JP4103826B2 - Authenticity determination method, apparatus and program - Google Patents

Description

  The present invention relates to a true / false determination method, a true / false determination apparatus, and a program, and in particular, a true / false determination method for determining the authenticity of a solid in which readable and unique features having randomness are distributed along a surface, The present invention relates to a true / false determination apparatus to which the true / false determination method can be applied, and a program for causing a computer to function as the true / false determination apparatus.

  In recent years, counterfeiting and misuse of copy materials have been deterred against the backdrop of increasing use cases of copies of bills and securities copied with copy machines and printers as performance of copiers and printers has improved. In order to do so, the accuracy of various paper documents (in addition to the banknotes and securities mentioned above, such as passports, various rights documents, resident's cards, birth certificates, insurance certificates, guarantees, confidential documents, etc.) is determined with high accuracy. The establishment of the technology that can be done is awaited.

  As a technique for determining the authenticity of a paper document, Patent Document 1 provides a light emitting substance that emits light in a specific wavelength region when irradiated with light in a specific wavelength region on paper partially different in opacity, A technique is disclosed in which authenticity determination is performed by irradiating light of a specific wavelength onto the paper and receiving light emitted from a luminescent material by a light receiving sensor. Patent Document 2 discloses a technique for integrating a conductive labeling substance (transparent conductive polymer or conductive pigment or a combination thereof) into a paper web for documents, securities and banknotes. Yes.

  In Patent Document 3, frequency analysis is performed on image data obtained by light transmitted or reflected through a sample paper sample, integrated data for a plurality of wavelength ranges is obtained from the data after frequency analysis, and standard paper is obtained. A technique for determining whether a sample paper sample is the same as a standard paper sample by obtaining a correlation with the integrated data of the sample is disclosed.

Furthermore, Patent Document 4 uses a fact that the transparency of the paper is randomly changed due to the randomness of the entanglement of the fibrous material forming the paper, and a predetermined area of the tag T is defined as a plurality of squares. The transparency of six rectangular areas selected at random is divided into areas, and the detected transparency is recorded as information together with the addresses of the individual rectangular areas, and is specified by the recording information at the time of authenticity judgment. A technique is disclosed in which the authenticity is determined by detecting the transparency of each rectangular area and comparing the detection result with the transparency represented by the recorded information.
JP 2000-094865 A Japanese translation of PCT publication No. 2002-518608 JP 2000-146952 A Japanese Examined Patent Publication No. 6-16312

  However, since the techniques described in Patent Document 1 and Patent Document 2 prevent the forgery by adding a special substance to paper for authenticity determination, the cost increases and the authenticity determination is performed. In addition, there is a problem that a special device for detecting a special substance is required.

  In addition, the technique described in Patent Document 3 is a technique for identifying paper by utilizing the periodicity of a pattern such as a paper texture or a papermaking wire formed in the papermaking process, and this technique is used for authenticity determination of a paper document. If applied, for example, if a malicious person obtains paper with the same production lot as the original paper document and copies the original content, this copy will be erroneously determined as the original. is there.

  On the other hand, the technique described in Patent Document 4 uses the fact that the transparency of the paper changes randomly due to the randomness of the entanglement of the fibrous material forming the paper. There is no need to add a substance, and there is an advantage that individual papers can be distinguished even if the production lot is the same paper. However, since this technology compares the transparency of a minute area on the paper and performs authenticity determination, the position of the read area on the paper is different between reading for information recording and reading for authenticity determination. If the orientation is slightly different, there is a high possibility that the authenticity of the paper is erroneously determined. In addition, the detected value of transparency is also affected by a mismatch in the amount of light source at the time of reading, or discoloration of paper due to ultraviolet rays or the like. On the other hand, the technique described in Patent Document 4 does not consider any countermeasures against a decrease in the accuracy of authenticity determination due to the above-described event, and has a problem that the accuracy of authenticity determination is not sufficient. It was.

  The present invention has been made in consideration of the above-described facts, and an object thereof is to obtain a true / false determination method, a true / false determination apparatus, and a program that can easily and accurately determine the authenticity of a solid.

  The inventors of the present application believe that characteristics unique to a solid having randomness and distributed along the surface of the solid, such as the transparency of the paper changing at random, are useful for determining the authenticity of the solid. In order to improve the accuracy of true / false judgments using features unique to a solid object, the area of the comparison target area (the region on the solid where the unique features are distributed) for authenticity judgment The difference between the true solid and the solid to be judged, and the correlation value is repeatedly calculated while moving the small area within the large area. In addition to the above, a hypothesis was made that it would be effective to perform true / false judgments using feature quantities representing the degree of distribution of many correlation values obtained.

