template image.As shown in figure 11, in the positive and negative separating treatment of template image, carry out template image 51 is separated into as the t+ image 51a of positive template image with as the processing of the t-image 51b of negative template image.

As shown in Figure 9, in the positive and negative separating treatment of template image, at first the starting point pixel to template image 51 moves (step S601).This starting point pixel for example is x=0, the pixel of y=0.Then, if the threshold value (T of the concentration value of this pixel for stipulating _t) above (step S602 " affirms "), then the pixel value with the same coordinate of t+ image 51a is made as 1 (step S603).On the other hand, if this concentration value less than the regulation threshold value (T _t) (step S602 " negates "), then the pixel value with the same coordinate of t-image 51b is made as 1 (step S604).

Then, under the situation of also whole pixels of template image 51 not being finished positive and negative separating treatment (step S605 " negates "), move to the next one and gaze at pixel (step S606), repeat the following processing of step S602.On the other hand, under the situation of positive and negative separating treatment that whole pixels are through with (step S605 " affirms "), end process.By the positive and negative separating treatment of this template image, t+ image 51a is generated as that the marginal portion is 1, background parts is 0 bianry image, and t-image 51b is generated as that the marginal portion is 0, background parts is 1 bianry image.

Then, the positive and negative separation associated picture that uses Figure 10 and Figure 11 to illustrate that positive and negative separation associated picture generating portion 120 is carried out generates and handles.Figure 10 is that positive and negative separation associated picture generates the process flow diagram of handling.

As shown in figure 11, generate in the processing at positive and negative separation associated picture, r+ image 111a, r-image 111b, t+ image 51a and the t-image 51b that will generate in standardization correlation value calculation part 110 generates A+ area image 121, A-area image 122, B+ area image 123 and B-area image 124 as input picture.

For example, with r+ image 111a and, t+ image 51a is as under the situation of input picture, as shown in figure 10, at first the starting point pixel to each image moves (step S701).Then, the pixel value of the t+ image 51a in this pixel is (step S702 " affirms ") under 1 the situation, with the pixel value of A+ area image 121 pixel value (step S703) as r+ image 111a.On the other hand, the pixel value of the t+ image 51a in this pixel is not (that is, being under 0 the situation) (step S702 " negates ") under 1 the situation, and the pixel value of A+ area image 121 is made as 0 (step S704).

Then, under the situation of also whole pixels not being finished area image generation processing (step S705 " negates "), move to the next one and gaze at pixel (step S706), repeat the following processing of step S702.On the other hand, generate under the situation about handling (step S705 " affirms "), generate A+ area image 121, end process at area image that whole pixels are through with.

Equally, generate A-area image 122, generate B+ area image 123, generate B-area image 124 by r-image 111b and t-image 51b by r+ image 111a and t-image 51b by r-image 111b and t+ image 51a.

Then, use Figure 12～Figure 14 that the expansion process that expansion process part 130 is carried out is described.Figure 12 is the key diagram that is used for illustrating the pattern mask that uses in expansion process, and Figure 13 is the process flow diagram of expansion process, and Figure 14 is the key diagram that is used to illustrate the image that is generated by expansion process.

In this expansion process, the processing that the isolated point (pixel) of the noise-like that comprises in the area image (A-area image 122 and B-area image 124) that carries out bearing moves to positive area image (A+ area image 121 and B+ area image 123).By carrying out this processing, can improve the precision of the value of checking.

As shown in figure 12, in this expansion process, use two pattern masks of positive region pattern mask 130a and negative region pattern mask 130b.8 zones that each pattern mask has P5 and M5 and these zones are surrounded.For example, carrying out from A-area image 122 under the situation of the expansion process of A+ area image 121, the pixel of gazing at of the M5 of negative region pattern mask 130b and A-area image 122 merges, and merges with the P5 of positive region pattern mask 130a and corresponding to the pixel of gazing at pixel.Then, successively the pixel value of the pixel value of M5 and P1～P9 is compared and carry out expansion process.

Then, be example to carry out to the situation of the expansion process of A+ area image 121 from A-area image 122, use Figure 13 that the treatment step of this expansion process is described.At first, the starting point pixel to each image (121 and 122) moves (step S801).This starting point pixel for example is x=0, the pixel of y=0.Then, 9 zones (P1～P9) n is set 1 (step S802) in order to switch positive region mask 130a successively.That is, in the moment that step S802 finishes, the zone that becomes the positive region pattern mask 130a of object is P1.

Then, the value of Pn and the value of M5 are compared, under the situation of value of P1 (step S803 " affirms "), be set at 0 (step S805) with the value of M5 displacement P5 and with the value of M5 greater than the value of M5.That is, the pixel with M5 moves to the pixel of P5.On the other hand, be under the situation below the value of M5 (step S803 " negates ") in the value of Pn, the value of n is added 1 (step S804), be under the situation below 9 (step S806 " negates ") in the value of n, and carry out step S803 once more.

Like this, as long as a value greater than M5 is arranged in the value of P1～P9, then the pixel with M5 moves to P5.On the other hand, all be under the situation below the value of M5 (step S806 " affirms ") in the value of P1～P9, do not carry out moving of pixel.

Then, under not to the situation of whole pixel end process of A-area image 122 (step S807 " negates "), gaze at pixel to the next one and move (step S808), carry out the later processing of step S802.On the other hand, under the whole pixels to A-area image 122 are through with situation about handling (step S807 " affirms "), finish this expansion process.

As shown in figure 14, by this expansion process, A+ area image 121, A-area image 122, B+ area image 123 and B-area image 124 respectively by image transformation for the A+ area image 131 that expands, expand A-area image 132, expand the B+ area image 133 and the B-area image 134 that expanded.In addition, because the isolated point on the A-area image 122 moves to A+ area image 121, so the marginal portion of the A+ area image 131 that expanded is compared with A+ area image 121, area increases.On the other hand, the marginal portion of the A-area image 132 that expanded is compared with A-area image 122, and area reduces.

Then, the value of the checking computing of using Figure 15 explanation value of checking calculating section 140 to carry out.Figure 15 is the key diagram that is used to illustrate that the piece of expansion area image (131～134) is cut apart.As shown in the drawing, the value of checking calculating section 140 at first with each expansion area image (131～134) be divided on the horizontal direction on 4, vertical direction totally 16 pieces of 4, generate A+ area image piece 141, A-area image piece 142, B+ area image piece 143 and B-area image piece 144.

Then, check the calculating that calculating section 140 uses formula (5) value of checking (Z).Here, suppose each coefficient a of formula (5) _Ij, b _Ij, c _IjAnd d _IjUse learning sample to wait and ask optimum solution by linear discriminatory analysis.Specifically, because according to the different coins that extract the coins at edge easily and be difficult to extract the edge that exist of the design of the relief pattern of coin, so these coefficients are got different values for the classification of each coin.Thereby check by these coefficient optimizations being carried out high-precision image by learning sample.

And the value of checking calculating section 140 uses the coefficient a that has set optimum value _Ij, b _Ij, c _IjAnd d _IjCalculate the value of checking (Z) with each image block (141～144), under this value of checking is situation more than the threshold value, be judged to be real coin, under situation, be judged to be and counterfeit coins less than threshold value.In addition, in embodiment 1, illustrated with each image segmentation to be 16 situation, but the piece number can be for counting arbitrarily.

In addition, in formula (5), if with coefficient c _IjAnd d _IjBe set at 0, then can only calculate the value of checking (Z) from A+ area image piece 141 and A-area image piece 142.In addition, if with coefficient a _IjAnd b _IjBe set at 0, then can only calculate the value of checking (Z) from B+ area image piece 143 and B-area image piece 144.