  There are two types of false judgment in authenticity determination: a true article is mistaken as a fake and a false article is mistaken as a genuine article (note that the probability of falsely judging a genuine article as a fake is FRR (: False Rejection Rate)). The probability of misjudging a fake as a real is called FAR (: False Acceptance Rate)). In order to verify whether the above hypothesis is effective even in the case where the above-mentioned hypothesis is highly likely to be erroneously determined as a fake and the high probability that a fake is erroneously determined to be a genuine article, The experiment was conducted.

  That is, first, a flatbed scanner reads a reference area of 32 × 32 dots (about 2 mm × about 2 mm) of an unprinted portion of paper (original) with a resolution of 400 dpi and a gray scale of 8 bits. The output image data (this image data represents a random change in the transparency of the paper in the reference area on the paper (original) due to the randomness of the entanglement of the fibrous material forming the paper) Stored as reference data. In FIG. 1A, the reference data is visualized as a “reference image” (contrast is corrected so that visual confirmation is easy).

  Since it is impossible to control the entanglement of the fibrous material forming the paper at the time of manufacture, the entanglement of the fibrous material forming the paper can be regarded as random. The entanglement of the fibrous material forming the paper can be observed using a transmission light microscope. On the other hand, in the “reference image” shown in FIG. 1 (A), although the entanglement of the fibrous material cannot be confirmed, it is caused by the randomness of the entanglement of the fibrous material (paper generated by various conditions when the paper is rolled) Since the pattern of random light and darkness that reflects random changes in the transparency of the paper (which may also be affected by unevenness on the surface) has occurred, the reference data corresponding to the reference image is the reference on the paper (original) It can be confirmed that the information is a characteristic unique to the paper (original) in the area, that is, information representing a random change in transparency in the reference area on the paper (original).

  Next, as a comparative example, a 64 × 64 dot (about 4 mm × about 4 mm) collation area (area including the above-described reference area) of the paper used as the original is read, and the image data output from the scanner Was stored as the first verification data. The first collation data represents a random change in the transparency of the paper in the collation area of the paper (original). In FIG. 1B, the first collation data is visualized as a “collation image”.

  Further, as a case where the probability that a genuine article is erroneously determined to be a fake is high, the paper used as the original is slightly shifted in position and slightly rotated with respect to the time when the first collation data is acquired, and placed on the scanner platen. In the mounted state, the collation area of 64 × 64 dots is read (the area and the direction slightly different from the reading area at the time of the first collation data acquisition is read), and from the scanner The output image data was stored as second verification data. Further, as another comparative example, a collation area of 64 × 64 dots out of paper different from the paper used as the original is read, and the image data output from the scanner is stored as third collation data.

  Next, correlation values between the first to third matching images represented by the first to third matching data and the reference image represented by the reference data were calculated. Specifically, as shown in FIG. 2 as an example, a partial area having the same size as the reference image (indicated as “correlation value calculation range” in FIG. 2) is extracted from the collation image to be calculated, The correlation value between the reference image and the reference image is calculated by the normalized correlation method (see the following equation (1)), and the position of the partial area on the collation image is shifted by 1 dot (pixel) in the X and Y directions. While repeating.

Where F is a reference image (a set of reference data), fi is a brightness value of each pixel of the reference image, N is the total number of pixels of the reference image (and a partial area of the collation image), and G is a partial area of the collation image ( ), Gi is the lightness value of each pixel in the partial region of the collation image, f _AVE is the average value of the lightness values of individual pixels in the reference image, and g _AVE is the lightness value of each pixel in the partial region of the collation image Is the average value. By performing each of the above calculations using the first to third matching images as the target images to be calculated, assuming that the number of dots in the reference image is m × n and the number of dots in the matching image is M × N, a single matching is performed. (M−m + 1) × (N−n + 1) correlation values are obtained per image.

Subsequently, the normalized score of the maximum value of the correlation value was calculated for each of the first to third matching images as a feature amount representing the distribution degree of the correlation value according to the following equation (2).
Normalized score = (maximum correlation value-average correlation value) / standard deviation of correlation value (2)
3A to 3B show the correlation value between the maximum correlation value and the normalized score calculation result of the correlation value, and the relationship between the position of the partial area on the collation image and the correlation value. It is shown with the chart shown in.