Like this, the value of checking calculating section 140 is checked thereby can carry out image expeditiously by adjust the value of each coefficient of image block number or formula (5) according to the ability of the kind of coin or hardware.

In addition, in the value of the checking calculating section 140 of embodiment 1, structure also can constitute by other method and check judgement for calculating the value of checking (Z) by formula (5), but be not limited thereto after piece cuts apart in that each area image having been carried out.For example, also can use other method such as multilayer neural network, support vector machine, secondary recognition function.

After, before being carried out positive and negative separation, standardization correlation image 111 carries out under the situation of expansion process, and use Figure 16～Figure 18 to describe.Figure 16 is the key diagram that the image of this expansion process of explanation generates step, and Figure 17 is the key diagram of the pattern mask that is used for illustrating that this expansion process is used, and Figure 18 is the process flow diagram of this expansion process.

In above-mentioned expansion process, generating each area image (121～124) afterwards, pixel is moved to positive area image (for example, the A+ area image 121) from negative area image (for example, A-area image 122).But this expansion process can use the template image 51 before standardization correlation image 111 and positive and negative separation before the positive and negative separation to carry out.

As shown in figure 16, standardization correlation value calculation part 110 at first generates standardization correlation image 111 by input picture 33 and template image 51.Then, this expansion process is carried out expansion process with the standardization correlation image 111 that generates as input, generates the standardization correlation image 135 that has expanded.Then, this standardization correlation image 135 that has expanded r-image 135b of being separated into the r+ image 135a that has expanded and having expanded.Then, with the r+ image 135a that has expanded, the r-image 135b, the t+ image 51a that have expanded and t-image 51b as input, carry out the processing of positive and negative separation associated picture generating portion 120, output expanded A+ area image 131, expand A-area image 132, expand the B+ area image 133 and the B-area image 134 that expanded.

As shown in figure 17, in this expansion process, use two pattern masks of input picture mask 130c and template image mask 130d.8 zones that each pattern mask has S5 and T5 and these zones are surrounded.For example, using template image 51 and standardization correlation image 111 to carry out under the situation of expansion process, the S5 of input picture mask 130c and the pixel of gazing at of standardization correlation image are merged, merge with the T5 of template image mask 130d and corresponding to the pixel of gazing at pixel.Then, compare with reference to the pixel value in the zone of S1～S9 and T1～T9 and carry out expansion process.

Use Figure 18 that the treatment step of this expansion process is described.At first, the starting point pixel to each image (111 and 51) moves (step S901).This starting point pixel for example is x=0, the pixel of y=0.Then, value at S5 is under the situation about bearing, the standardization correlation that is corresponding pixel is under the negative situation (step S902 " negates "), for 9 zones of switching input picture mask 130c successively (S1～S9) and 9 zones of template image mask 130d (T1～T9) and with n setting 1 (step S903).

Then, in the value of Tn greater than threshold value (T _t) situation under (step S904 " affirms "), judge whether the value of Sn is (step S905) more than 0, if the value of Sn is (step S905 " affirms ") more than 0, then the value of this Sn and the absolute value of S5 are compared (step S906).Then, if the value of Sn greater than the absolute value (step S906 " affirms ") of S5, is then replaced the value (step S907) of Sn with the absolute value of S5.

That is, in the Sn of the periphery of S5, in the value that has Tn greater than threshold value (T _t) and the value of Sn be more than 0, and under the situation of the value of Sn greater than the zone (Sn) of the absolute value of S5, judge that the pixel of this S5 is an isolated point, get S5 value absolute value and with the value counter-rotating of S5.Then, if the whole pixels of standardization correlation image 111 are not finished expansion process (step S910 " negates "), then move (step S911) and repeat processing below the step S902 to gazing at pixel.On the other hand, under the situation of expansion process that whole pixels are through with (step S910 " affirms "), finish this expansion process.

On the other hand, the value at Tn is threshold value (T _t) following (step S904 " negates "), or the value of Sn is negative (step S905 " negates "), or the value of Sn is under the situation of absolute value following (step S906 " negates ") of S5, n is added 1 (step S908), if n is (step S909 " negates ") below 9, then repeat the following processing of step S904.On the other hand, if n then carries out the processing of step S910 greater than 9 (step S909 " affirms ").

Like this, even before the positive and negative separation of standardization correlation image 111, carried out under the situation of expansion process, also can obtain expansion area image (131～134).In this case, owing to use the standardization correlation image 111 before the positive and negative separation, so compare with the expansion process after generating area image (121～124), can cut down the picture number of the object that becomes expansion process, so can carry out expansion process more efficiently.

As mentioned above, image by embodiment 1 is checked device, image checking method and image check program, to implement that edge extracting is handled and the edge standardization and carry out feature extraction after input picture and the template image of having implemented standardization in advance check, generate standardization correlation image, whether be more than the threshold value according to the pixel value in each image simultaneously, and standardization correlation image and template image are separated into positive standardization correlation image and negative standardization correlation image respectively, with positive template image and negative template image, pass through the combination of this image then, generate positive feature regional images, negative feature regional images, positive background area image and negative background area image, and then, enforcement is carried out moving to the pixel of positive feature regional images from negative feature regional images, the expansion process that moves to the pixel of positive feature regional images from negative background area image, with these the area image of expansion process carry out that piece is cut apart and by calculating values of checking such as linear discriminant analysiss and check judgement, so can be with whole pixels of input picture and template image as checking object, get rid of the influence of the isolated point of following correlation value calculation simultaneously, characteristic area not only, and the correlation of background area is reflected in the value of checking also well-balancedly, so can carry out the rate of checking that image was checked and can be improved to high-precision image.

In addition, in embodiment 1, illustrated that the input picture to coin carries out the situation that image is checked, but the present invention is not limited to this, for example, also can be applied to checking of banknote image, or the parts among the FA (FactoryAutomation) etc. or the image of product are checked.

Embodiment 2

The situation that the image that uses orthogonal coordinate system to carry out coin etc. is checked has been described in the foregoing description 1.In present embodiment 2, illustrate and use polar coordinate system to replace the image of orthogonal coordinate system to check.Check by the image that uses polar coordinate system can carry out circular object such as coin more efficiently.

Figure 19 is the functional-block diagram that the image of expression embodiment 2 is checked the structure of device.As shown in the drawing, this image is checked device 201 and is comprised that image importation 210, image cut out part 220, edge extracting part 230, matching treatment part 240, registered images storage area 250, the relevant judgement of positive and negative separation part 300, this matching treatment part 240 comprises: polar coordinate transform part 240a, rotation angle test section 240b, positive and negative judgement part 240c, this positive and negative separation is relevant judges that part 300 comprises: standardization correlation value calculation part 310, positive and negative separation associated picture generating portion 320, expansion process part 330, the value of checking calculating section 340.

Image importation 210 is the importations that are used for becoming in the input picture capture apparatus of coin of checking object, the image of input is outputed to image cut out part 220.Specifically, handle input picture image importation 210 as the aggregate of the pixel of stated number.For example, input picture is identified as the gray level image of the concentration value with 256 gray shade scales, outputs to image as the rectangular image of prescribed level and cut out part.

Image cuts out part 220 and obtains this rectangular image from image importation 210, only cut out with the external square area of coin image in image, the image that cuts out is outputed to edge extracting part 230.

Figure 20 is used to illustrate that this image cuts out the key diagram of the processing summary of part 220.As shown in the drawing, image cuts out part 220 and scans input picture of obtaining from image importation 210 211 and the concentration value that adds up whole pixels in the horizontal direction, generates horizontal direction projection 221.In addition, scan input picture 211 in vertical direction, and generate vertical direction projection 222 with same step.Then, image cuts out part 220 horizontal scan direction projections 221 and vertical direction projection 222, calculates the rising coordinate and the decline coordinate of the concentration value of accumulative total.Then, shown in four dotted lines of this figure, each the coordinate institute area surrounded that calculates is cut out as cutting out image 223, and this is cut out image 223 output to edge extracting part 230.

Turn back to the explanation of Figure 19, edge extracting part 230 is described.Edge extracting part 230 cuts out part 220 from image and obtains and cut out image 223, and calculates and cut out the concentration change (edge strength) of image 223 to avoid other influence of individual difference based on brightness that cuts out image 223 or color.In addition, carry out the standardization of edge strength for the deviation of the edge strength that suppresses to calculate.Specifically, thus by using the edge extracting of Sobel operator to handle edge calculation intensity, with the result of calculation standardization to cutting out image 223.In addition, in embodiment 2, suppose to use the Sobel operator, carry out edge extracting but also can use the Roberts operator to wait.

Figure 21 is the key diagram that is used to illustrate the Sobel operator.As shown in the drawing, edge extracting part 230 usage level direction edge calculations are carried out the calculating of edge strength with 230a and vertical direction edge calculations with two Sobel operators of 230b.Specifically, for each Sobel operator of the whole picture element scans that cut out image 223 (230a and 230b), obtain horizontal direction edge calculations Gx and vertical direction edge calculations Gy as a result as a result.Then, the edge strength (G) in calculating each pixel is afterwards with this edge strength standardization (E).

[equation 5]

G＝|G _x|+|G _y| ...(6)

$E=c\×\frac{G}{\mathrm{\ΣG}}...\left(7\right)$

Shown in formula (6), the edge strength in each pixel (G) is represented as the horizontal direction edge calculations absolute value of Gx and the vertical direction edge calculations absolute value sum of Gy as a result as a result.In addition, shown in formula (7), the standardization edge strength (E) in each pixel be to each coin kind set the constant c of value of regulation and edge strength (G) long-pending divided by whole results of the summation of the edge strength (G) of pixels.

Like this, by carrying out the standardization of edge strength, can be suppressed at the new coin at easy extraction edge and be difficult to extract the deviation that edge strength takes place between the coin circulation at edge, new and old irrelevant so can carry out checking of various coins accurately with coin.

Figure 22 is the key diagram that is used to illustrate the summary of the edge extracting processing of being undertaken by edge extracting part 230 (image transformation processing).As shown in the drawing, cut out image 223 by using the edge strength computing of Sobel operator to be edge extracting image 231 by image transformation.Then, edge extracting image 231 is transformed to edge standardized images 232 by the edge strength standardization of using formula (6) and formula (7).Edge extracting part 230 outputs to matching treatment part 240 with this edge standardized images 232.

Each pixel value of edge extracting image 231 shown in this figure is for example got 0～255 value, and 0 gets corresponding to black gray-scale value, and 255 is corresponding to white gray-scale value.In the edge extracting image 231 of this figure, white part is the marginal portion that extracts, and black part is a background parts.In addition, each pixel of edge standardized images 232 is for example got 0～255 value, and 0 gets corresponding to black gray-scale value, and 255 is corresponding to white gray-scale value.In addition, in the edge of this figure standardized images, white part is equivalent to the marginal portion, and black part is equivalent to background parts, and this and edge extracting image 231 are same.

Return the explanation of Figure 19, matching treatment part 240 is described.Matching treatment part 240 obtains edge standardized images 232 from edge extracting part 230, and obtains the edge standardization and the template image of polar coordinate transform from registered images storage area 250.Then, this edge standardized images 232 is carried out polar coordinate transform, move by template image parallel and to detect this image of polar coordinate transform and fleet angle of template image, carry out positive and negative judgement simultaneously, positive and negative separation is relevant judges part 300 thereby edge standardized images 232 and the template image of having proofreaied and correct fleet angle outputed to.In addition, in embodiment 2, thereby illustrate by making the parallel situation that move to detect fleet angle of template image, thus but also can be by carrying out parallel moving and detect fleet angle to edge standardized images 232 having been carried out the image behind the polar coordinate transform.

Polar coordinate transform part 240a is the processing section that is used for edge standardized images 232 is carried out polar coordinate transform.Specifically, the central point of edge calculation standardized images 232, with this central point as polar initial point.Then, according to rotation angle θ and decentering point determine each pixel apart from ρ, by with each pixel to ρ-θ spatial movement, thereby carry out polar coordinate transform.This conversion is used

[equation 6]

x＝ρ·cos(θ) ...(8)

y＝ρ·sin(θ) ...(9)。

Figure 23 is the key diagram that is used to illustrate the processing summary that these polar coordinates change.With the central point of x-y space (edge standardized images 232) as initial point, with (x y) represents the coordinate of each pixel, and this x and y and above-mentioned ρ and θ have the relation shown in formula (8) and the formula (9).Thereby, by with each pixel in the edge standardized images 232 (x, y) be transformed to the relation that satisfies formula (8) and (9) (ρ, θ), thereby polar coordinate transform part 240a generates coordinate transform to the utmost edge standardized images 233.

In addition, in the figure, what show decentering point gets 10～100 value apart from ρ, and rotation angle θ gets the situation of 0～255 value, but the scope of these values can at random be set.

Return the explanation of Figure 19, rotation angle test section 240b is described.Rotation angle test section 240b detects coordinate transform to the utmost edge standardized images 233 and handles by same polar coordinate transform and carried out the fleet angle of the template image of polar coordinate transform in advance, and carries out processing that the fleet angle of two images is proofreaied and correct.Figure 24 is the key diagram that is used to illustrate the processing summary of rotation angle test section 240b.

As shown in the drawing, in ρ-θ space, template image 251 and θ coordinate axis are moved abreast.Then, calculate the fleet angle (φ) of θ coordinate direction and the consistent degree M (φ) of template image 251 in each fleet angle (φ) and edge standardized images 232, and obtain the rotationangle of this unanimity degree M (φ) for maximum _MaxIn addition, this unanimity degree M (φ) passes through

[equation 7]

$M\left(\mathrm{\φ}\right)=\underset{k}{\mathrm{\Σ}}\underset{\mathrm{\θ}=0}{\overset{255}{\mathrm{\Σ}}}t(k,\mathrm{\θ}-\mathrm{\φ}).s(k,\mathrm{\θ})...\left(10\right)$

Calculate.

Shown in formula (10), consistent degree M (φ) in each fleet angle (φ) is departing from template image 251 under the situation of φ, concentration value t (the k of each pixel of template image 251, (k is θ) to the long-pending summation of each pixel for the concentration value s of each pixel of θ-φ) and edge standardized images 232.Here, k has carried out the selective value of selecting to the distance apart from occurring feature among the ρ easily of the central point in the isolated edge standardized images 232.For example, by 16 ρ values that occur feature easily of selection in the ρ (0～100) of coordinate transform to the utmost edge standardized images 233 shown in Figure 23, thereby select this k.