  As shown in FIG. 3A, when the collation area including the reference area on the same paper is read without any deviation in position and orientation, the maximum correlation value is very high. In addition, the distribution of correlation values also shows a value that is very low compared to the maximum value in the portion other than the peak portion where the correlation value is maximum. The normalized score is also very high. In addition, when a paper different from the original is read, the maximum correlation value is very low as shown in FIG. 3C, and the correlation value distribution as a whole, including the peak portion, is also shown. Since the correlation value indicates a low value, the normalized score of the maximum correlation value is also a very low value.

  On the other hand, when the collation area including the reference area on the same paper is read with a slightly different position and orientation (corresponding to the case where the probability that a genuine article is erroneously determined to be fake) is high, the maximum correlation value and the correlation value As shown in FIG. 3B, the maximum normalized score is an intermediate value between the case where the same paper is read without deviation in position and orientation and the case where different paper is read. Therefore, an intermediate value between the value shown in FIG. 3B and the value shown in FIG. 3C is adopted as the maximum correlation value and the normalized score threshold value of the maximum correlation value (for example, Correlation value threshold value ≈ 0.3, correlation value maximum value normalized score threshold ≒ 5.0), correlation value maximum value is compared with threshold value and correlation value maximum value normalized value If authenticity is determined by comparing the score with a threshold, there is a high probability that the paper will be misidentified as a fake, such as when the paper position and orientation at the time of collation area reading are slightly shifted. Therefore, it can be understood that the determination accuracy of the authenticity determination may be improved as compared with the case where the determination is performed using only the maximum correlation value.

  In addition, the inventors of the present application use the same scanner as the above experiment, and set an arbitrary 32 × 32 dot (about 2 mm × about 2 mm) reference area of A4 blank paper (original) with the same resolution and gradation. In addition to obtaining reference data by reading, as a first comparative example, the entire surface of the paper used as the original was read, and data of a collation area of 64 × 64 dots was extracted and extracted from the image data obtained by the reading. The calculation of the correlation value between the partial area data further extracted from the collation data and the reference data according to the equation (1) is repeated while shifting the position of the partial area within the collation area by one dot (by this, 10 million More than one correlation value was obtained).

  Further, as a second comparative example, the reading of the substantially entire surface of the paper used as the original is performed again after slightly shifting the position and slightly rotating the direction, and obtained by reading in the same manner as in the first comparative example. Extracting 64 × 64 dot collation area data from the obtained image data, and further calculating the correlation value between the partial area data extracted from the extracted collation data and the reference data according to the expression (1). The position of the partial area inside was repeated while shifting by one dot. Further, as a third comparative example, a paper different from the paper used as the original was used, and the reading / correlation value was calculated in the same manner as in the first and second comparative examples.

  Then, as a case where there is a high probability that a fake is erroneously determined to be genuine, the reference area of the paper used as the original is intentionally read with an excessive amount of light, and the change in transparency in the reference area is partially whitened. The second reference data representing the image that has been captured is acquired, the entire surface of the paper used in the third comparative example is read, and the data of the collation area of 64 × 64 dots is extracted from the image data obtained by the reading. Then, the correlation value between the partial region data further extracted from the extracted collation data and the second reference data is calculated by the normalized correlation method according to the equation (1). Repeated while shifting the dots.

  FIGS. 4 to 7 show the distribution of correlation values (charts with the horizontal axis representing the correlation value and the vertical axis representing the logarithm of the frequency) obtained by the above experiment. 4 shows the correlation value distribution obtained in the first comparative example, and FIG. 5 shows the correlation value distribution obtained in the second comparative example. In any distribution, the majority of the correlation values are 0 or a value close to 0. However, the maximum correlation value in the first comparative example is 1.00 and the second comparative example is also included. Since the maximum correlation value at 0.657 is 0.657, which is a high value, it can be understood that the real object can be determined as the true object even if only the maximum correlation value is used. FIG. 6 shows the distribution of correlation values obtained in the third comparative example. All correlation values are less than a predetermined value (for example, 0.3), and the maximum correlation value is a low value of 02.54. Thus, even if only the maximum correlation value is used as described above, it is possible to determine a fake as a fake.

  On the other hand, FIG. 7 shows a distribution of correlation values obtained by an experiment that assumes a high probability that a fake is erroneously determined to be a real thing. A high correlation value greater than or equal to a predetermined value (eg, 0.3 or more) is shown. The data shown is also included (the maximum correlation value is 0.348), and if the authenticity determination is performed using only the maximum correlation value, there is a possibility that a fake is erroneously determined to be a real object. On the other hand, as apparent from the comparison of the distribution of FIG. 7 with the distribution of FIG. 6, the correlation value distribution shown in FIG. The standard deviation of the correlation value in the distribution is larger than the distribution of FIG. 6, and the normalized score value of the maximum correlation value in the distribution of FIG. 6 is smaller than the distribution of 6 (the normalized score of the maximum correlation value in the distribution of FIG. 6 is 5.32 and the normalized score of the maximum correlation value in the distribution of FIG. 7 is 4.91). It can be understood that the erroneous determination can be avoided.

  As described above, even in a case where the probability that a fake is erroneously determined to be true is high (in the case of FIG. 7), the authenticity determination is performed using the maximum correlation value and the normalized score of the maximum correlation value. For example, it is possible to avoid erroneous determination. Therefore, in addition to the maximum correlation value, if a true / false determination is performed using a feature value indicating the distribution degree of the correlation value, such as a normalized score of the maximum correlation value. It was confirmed that the accuracy of the true / false judgment can be improved.

Based on the above, the true / false determination method according to the invention described in claim 1 is a true / false determination method for determining the authenticity of a solid in which readable and unique features having randomness are distributed along a surface. In addition to obtaining reference data representing features distributed on the true solid, obtained by reading the features of the true solid in advance, and reading the features of the solid to be judged, Collation data representing the feature distributed above is obtained, and based on the reference data and the collation data, a feature distributed in a first region of a predetermined size on one of the true solid and the solid to be determined is represented. Calculating a correlation value between the data and data representing characteristics distributed in a second region of the same size as the first region on the other solid, and determining the position of the second region on the other solid From the predetermined size Repeated while moving a large area, the maximum value of the plurality of correlation values obtained by the calculation is equal to or more than the first predetermined value, and the correlation value obtained by subtracting the average value of the correlation value from the maximum value of the correlation values The authenticity of the object to be determined is determined based on whether or not the normalized score of the maximum correlation value obtained by dividing by the standard deviation of the value is equal to or greater than a second predetermined value . .

  According to the first aspect of the present invention, the authenticity of a solid in which readable and unique features having randomness are distributed along the surface is determined. As a typical example of the solid according to the present invention, random changes in light transmittance or light reflectance are distributed along the surface as a unique feature due to the randomness of the entanglement of the fibrous material. Paper (for example, printed matter on which paper is printed in some detail, specifically, documents in which originals such as public documents and securities are present) can be mentioned, but the present invention has a unique characteristic that can be read with randomness. It can be applied to solids distributed along the surface. Specifically, for example, a card made of a synthetic resin in which random irregularities are distributed along the surface (for example, a telephone card or a highway card on the back surface). Random irregularities), CD (Compact Disc), DVD (Digital Versatile Disc), etc. (CD and DVD also have random irregularities on the surface).

  In the invention described in claim 1, reference data representing characteristics distributed on the true solid is obtained in advance by reading the characteristics of the true solid (that is, the original) in advance. In the invention described in claim 1, This reference data is acquired. The reference data may be carried on, for example, the solid itself (for example, recorded in a coded state), or the reference data is stored in a storage means separate from the solid, and the solid itself is stored in the storage means. Identification information for specifying the corresponding reference data among the stored data may be carried.

  According to the first aspect of the present invention, the collation data representing the characteristics distributed on the determination target solid is obtained by reading the characteristics of the determination target solid. When the characteristics of the solid are optically readable, the collation data (and the above-described reference data) is, for example, as described in claim 2 in which light is applied to the determination target solid (or true solid). Can be used, and image data obtained by reading reflected light or transmitted light can be used. In addition, the solid feature may be a feature that can be read by other reading methods (for example, a reading method using electromagnetic waves or magnetism in a wavelength region other than visible light).

According to the first aspect of the present invention, the characteristics distributed in the first region of a predetermined size on one of the true solid and the solid to be determined are represented based on the acquired reference data and the obtained collation data. The calculation of the correlation value between the data and the data representing the feature distributed in the second area of the same size as the first area on the other solid means that the position of the second area on the other solid is determined from the predetermined size. The maximum value of the plurality of correlation values obtained by calculation is repeated while moving within a larger area, and a value obtained by subtracting the average value of the correlation values from the maximum value of the correlation values. Based on whether or not the normalized score of the maximum correlation value obtained by dividing by the standard deviation of the correlation value is greater than or equal to a second predetermined value , the authenticity of the object to be determined is determined.