As shown in figure 24, the curve of the value of this M (φ) for having peaked chevron in a certain rotation angle.Rotation angle test section 240b obtains the φ of this M (φ) for maximum (apex portion of chevron) _MaxValue, like this, rotation angle test section 240b is because parallel the moving of the ρ-θ image by having implemented polar coordinate transform proofreaied and correct fleet angle, can cut down calculated amount so compare with the method for proofreading and correct fleet angle by the rotation of x-y image.

Return the explanation of Figure 19, positive and negative judgement part 240c is described.At first, positive and negative judgement part 240c calculates the maximal value M (φ of above-mentioned consistent degree M (φ) between with template image and coordinate transform to the utmost edge standardized images 233 with template image and reverse side in the front of polar coordinate transform _Max), and by this M (φ _Max) ask standardized correlation coefficients R.Specifically, this standardized correlation coefficients R passes through

[equation 8]

$R=\frac{\underset{\mathrm{\θ}}{\mathrm{\Σ}}(t(\mathrm{\θ}-{\mathrm{\θ}}_{\max })-\frac{\mathrm{\Σt}}{N})(s\left(\mathrm{\θ}\right)-\frac{\mathrm{\Σs}}{N})}{\sqrt{\{{\mathrm{\Σt}}^2-\frac{{\left(\mathrm{\Σt}\right)}^2}{N}}\}.\{{\mathrm{\Σs}}^2-\frac{{\left(\mathrm{\Σs}\right)}^2}{N}\}}=\frac{N.M\left({\mathrm{\φ}}_{\max }\right)-\mathrm{\Σt}.\mathrm{\Σs}}{\sqrt{\{N.{\mathrm{\Σt}}^2-{\left(\mathrm{\Σt}\right)}^2\}\·\{N\·\mathrm{\Σ}{s}^2-{\left(\mathrm{\Σs}\right)}^2\}}}...\left(11\right)$

Obtain.In addition, the N in the formula (11) represents to become judgement object pixels number.

Then, the big template image of this positive and negative judgement part 240c choice criteria coefficient R together outputs to the positive and negative relevant judgement part 300 of separating with coordinate transform to the utmost edge standardized images 233.For example, reverse side with the standardized correlation coefficients R of template image and coordinate transform to the utmost edge standardized images 233 than positive with the big situation of the standardized correlation coefficients R of template image and coordinate transform to the utmost edge standardized images 233 under, reverse side is outputed to the positive and negative relevant judgement part 300 of separating with template image and coordinate transform to the utmost edge standardized images 233.Here, judge that to positive and negative separation is relevant the template image of part 300 outputs is parallel move angle φ _MaxAnd proofreaied and correct template image with the fleet angle of coordinate transform to the utmost edge standardized images 233.

Turn back to the explanation of Figure 19, registered images storage area 250 is described.Registered images storage area 250 is stored a plurality of template images corresponding with the various coins of registered in advance, and provides these template images to matching treatment part 240.The deviation that causes for the individual difference that such template image is suppressed coin and use the coin image of same kind is synthetic a plurality of and the average image that obtain.By using this average image, make wait in year each coin intrinsic relief pattern part and template image counterpart correlation for and the relevant correlation of the average image (mean value), so be difficult to produce influence when checking.That is, even can prevent real coin but be judged to be situations about counterfeiting coins owing to make a year difference.

Such template image for be implemented with input picture that polar coordinate transform is handled with implementing that polar coordinate transform is handled and the input picture of edge standardization is checked equally and the edge standardization after, be registered in the registered images storage area 250.In addition, the front of a plurality of each metal species of registration and the template image of reverse side in the registered images storage area 250.

Return the explanation of Figure 19, the relevant judgement of positive and negative separation part 300 is described.Positive and negative separation is relevant judges the template image 251 (below be called ' template image 251 ') that part 300 obtains coordinate transform to the utmost edge standardized images 233 shown in Figure 24 (below be called ' input picture 233 ') and proofreaied and correct fleet angle from matching treatment part 240, by these images being checked thereby whether the coin that carries out input picture 233 is the judgement of checking of real coin, and export this result of determination.

The correlation of each pixel of the correspondence of standardization correlation value calculation part 310 calculating input images 233 and template image 251 generates standardization correlation image with this correlation standardization.Specifically, be that ((k θ) and proofreaies and correct concentration value t (k, the θ-φ of the template image 251 of fleet angle to use the concentration value s of input picture 233 for k, each pixel θ) for coordinate figure _Max), by

[equation 9]

$r(k,\mathrm{\θ})=\frac{(t(k,\mathrm{\θ}-{\mathrm{\φ}}_{\max })-\frac{\mathrm{\Σt}}{n})(s(k,\mathrm{\θ})-\frac{\mathrm{\Σs}}{n})}{\sqrt{\{{\mathrm{\Σt}}^2-\frac{{\left(\mathrm{\Σt}\right)}^2}{n}\}.\{{\mathrm{\Σs}}^2-\frac{{\left(\mathrm{\Σs}\right)}^2}{n}\}}}...\left(12\right)$

Calculate each pixel standardization correlation r (k, θ).In addition, (k θ) for example gets-1.0～+ 1.0 value to the standardization correlation r of each pixel shown in the formula (12).In addition, the n remarked pixel number in the formula (12).

And whether standardization correlation value calculation part 310 is to be separated into positive standardization correlation image (r+ image) and negative standardization correlation image (r-image) more than 0 according to the pixel value of this standardization correlation image.In addition, about template image 251, whether be the threshold value (T of regulation according to each pixel value _t) the above and template image (t-image) that is separated into positive template image (t+ image) and bears.

In addition, the pixel value of r+ image is for example got 0.0～1.0 value, and the pixel value of r-image is for example got 0.0～1.0 value by the absolute value of getting each pixel value.In addition, the pixel value of t+ image and t-image is for example got 0 or 1 two-value.That is, t+ image and t-image have the effect as the employed pattern mask of image transformation of each standardization correlation image.

Here, the meaning of each image is described, r+ graphical representation is becoming the pixel that relevant (similar) arranged between the image of checking object, if very strong being correlated with arranged, then this pixel is got big value.In addition, r-graphical representation is becoming the pixel that does not have relevant (dissmilarity) between the image of checking object, if very strong negative correlation is arranged, then this pixel is got big value.And, the marginal portion of t+ graphical representation template image, 1 value is got in the marginal portion, and background parts is got 0 value.In addition, the background parts of t-graphical representation template image (not being the part at edge), background parts is got 1 value, and 0 value is got in the marginal portion.

Positive and negative separation associated picture generating portion 320 generates positive and negative separation associated picture by the combination of r+ image, r-image, t+ image and t-image that standardization correlation value calculation part 310 generates.Specifically, generate the A+ area image, generate the A-area image, generate the B+ area image, generate the B-area image from r-image and t-image from r+ image and t-image from r-image and t+ image from r+ image and t+ image.

The meaning of each area image is described here.Figure 25 is the key diagram that is used to illustrate these 4 zones.As shown in the drawing, the A+ area image is the area image with the r+ image and the t+ doubling of the image, and expression has relevantly with the marginal portion, promptly the situation that the edge has appearred in edge should occur, corresponding to the positive feature regional images in the technical scheme.The A-area image is the area image with the r-image and the t+ doubling of the image, and expression does not have relevantly with the marginal portion, promptly the situation that the edge does not appear in edge should occur, corresponding to the negative feature regional images in the technical scheme.The B+ area image is the area image with the r+ image and the t-doubling of the image, and expression has relevantly with background parts, promptly the situation that the edge does not appear in edge should not occur, corresponding to the positive background area image in the technical scheme.The B-area image is the area image with the r-image and the t-doubling of the image, and expression does not have relevantly with background parts, promptly the situation that the edge does not appear in edge should not occur, corresponding to the negative background area image in the technical scheme.

Return the explanation of Figure 19, expansion process part 330 is described.The pattern mask of expansion process part 330 use regulations moves the pixel of A-area image to the A+ area image, the pixel with the B-area image moves to the B+ area image simultaneously.Carrying out this expansion process is owing to be the isolated point that negative correlation has appearred having in noise-like ground in the standardization correlation.That is, by carrying out this expansion process, the influence that can suppress this isolated point feeds through to the result of determination of the value of checking.

The value of checking calculating section 340 for example is divided on the horizontal direction 16 respectively with A+ area image, A-area image, B+ area image and B-area image, totally 64 pieces of 4 on the vertical direction, by

[equation 10]

$Z=\underset{j=0}{\overset{3}{\mathrm{\Σ}}}\underset{i=0}{\overset{15}{\mathrm{\Σ}}}(a_{\mathrm{ij}}{A}_{\mathrm{ij}}^{+}+b_{\mathrm{ij}}{A}_{\mathrm{ij}}^{-}+c_{\mathrm{ij}}{B}_{\mathrm{ij}}^{+}+d_{\mathrm{ij}}{B}_{\mathrm{ij}}^{-})---\left(13\right)$

Formula (13) is asked the value of checking (Z).Here, coefficient a _Ij, b _Ij, c _IjAnd d _IjUse learning sample to ask optimum solution by linear discriminatory analysis.In addition, as the A of the piece value of each area image ⁺ _Ij, A ^- _Ij, B ⁺ _IjAnd B ^- _IjThe summation of representing the pixel value in each piece.

And if this value of checking (Z) is more than the threshold value, then to be judged to be the coin of input picture 233 are real coins to the value of checking calculating section 340, and if less than threshold value the value of checking calculating section 340 be judged to be and counterfeit coins, export this result of determination then.

After, further specify the relevant processing of judging part 300 of positive and negative separation shown in Figure 19.At first, the relevant positive and negative separating treatment of the standardization of using Figure 26 and Figure 29 description standard correlation value calculation part 310 to carry out.Figure 26 is the process flow diagram of the positive and negative separating treatment of standardization correlation, and Figure 29 is used for illustrating the relevant key diagram of judging the image generation step of part 300 of positive and negative separation.

As shown in figure 29, standardization correlation value calculation part 310 at first generates standardization correlation image 311 from input picture 233 and template image 251.Then, carry out the positive and negative separating treatment of standardization correlation with the standardization correlation image 311 that generates as input, this standardization correlation image 311 is separated into as the r+ image 311a of positive correlation image with as the r-image 311b of the correlation image of bearing.

As shown in figure 26, in the positive and negative separating treatment of standardization correlation, at first the starting point pixel to standardization correlation image 311 moves (step S1501).This starting point pixel for example is k=0, the pixel of θ=0.Then, use formula (12) calculate this pixel standardization correlation r (k, θ) (step S1502) is if the r (k that calculates, θ) be (step S1503 " affirms ") more than 0, then with the pixel value (step S1504) of this pixel value as the same coordinate of r+ image 311a.On the other hand, if (k is θ) less than 0 (step S1503 " negates " r that calculates), then with the absolute value of the pixel value of this pixel pixel value (step S1505) as the same coordinate of r-image 311b.

Then, under the situation of also whole pixels of standardization correlation image 311 not being finished positive and negative separating treatment (step S1506 " negates "), move to the next one and gaze at pixel (step S1507), repeat the following processing of step S1502.On the other hand, under the situation of positive and negative separating treatment that whole pixels are through with (step S1506 " affirms "), end process.By the positive and negative separating treatment of this standardization correlation, r+ image 311a and r-image 311b are generated as the image with the pixel of getting 0.0～1.0 pixel value.In addition, in embodiment 2, illustrated that the pixel value of the pixel of r-image 311b is got 0.0～1.0 pixel value, but this pixel value also can be got-1.0～0.0 value.

Then, the positive and negative separating treatment of template image of using Figure 27 and Figure 29 description standard correlation value calculation part 310 to carry out.Figure 27 is the process flow diagram of the positive and negative separating treatment of template image.As shown in figure 29, in the positive and negative separating treatment of template image, carry out template image 251 is separated into as the t+ image 251a of positive template image with as the processing of the t-image 251b of negative template image.

As shown in figure 27, in the positive and negative separating treatment of template image, at first the starting point pixel to template image 251 moves (step S1601).This starting point pixel for example is k=0, the pixel of θ=0.Then, if the threshold value (T of the concentration value of this pixel for stipulating _t) above (step S1602 " affirms "), then the pixel value with the same coordinate of t+ image 251a is made as 1 (step S1603).On the other hand, if the concentration value of this pixel less than the regulation threshold value (T _t) (step S1602 " negates "), then the pixel value with the same coordinate of t-image 251b is made as 1 (step S1604).

Then, also not under the situation of whole pixels of template image 251 being finished positive and negative separating treatment (step S1605 " negates "), move to the next one and gaze at pixel (step S1606), repeat the following processing of step S1602.On the other hand, under the situation of positive and negative separating treatment that whole pixels are through with (step S1605 " affirms "), end process.By the positive and negative separating treatment of this template image, t+ image 251a is generated as that the marginal portion is 1, background parts is 0 bianry image, and t-image 251b is generated as that the marginal portion is 0, background parts is 1 bianry image.

Then, the positive and negative separation associated picture that uses Figure 28 and Figure 29 to illustrate that positive and negative separation associated picture generating portion 320 is carried out generates and handles.Figure 28 is that positive and negative separation associated picture generates the process flow diagram of handling.

As shown in figure 29, generate in the processing at positive and negative separation associated picture, r+ image 311a, r-image 311b, t+ image 251a and the t-image 251b that will generate in standardization correlation value calculation part 310 generates A+ area image 321, A-area image 322, B+ area image 323 and B-area image 324 as input picture.

For example, with r+ image 311a and, t+ image 251a is as under the situation of input picture, as shown in figure 28, at first the starting point pixel to each image moves (step S1701).Then, the pixel value of the t+ image 251a in this pixel is (step S1702 " affirms ") under 1 the situation, with the pixel value of A+ area image 321 pixel value (step S1703) as r+ image 311a.On the other hand, the pixel value of the t+ image 251a in this pixel is not (that is, being under 0 the situation) (step S1702 " negates ") under 1 the situation, and the pixel value of A+ area image 321 is made as 0 (step S1704).

Then, under the situation of also whole pixels not being finished area image generation processing (step S1705 " negates "), move to the next one and gaze at pixel (step S1706), repeat the following processing of step S1702.On the other hand, generate under the situation about handling (step S1705 " affirms ") at area image that whole pixels are through with, owing to generate A+ area image 321, so end process.

Equally, generate A-area image 322, generate B+ area image 323, generate B-area image 324 by r-image 311b and t-image 251b by r+ image 311a and t-image 251b by r-image 311b and t+ image 251a.

Then, use Figure 30～Figure 32 that the expansion process that expansion process part 330 is carried out is described.Figure 30 is the key diagram that is used for illustrating the pattern mask that uses in expansion process, and Figure 31 is the process flow diagram of expansion process, and Figure 32 is the key diagram that is used to illustrate the image that is generated by expansion process.

In this expansion process, the processing that the isolated point (pixel) of the noise-like that comprises in the area image (A-area image 322 and B-area image 324) that carries out bearing moves to positive area image (A+ area image 321 and B+ area image 323).By carrying out this processing, can improve the precision of the value of checking.