As described above, in the invention described in claim 1, since the authenticity of a solid is determined based on the maximum correlation value and the normalized score of the maximum correlation value, it is clear from the results of the above-described experiment. In addition, the authenticity of the solid can be determined with high accuracy. Further, it is not necessary to add a special substance to the solid for authenticity determination, and a special device for detecting the special substance is not required, so that the authenticity of the solid can be easily determined.

  When the solid feature is optically readable as described above, if the solid is a sheet-like or flat plate-like medium, for example, as described in claim 3, the solid feature is converted into a flat bed. It is preferable to read with a type scanner. Since the flood bed type scanner is widely used, it can be obtained at low cost, and the present invention can be easily implemented.

  In the first aspect of the present invention, any of the known calculation methods may be applied to the calculation of the correlation value. For example, as described in the fourth aspect, the correlation value is calculated by the normalized correlation method. It is preferable. As is clear from the above equation (1), in the normalized correlation method, the correlation value is calculated using the brightness value normalized by subtracting the average value of the brightness value from the brightness value to be calculated. For example, in an aspect in which solid features are optically read, even when the light source light amount is different between the reading for obtaining the reference data and the reading for obtaining the collation data, the difference in the light amount of the light source, etc. The resulting difference in brightness value is corrected by normalization, and an accurate correlation value can be obtained in which the influence of the difference in the amount of light source is eliminated.

  By the way, in the authenticity determination of a solid in which the characteristics of a solid that is optically readable with randomness are distributed along the surface, due to adhesion of dirt or the like on the true solid or the determination target solid If there is a shaded part, the reference data and collation data (representing the characteristics of the solid with randomness on the solid) obtained by optically reading the true solid or the solid to be judged are Noise components corresponding to the light and shade portions are mixed. Then, the above-described shade portions are present on the true solid and the determination target solid, respectively, and the features (eg, size, shape, concentration, etc.) of the shade portions existing on the determination target solid are true. If it is similar to the features of the light and shaded parts existing on the solid, the maximum correlation value and the feature value indicating the correlation value distribution will be a value indicating high correlation, so that the object to be judged is a fake However, there is a problem that the probability of being erroneously determined as a real product increases. For example, when a true solid is read to obtain reference data, a gray portion existing on the true solid is detected, and an area where the gray portion on the true solid does not exist is set as a reading range. This can be solved by setting, but if there is no area where there is no shading, it is difficult to set the reading range as described above, and the position of the reading range There is also a drawback that it cannot be applied to a mode in which is fixedly determined in advance.

In general, when there is a light and dark part due to adhesion of dirt or the like on a solid in which the characteristics of the solid having optical randomness and being optically readable are distributed along the surface, the solid is optically The noise component corresponding to the above-described shaded portion mixed in the reference data and collation data obtained by reading the image data is distributed within a certain gradation value range in the gradation value distribution of the reference data and collation data. ing. Based on the above, the invention according to claim 5 is the invention according to claim 2, in which at least one of the reference data and the collation data is estimated to contain a noise component based on a gradation value distribution. It is characterized in that a correlation value is calculated after setting a tone value range and excluding data belonging to the set tone value range.

According to the fifth aspect of the present invention, a gradation value range in which a noise component is estimated to be included based on a gradation value distribution is set for at least one of the reference data and the collation data, and the set gradation is set. Since data belonging to the value range is excluded, there is a shaded part caused by adhesion of dirt or the like on the true solid or the solid to be judged, and the noise component corresponding to this shaded part is the reference data and verification data. Even if it is mixed in, it is possible to prevent the feature value indicating the maximum correlation value or the correlation value distribution from being affected by this effect, and to prevent the value from showing a high correlation. The determination can be performed with high accuracy. Further, the gradation value range in which the noise component is distributed changes under the influence of a change in illumination conditions at the time of reading for obtaining the reference data or the collation data. However, in the invention according to claim 5 , A gradation value range that is estimated to contain a noise component based on the distribution of gradation values is set for at least one of the data and the collation data, so that it is not affected by variations in the illumination conditions, etc. The tone value range estimated to contain the noise component can be set appropriately, and the data corresponding to the noise component can be accurately excluded from the calculation target of the correlation value.

The noise component according to the invention described in claim 5 is not limited to the noise component corresponding to the light and shade portion caused by dirt or the like attached on the solid, for example, a character portion printed on paper as a solid This data is also a noise component that adversely affects solid authenticity determination, and the noise component corresponding to this character portion is also excluded from the correlation value calculation target according to the invention described in claim 5. be able to.