As shown in figure 30, in this expansion process, use two pattern masks of positive region pattern mask 330a and negative region pattern mask 330b.8 zones that each pattern mask has P5 and M5 and these zones are surrounded.For example, carrying out from A-area image 322 under the situation of the expansion process of A+ area image 321, the pixel of gazing at of the M5 of negative region pattern mask 330b and A-area image 322 merges, and merges with the P5 of positive region pattern mask 330a and corresponding to the pixel of gazing at pixel.Then, successively the pixel value of the pixel value of M5 and P1～P9 is compared and carry out expansion process.

Then, be example to carry out to the situation of the expansion process of A+ area image 321 from A-area image 322, use Figure 31 that the treatment step of this expansion process is described.At first, the starting point pixel to each image (321 and 322) moves (step S1801).This starting point pixel for example is k=0, the pixel of θ=0.Then, 9 zones (P1～P9) n is set 1 (step S1802) in order to switch positive region mask 330a successively.That is, in the moment that step S1802 finishes, the zone that becomes the positive region pattern mask 330a of object is P1.

Then, the value of Pn and the value of M5 are compared, under the situation of value of P1 (step S1803 " affirms "), be set at 0 (step S1805) with the value of M5 displacement P5 and with the value of M5 greater than the value of M5.That is, the pixel with M5 moves to the pixel of P5.On the other hand, be under the situation below the value of M5 (step S1803 " negates ") in the value of Pn, the value of n is added 1 (step S1804), be under the situation below 9 (step S1806 " negates ") in the value of n, and carry out step S1803 once more.

Like this, as long as a value greater than M5 is arranged in the value of P1～P9, then the pixel with M5 moves to P5.On the other hand, all be under the situation below the value of M5 (step S1806 " affirms ") in the value of P1～P9, do not carry out moving of pixel.

Then, under not to the situation of whole pixel end process of A-area image 322 (step S1807 " negates "), gaze at pixel to the next one and move (step S1808), carry out the later processing of step S1802.On the other hand, under the whole pixels to A-area image 322 are through with situation about handling (step S1807 " affirms "), finish this expansion process.

Shown in figure 32, by this expansion process, A+ area image 321, A-area image 322, B+ area image 323 and B-area image 324 respectively by image transformation for the A+ area image 33 1 that expands, expand A-area image 332, expand the B+ area image 333 and the B-area image 334 that expanded.In addition, because the isolated point on the A-area image 322 moves to A+ area image 321, so the marginal portion of the A+ area image 331 that expanded is compared with A+ area image 321, area increases.On the other hand, the marginal portion of the A-area image 132 that expanded is compared with A-area image 322, and area reduces.

Then, the value of the checking computing of using Figure 33 explanation value of checking calculating section 340 to carry out.Figure 33 is used for explanation to cut apart the key diagram that the A+ area image 331 that expands is carried out the example of the situation that piece cuts apart about the piece of expansion area image (331～334).As shown in the drawing, the value of the checking calculating section 340 A+ area image 331 that at first each expanded is divided on the horizontal direction on 16, vertical direction totally 64 pieces of 4.Equally, to the A-area image 332 that expands, the expand B+ area image 333 and the B-area image 334 that expanded also carries out piece and cuts apart.

Then, the value of checking calculating section 340 uses the calculating of formula (13) value of checking (Z).Here, suppose each coefficient a of formula (13) _Ij, b _Ij, c _IjAnd d _IjUse learning sample to wait and ask optimum solution by linear discriminatory analysis.Specifically, because according to the different coins that extract the coins at edge easily and be difficult to extract the edge that exist of the design of the relief pattern of coin, so these coefficients are got different values for the classification of each coin.Thereby check by these coefficient optimizations being carried out high-precision image by learning sample.

And the value of checking calculating section 340 uses the coefficient a that has set optimum value _Ij, b _Ij, c _IjAnd d _IjCalculate the value of checking (Z) with each image block, under this value of checking is situation more than the threshold value, be judged to be real coin, under situation, be judged to be and counterfeit coins less than threshold value.In addition, in embodiment 2, illustrated with each image segmentation to be 64 situation, but the piece number can be several arbitrarily.

In addition, in formula (13), if with coefficient C _IjAnd d _IjBe set at 0, then can only calculate the value of checking (Z) from A+ area image piece and A-area image piece.In addition, if with coefficient a _IjAnd b _IjBe set at 0, then can only calculate the value of checking (Z) from B+ area image piece and B-area image piece.

Like this, the value of checking calculating section 340 is checked thereby can carry out image expeditiously by adjust the value of each coefficient of image block number or formula (13) according to the ability of the kind of coin or hardware.

In addition, in the value of the checking calculating section 340 of embodiment 2, structure is for carrying out each area image after piece cuts apart, calculate the value of checking (Z) by formula (13), but be not limited thereto, for example, also can use other method such as multilayer neural network, support vector machine, secondary recognition function.

After, before being carried out positive and negative separation, standardization correlation image 311 carries out under the situation of expansion process, and use Figure 34～Figure 36 to describe.Figure 34 is the key diagram that the image of this expansion process of explanation generates step, and Figure 35 is the key diagram of the pattern mask that is used for illustrating that this expansion process is used, and Figure 36 is the process flow diagram of this expansion process.

In above-mentioned expansion process, generating each area image (321～324) afterwards, pixel is moved to positive area image (for example, the A+ area image 321) from negative area image (for example, A-area image 322).But this expansion process can use the template image 251 before standardization correlation image 311 and positive and negative separation before the positive and negative separation to carry out.

As shown in figure 34, standardization correlation value calculation part 310 at first generates standardization correlation image 311 by input picture 233 and template image 251.Then, this expansion process is carried out expansion process with the standardization correlation image 311 that generates as input, generates the standardization correlation image 335 that has expanded.Then, this standardization correlation image 335 that has expanded r-image 335b of being separated into the r+ image 335a that has expanded and having expanded.Then, with the r+ image 335a that has expanded, the r-image 335b, the t+ image 251a that have expanded and t-image 251b as input, carry out the processing of positive and negative separation associated picture generating portion 320, output expanded A+ area image 331, expand A-area image 332, expand the B+ area image 333 and the B-area image 334 that expanded.

As shown in figure 35, in this expansion process, use two pattern masks of input picture mask 330c and template image mask 330d.8 zones that each pattern mask has S5 and T5 and these zones are surrounded.For example, using template image 251 and standardization correlation image 311 to carry out under the situation of expansion process, the S5 of input picture mask 330c and the pixel of gazing at of standardization correlation image are merged, merge with the T5 of template image mask 330d and corresponding to the pixel of gazing at pixel.Then, compare with reference to the pixel value in the zone of S1～S9 and T1～T9 and carry out expansion process.

Use Figure 36 that the treatment step of this expansion process is described.At first, the starting point pixel to each image (311 and 251) moves (step S1901).This starting point pixel for example is k=0, the pixel of θ=0.Then, value at S5 is under the situation about bearing, the standardization correlation that is corresponding pixel is under the negative situation (step S1902 " negates "), for 9 zones of switching input picture mask 330c successively (S1～S9) and 9 zones of template image mask 330d (T1～T9) and to n setting 1 (step S1903).

Then, in the value of Tn greater than threshold value (T _t) situation under (step S1904 " affirms "), judge whether the value of Sn is (step S1905) more than 0, if the value of Sn is (step S1905 " affirms ") more than 0, then the value of this Sn and the absolute value of S5 are compared (step S1906).Then, if the value of Sn greater than the absolute value (step S1906 " affirms ") of S5, is then replaced the value (step S1907) of Sn with the absolute value of S5.