In the invention according to claim 5 , the setting of the gradation value range estimated to contain noise components and the exclusion of data belonging to the set gradation value range may be performed only on the reference data. The calculation of the correlation value in the present invention is distributed in the first region on one solid, although it may be performed only on the collation data or may be performed on each of the reference data and the collation data. Calculating the correlation value between the data representing the feature to be detected and the data representing the feature distributed in the second region on the other solid, and determining the position of the second region on the other solid to be larger than a predetermined size. In consideration of what is performed by repeating while moving within, the exclusion of data belonging to the gradation value range estimated to contain noise components is performed on data representing features distributed in the first region. The operation is negative But desirable because it is reduced. In addition, the reference data obtained by optically reading the characteristics of the true solid is temporarily recorded on a predetermined medium (the solid itself or a storage medium separate from the solid) and then read out from the predetermined medium. In order to reduce the capacity of the reference data, it is desirable to apply the reference data as data representing the characteristics distributed in the first region, and in this case, the floor that is estimated to contain noise components is used. It is desirable to exclude data belonging to the gradation range only for the reference data.

As described above, the true solid feature is obtained by optically reading the reference data recorded on the predetermined medium and read out from the predetermined medium. In the case of excluding data belonging to the gradation value range estimated to contain noise, for example, the reference data after excluding data belonging to the gradation value range estimated to contain noise components is used. Although it may be recorded on a predetermined medium, considering that there is a possibility that the reference data recorded on the predetermined medium may be used for applications other than solid authenticity determination, for example, as described in claim 6 , Based on the distribution of gradation values of the reference data acquired by reading from a predetermined medium, a gradation value range that is estimated to contain noise components with respect to the reference data is set, and the set gradation value range Belongs to Data after excluding from the reference data, it is preferable to perform the calculation of the correlation values. According to the sixth aspect of the present invention, data belonging to the gradation value range estimated to contain a noise component is excluded from the reference data read from the predetermined medium, and therefore recorded on the predetermined medium. The reference data to be included includes data belonging to the gradation value range, and the reference data read from a predetermined medium can be used for purposes other than solid authenticity determination. That is, when an area printed with characters or figures such as serial numbers and barcodes is used as the reference data, the ink or toner-covered part becomes a noise component for authenticity determination, but other than authenticity determination Can be important information for the application.

In the invention according to claim 5 or claim 6, as the gradation value range estimated to contain a noise component, for example, as described in claim 7 , the maximum or minimum gradation value The range from the value until the cumulative frequency reaches the predetermined value, or the average value of the gradation values is AVE, the standard deviation of the distribution of gradation values is σ, and the predetermined value is n, the gradation value is AVE + nσ or more Or the range below AVE-n (sigma) can be set.

The true / false determining apparatus according to the invention described in claim 8 is a true / false determining apparatus for determining the authenticity of a solid in which readable and unique features having randomness are distributed along a surface. By obtaining the reference data representing the characteristics distributed on the true solid obtained by reading the characteristics of the solid in advance, and by reading the characteristics of the determination target solid, Reading means for obtaining collation data representing a feature distributed in a feature, and features distributed in a first region of a predetermined size on one of a solid that is a true solid and a solid to be determined based on the reference data and the collation data And calculating the correlation value between the data representing the feature distributed in the second region of the same size as the first region on the other solid, and calculating the correlation value of the second region on the other solid Position the specified size Calculating means for repeatedly while also moving in a large region Ri, the maximum value of the computed plurality of correlation values at a first predetermined value or more, and, by subtracting the average value of the correlation value from the maximum value of the correlation values and normalized score of the maximum value of the correlation value obtained by dividing the value by the standard deviation of the correlation value is based on whether the second predetermined value or more, determination means for determining authenticity of the determination target solid Thus, as in the first aspect of the invention, the authenticity of the solid can be determined easily and with high accuracy.

According to a ninth aspect of the present invention, there is provided a program for reading a computer connected to a reader capable of reading a characteristic unique to a solid distributed along the surface of the solid and having randomness, and to read a characteristic of the true solid in advance. The acquisition means for acquiring the reference data representing the characteristics distributed on the true solid obtained by the processing, the characteristics of the determination target solid being read by the reader, and the distribution on the determination target solid Read control means for obtaining collation data representing the feature to be represented, and representing features distributed in a first region of a predetermined size on one of the solids to be determined and the solids to be determined based on the reference data and the collation data Computing a correlation value between the data and data representing features distributed in a second region of the same size as the first region on the other solid, the first on the other solid Calculating means for repeatedly while the position of the region is moved at the predetermined large area than the size, and the maximum value of the calculated plurality of correlation values at a first predetermined value or more, and, from the maximum value of the correlation values Based on whether or not the normalized score of the maximum correlation value obtained by dividing the value obtained by subtracting the average value of the correlation values by the standard deviation of the correlation values is equal to or greater than the second predetermined value , the object to be determined It is made to function as a determination means for determining the true or false of.

According to a ninth aspect of the present invention, there is provided a program connected to a reading device capable of reading a characteristic unique to a solid distributed along the surface of the solid and having randomness. Since it is a program for causing the computer and the determination unit to function, the computer and the reading device function as the authenticity determination device according to the eighth aspect when the computer executes the program according to the ninth aspect. Thus, as in the first and eighth aspects of the invention, the authenticity of a solid can be determined easily and with high accuracy.

As described above, the present invention acquires reference data obtained by reading in advance a unique readable characteristic having randomness distributed along the surface of a true solid, The collation data is obtained by reading the feature of the solid, and the data representing the feature distributed in the first region of a predetermined size on one of the true solid and the solid to be determined, and the first on the other solid The calculation of the correlation value with the data representing the characteristics distributed in the second area of the same size as the area is repeated while moving the position of the second area on the other solid within the area larger than the predetermined size. The maximum value of the plurality of correlation values obtained by the calculation is not less than the first predetermined value, and the value obtained by subtracting the average value of the correlation values from the maximum value of the correlation values is divided by the standard deviation of the correlation values. Of the maximum correlation value Maraizudo score is based on whether the second predetermined value or more, since determining the authenticity of the determination target solid, it is possible to perform the authenticity determination of the solid easily and highly accurately, an excellent effect that Have.

  Hereinafter, an example of an embodiment of the present invention will be described in detail with reference to the drawings. FIG. 8 shows a color printer 10 according to the present embodiment. The color printer 10 includes a photosensitive drum 12 as an image carrier, and the photosensitive drum 12 is charged by a charger 14. Above the photosensitive drum 12, a light beam scanning device that emits a light beam that is modulated according to the image to be formed and deflected along the main scanning direction (direction parallel to the axis of the photosensitive drum 12). 16 is arranged. The light beam emitted from the light beam scanning device 16 scans the circumferential surface of the photosensitive drum 12 in the main scanning direction, and at the same time, the photosensitive drum 12 is rotated to perform sub-scanning. An electrostatic latent image is formed on the peripheral surface.

  Further, a multi-color developing unit 18 is disposed on the right side of the photosensitive drum 12 in FIG. The multicolor developing unit 18 includes developing units 18A to 18D loaded with toners of any one of C (cyan), M (magenta), Y (yellow), and K (black). The electrostatic latent image formed in (1) is developed into one of C, M, Y, and K colors. In the formation of a full-color image in the color printer 10, an electrostatic latent image is formed on the same area on the photosensitive drum 12 and developed in different colors, and each color is formed on the area. The toner images are sequentially superimposed.

  An endless transfer belt 20 is disposed in the vicinity of the photosensitive drum 12, and a sheet tray 24 for storing the recording sheet 22 is disposed below the position where the transfer belt 20 is disposed. The peripheral surface of the transfer belt 20 is in contact with the peripheral surface of the photosensitive drum 12 on the downstream side of the developing position by the multicolor developing unit 18 along the rotation direction of the photosensitive drum 12, and is formed on the photosensitive drum 12. The toner image thus transferred is temporarily transferred to the transfer belt 20, and is then transferred again to the recording paper 22 that is pulled out from the paper tray 24 and conveyed to the position where the transfer belt 20 is disposed. A fixing device 26 is disposed in the middle of the conveyance path of the recording paper 22 toward the outside of the machine to the color printer 10, and the toner image is fixed on the recording paper 22 to which the toner image is transferred by the fixing device 26. Later, it is discharged out of the machine body to the color printer 10.

  In addition, a reading unit 28 is provided in the middle of the conveyance path (indicated by an imaginary line in FIG. 8) of the recording paper 22 from the paper tray 24 to the position where the transfer belt 20 is arranged. The reading unit 28 includes a light emitter 28A for irradiating the recording paper 22 with light and a light receiver 28B for receiving the light emitted from the light emitter 28A and reflected by the recording paper 22, and the signal output from the light receiver 28B. Is converted to digital data and output (not shown), and distributed along the surface of the recording paper 22 due to the randomness of the entanglement of the fibrous material forming the recording paper 22 It is possible to read a random change in the light reflectance at a predetermined resolution (for example, 400 dpi) and a predetermined gradation (for example, 8-bit gray scale).