That is, in the Sn of the periphery of S5, in the value that has Tn greater than threshold value (T _t) and the value of Sn be more than 0, and under the situation of the value of Sn greater than the zone (Sn) of the absolute value of S5, judge that the pixel of this S5 is an isolated point, get S5 value absolute value and with the value counter-rotating of S5.Then, if the whole pixels of standardization correlation image 311 are not finished expansion process (step S1910 " negates "), then move (step S1911) and repeat processing below the step S1902 to gazing at pixel.On the other hand, under the situation of expansion process that whole pixels are through with (step S1910 " affirms "), finish this expansion process.

On the other hand, the value at Tn is threshold value (T _t) following (step S1904 " negates "), or the value of Sn is negative (step S1905 " negates "), or the value of Sn is under the situation of absolute value following (step S1906 " negates ") of S5, n is added 1 (step S1908), if n is (step S1909 " negates ") below 9, then repeat the following processing of step S1904.On the other hand, if n then carries out the processing of step S1910 greater than 9 (step S1909 " affirms ").

Like this, even before the positive and negative separation of standardization correlation image 311, carried out under the situation of expansion process, also can obtain expansion area image (331～334).In this case, owing to use the standardization correlation image 311 before the positive and negative separation, so compare with the expansion process after generating area image (321～324), can cut down the picture number of the object that becomes expansion process, so can carry out expansion process more efficiently.

As mentioned above, image by embodiment 2 is checked device, image checking method and image check program, to implement that edge extracting is handled and the edge standardization and carry out feature extraction after the polar coordinate transform input picture and the polar coordinate transform template image of having implemented the edge standardization in advance check, and the fleet angle of image proofreaied and correct and generate standardization correlation image, whether be more than the threshold value according to the pixel value in each image simultaneously, and standardization correlation image and template image are separated into positive standardization correlation image and negative standardization correlation image respectively, with positive template image and negative template image, pass through the combination of this image then, generate positive feature regional images, negative feature regional images, positive background area image and negative background area image, and then, enforcement is carried out moving to the pixel of positive feature regional images from negative feature regional images, the expansion process that moves to the pixel of positive feature regional images from negative background area image, with these the area image of expansion process carry out that piece is cut apart and by calculating values of checking such as linear discriminant analysiss and check judgement, so can be with whole pixels of input picture and template image as checking object, get rid of the influence of the isolated point of following correlation value calculation simultaneously, characteristic area not only, and the correlation of background area is reflected in the value of checking also well-balancedly, so can carry out the rate of checking that image was checked and can be improved to high-precision image.

In addition, in embodiment 2, illustrated that the input picture to coin carries out the situation that image is checked, but the present invention is not limited to this, for example, the image that also can be applied to employed metal species such as play facility is checked, or the central portion among the FA (Factory Automation) etc. or the image of circular product are checked.In addition, the checking object thing is also not necessarily circular, has the coin of point symmetry shape or parts etc. for octagon or positive ten hexagons etc., also can use the present invention.

Utilizability on the industry

More than, image checking device of the present invention, image checking method and image check program are used for thing The image of product is checked, and is particularly suitable for checking of the currency such as currency or coin.

Claims (29)

1. an image is checked device, compares by the feature to image between a plurality of template images of the input picture of checking object thing and registered in advance, thereby image is checked, and it is characterized in that this image is checked device and comprised:

Whether correlation separation of images parts generate the correlation image by described input picture and described template image, be more than the threshold value and with this correlation separation of images to be positive correlation value image and negative correlation value image according to pixel value;

Whether the template image separating component is more than the threshold value described template image to be separated into positive template image and negative norm plate image according to pixel value;

Positive and negative separation associated picture generates parts, and the combination by described positive correlation value image and described negative correlation value image and described positive template image and described negative norm plate image generates a plurality of positive and negative separation associated pictures; And

Check judging part, use described positive and negative separation associated picture to check judgement.

2. image as claimed in claim 1 is checked device, it is characterized in that, described positive and negative separation associated picture generates parts and generates positive feature regional images and negative feature regional images, described positive feature regional images with to long-pending calculating of the described positive correlation value image of each pixel and described positive template image and the value that obtains as pixel value, described negative feature regional images with to long-pending calculating of the described negative correlation value image of each pixel and described positive template image and the value that obtains as pixel value.

3. image as claimed in claim 1 is checked device, it is characterized in that, described positive and negative separation associated picture generates parts and generates positive background area image and negative background area image, described positive background area image with to long-pending calculating of the described positive correlation value image of each pixel and described negative norm plate image and the value that obtains as pixel value, described negative background area image with to long-pending calculating of the described negative correlation value image of each pixel and described negative norm plate image and the value that obtains as pixel value.

4. image as claimed in claim 1 is checked device, it is characterized in that, described positive and negative separation associated picture generates parts and generates positive feature regional images, negative feature regional images, positive background area image and negative background area image, described positive feature regional images with to long-pending calculating of the described positive correlation value image of each pixel and described positive template image and the value that obtains as pixel value, described negative feature regional images with to long-pending calculating of the described negative correlation value image of each pixel and described positive template image and the value that obtains as pixel value, described positive background area image with to long-pending calculating of the described positive correlation value image of each pixel and described negative norm plate image and the value that obtains as pixel value, described negative background area image with to long-pending calculating of the described negative correlation value image of each pixel and described negative norm plate image and the value that obtains as pixel value.

5. check device as claim 2,3 or 4 described images, it is characterized in that, described positive and negative separation associated picture generates parts this surrounding pixel of gazing at the pairing respective pixel of pixel in the positive region image of gazing at pixel and the described positive correlation value image generation of use in the negative region image that uses described negative correlation value image generation is compared, gaze at greater than this at the pixel value of at least one this surrounding pixel under the situation of pixel value of pixel, carry out this is gazed at the expansion process that pixel moves to this respective pixel.

6. image as claimed in claim 1 is checked device, it is characterized in that, described input picture and described template image are to handle the edge image that has carried out image transformation by the edge extracting that uses the edge extracting operator.

7. image as claimed in claim 6 is checked device, it is characterized in that, described edge image is that the edge strength at the edge that will extract carries out standardization and the standardized images that obtains.

8. image as claimed in claim 1 is checked device, it is characterized in that, described template image is each the individual image for described checking object thing to be averaged and the average image that obtains.

9. image as claimed in claim 1 is checked device, it is characterized in that, described correlation image be with the correlation with each pixel of described input picture or described template image carry out standardization and the standardization correlation that obtains as the image of pixel value.

10. image as claimed in claim 1 is checked device, it is characterized in that, described check that judging part carries out to described positive and negative separation associated picture that piece is cut apart and the summation of calculating the pixel value in each piece as the piece value, thereby check judgement by whole described positive and negative separation associated pictures are calculated the value of checking with the long-pending Calais mutually of this piece value and weighting coefficient.

11. image as claimed in claim 1 is checked device, it is characterized in that, describedly checks judging part calculates described weighting coefficient by linear discriminant analysis value.

12. image as claimed in claim 1 is checked device, it is characterized in that, described checking object thing is a currency.

13. an image checking method compares by the feature to image between a plurality of template images of the input picture of checking object thing and registered in advance, thereby image is checked, and it is characterized in that this image checking method comprises:

Whether correlation separation of images step generates the correlation image by described input picture and described template image, be more than the threshold value and with this correlation separation of images to be positive correlation value image and negative correlation value image according to pixel value;

Whether the template image separating step is more than the threshold value described template image to be separated into positive template image and negative norm plate image according to pixel value;

Positive and negative separation associated picture generates step, and the combination by described positive correlation value image and described negative correlation value image and described positive template image and described negative norm plate image generates a plurality of positive and negative separation associated pictures; And

Check determination step, use described positive and negative separation associated picture to check judgement.

14. image check program, compare by feature between a plurality of template images of the input picture of checking object thing and registered in advance image, thereby image is checked, be it is characterized in that, this image check program makes computing machine carry out following steps:

Whether correlation separation of images step generates the correlation image by described input picture and described template image, be more than the threshold value and with this correlation separation of images to be positive correlation value image and negative correlation value image according to pixel value;

Whether the template image separating step is more than the threshold value described template image to be separated into positive template image and negative norm plate image according to pixel value;

Positive and negative separation associated picture generates step, and the combination by described positive correlation value image and described negative correlation value image and described positive template image and described negative norm plate image generates a plurality of positive and negative separation associated pictures; And

Check determination step, use described positive and negative separation associated picture to check judgement.

15. an image is checked device, compares by the feature to image between a plurality of template images of the input picture of circular object and registered in advance, thereby image is checked, and it is characterized in that this image is checked device and comprised:

Polar coordinate transform image production part spare is carrying out on the basis of polar coordinate transform described input picture and described template image, generates ρ-θ input picture and ρ-θ template image after rotation with two images departs from correction;

Whether correlation separation of images parts generate the correlation image by described ρ-θ input picture and described ρ-θ template image, be more than the threshold value and with this correlation separation of images to be positive correlation value image and negative correlation value image according to pixel value;

Whether the template image separating component is more than the threshold value described ρ-θ template image to be separated into positive template image and negative norm plate image according to pixel value;

Positive and negative separation associated picture generates parts, and the combination by described positive correlation value image and described negative correlation value image and described positive template image and described negative norm plate image generates a plurality of positive and negative separation associated pictures; And

Check judging part, use described positive and negative separation associated picture to check judgement.

16. image as claimed in claim 15 is checked device, it is characterized in that, described positive and negative separation associated picture generates parts and generates positive feature regional images and negative feature regional images, described positive feature regional images with to long-pending calculating of the described positive correlation value image of each pixel and described positive template image and the value that obtains as pixel value, described negative feature regional images with to long-pending calculating of the described negative correlation value image of each pixel and described positive template image and the value that obtains as pixel value.

17. image as claimed in claim 15 is checked device, it is characterized in that, described positive and negative separation associated picture generates parts and generates positive background area image and negative background area image, described positive background area image with to long-pending calculating of the described positive correlation value image of each pixel and described negative norm plate image and the value that obtains as pixel value, described negative background area image with to long-pending calculating of the described negative correlation value image of each pixel and described negative norm plate image and the value that obtains as pixel value.

18. image as claimed in claim 15 is checked device, it is characterized in that, described positive and negative separation associated picture generates parts and generates positive feature regional images, negative feature regional images, positive background area image and negative background area image, described positive feature regional images with to long-pending calculating of the described positive correlation value image of each pixel and described positive template image and the value that obtains as pixel value, described negative feature regional images with to long-pending calculating of the described negative correlation value image of each pixel and described positive template image and the value that obtains as pixel value, described positive background area image with to long-pending calculating of the described positive correlation value image of each pixel and described negative norm plate image and the value that obtains as pixel value, described negative background area image with to long-pending calculating of the described negative correlation value image of each pixel and described negative norm plate image and the value that obtains as pixel value.

19. check device as claim 16,17 or 18 described images, it is characterized in that, described positive and negative separation associated picture generates parts this surrounding pixel of gazing at the pairing respective pixel of pixel in the positive region image of gazing at pixel and the described positive correlation value image generation of use in the negative region image that uses described negative correlation value image generation is compared, gaze at greater than this at the pixel value of at least one this surrounding pixel under the situation of pixel value of pixel, carry out this is gazed at the expansion process that pixel moves to this respective pixel.

20. image as claimed in claim 15 is checked device, it is characterized in that, described ρ-θ input picture and described ρ-θ template image are to handle the edge image that has carried out image transformation by the edge extracting that uses the edge extracting operator.

21. image as claimed in claim 20 is checked device, it is characterized in that, described edge image is that the edge strength at the edge that will extract carries out standardization and the standardized images that obtains.

22. image as claimed in claim 15 is checked device, it is characterized in that, described template image is each the individual image for described circular object to be averaged and the average image that obtains.

23. image as claimed in claim 15 is checked device, it is characterized in that, described correlation image be with the correlation with each pixel of described ρ-θ input picture or described ρ-θ template image carry out standardization and the standardization correlation that obtains as the image of pixel value.

24. image as claimed in claim 15 is checked device, it is characterized in that, described check that judging part carries out to described positive and negative separation associated picture that piece is cut apart and the summation of calculating the pixel value in each piece as the piece value, thereby check judgement by whole described positive and negative separation associated pictures are calculated the value of checking with the long-pending Calais mutually of this piece value and weighting coefficient.

25. image as claimed in claim 15 is checked device, it is characterized in that, describedly checks judging part calculates described weighting coefficient by linear discriminant analysis value.

26. image as claimed in claim 15 is checked device, it is characterized in that, described ρ-θ input picture and described ρ-θ template image are parallel to be moved described polar coordinate transform image production part spare by making, thereby the rotation of proofreading and correct two images departs from.

27. image as claimed in claim 15 is checked device, it is characterized in that, described circular object is a coin.

28. an image checking method compares by the feature to image between a plurality of template images of the input picture of circular object and registered in advance, thereby image is checked, and it is characterized in that this image checking method comprises:

Polar coordinate transform image production part spare is carrying out on the basis of polar coordinate transform described input picture and described template image, generates ρ-θ input picture and ρ-θ template image after rotation with two images departs from correction;

Whether correlation separation of images step generates the correlation image by described ρ-θ input picture and described ρ-θ template image, be more than the threshold value and with this correlation separation of images to be positive correlation value image and negative correlation value image according to pixel value;

Whether the template image separating step is more than the threshold value described ρ-θ template image to be separated into positive template image and negative norm plate image according to pixel value;

Positive and negative separation associated picture generates step, and the combination by described positive correlation value image and described negative correlation value image and described positive template image and described negative norm plate image generates a plurality of positive and negative separation associated pictures; And

Check determination step, use described positive and negative separation associated picture to check judgement.

29. an image check program compares by the feature to image between a plurality of template images of the input picture of circular object and registered in advance, thereby image is checked, and it is characterized in that, this image check program makes computing machine carry out following steps:

Polar coordinate transform image production part spare is carrying out on the basis of polar coordinate transform described input picture and described template image, generates ρ-θ input picture and ρ-θ template image after rotation with two images departs from correction;

Whether correlation separation of images step generates the correlation image by described ρ-θ input picture and described ρ-θ template image, be more than the threshold value and with this correlation separation of images to be positive correlation value image and negative correlation value image according to pixel value;

Whether the template image separating step is more than the threshold value described ρ-θ template image to be separated into positive template image and negative norm plate image according to pixel value;

Positive and negative separation associated picture generates step, and the combination by described positive correlation value image and described negative correlation value image and described positive template image and described negative norm plate image generates a plurality of positive and negative separation associated pictures; And

Check determination step, use described positive and negative separation associated picture to check judgement.

Images