  A printer controller 30 is connected to the light beam scanning device 16. The printer controller 30 is connected to an operation unit (not shown) including a keyboard and a display, and a reading unit 28, and a personal computer (not shown) for inputting data to be printed on the recording paper 22. Are omitted) or connected via a network such as a LAN. The printer controller 30 includes a microcomputer, and controls the operation of each unit of the color printer 10 including the light beam scanning device 16.

FIG. 9 shows a personal computer (PC) 32 and a scanner 34 that can function as a true / false determination apparatus according to the present invention. Although not shown, the PC 32 includes a CPU, a ROM, a RAM, and an input / output port, which are connected to each other via a bus. In addition, a display, a keyboard, a mouse, and a hard disk drive (HDD) are connected to the input / output ports. The HDD stores an OS and various application software programs, and also stores a true / false determination program for performing a true / false determination process described later. This authenticity determination program corresponds to the program according to the ninth aspect of the invention.

  On the other hand, the scanner 34 is a flat bed type, and an original placed on an original table (not shown) is scanned with the same resolution (for example, 400 dpi) and the same gradation (for example, 8-bit gray scale) as the above-described reading unit 28. ). The scanner 34 is connected to an input / output port of the PC 32, and reading of the original by the scanner 34 is controlled by the PC 32, and image data obtained by the scanner 34 reading the original is input to the PC 32.

  Next, as an operation of this embodiment, processing in the color printer 10 will be described first. When the document to be printed on the recording paper 22 is an original, the color printer 10 according to the present embodiment performs printing as the original (reference data for use in authenticity determination of the document is also printed on the recording paper 22. Has a function. When printing using the color printer 10, the user transmits print data representing a document to be printed on the recording paper 22 from the PC to the color printer 10 and the document to be printed is a document used as an original. The color printer 10 is instructed to print the document to be printed as an original.

  When the above instruction is given, the printer controller of the color printer 10 performs reference data registration processing. Hereinafter, the reference data registration process will be described with reference to the flowchart of FIG. In step 100, the recording paper 22 on which the original document is printed is taken out from the paper tray 24 and conveyed to the arrangement position (reading position) of the reading unit 28. When the recording sheet 22 reaches the reading position, in the next step 102, the reading unit 28 causes the reading unit 28 to set a predetermined reference area (32 on the recording sheet 22 with a predetermined resolution (400 dpi) and a predetermined gradation (8-bit gray scale). Read x32 dots (area of size of about 2 mm x about 2 mm).

  Thereby, the reading unit 28 randomly changes the transparency of the paper in the reference region of the recording paper 22 to be read due to the randomness of the entanglement of the fibrous material forming the recording paper 22 to be read. Will be output. In this embodiment, since the reading resolution is 400 dpi, the reading gradation is 8-bit gray scale, and the reference area to be read is 32 × 32 dots, the size of the reference data is 1024 bytes, and each pixel (dot) is The gradation value (brightness value) is an integer value in the range of 0 to 255. FIG. 11 shows an example of an image obtained by visualizing the image represented by the reference data (contrast correction so that the visual observation is easy) based on the reference data obtained by the above reading.

  The reference area may be an arbitrary position on the recording paper 22, the position of the reference area on the recording paper 22 may be fixed, or the position of the reference area on the recording paper 22 may be changed depending on the document. Also good. However, when toner (or ink) adheres to the reference area on the recording paper 22 by printing after reading the reference area, the maximum correlation value calculated in the authenticity determination described later is greatly reduced. There is a very high possibility of erroneous determination. For this reason, when the position of the reference area is fixed, the position on the recording paper 22 where the toner is not likely to adhere (for example, a position corresponding to a position outside the printable range of the color printer 10) is set. When changing according to the document, it is desirable to determine a range of the recording paper 22 where toner or the like is not attached by printing based on the print data, and set the reference region within the determined range. In particular, in the authenticity determination process described later, an area wider than the reference area (for example, an area of 64 × 64 dots) is read as a collation area. Therefore, the reference area may be an area where toner or the like is not attached to the surrounding area by printing. desirable.

  The reference area can also be read after printing on the recording paper 22. In this case, the reference area includes a portion of the recording paper 22 to which toner or the like is attached by printing. Even if the toner is adhered to the reference area on the recording paper 22 by the printing performed after the reference area is read as described above, the false determination is made by the true / false determination. Although it is unlikely to occur, the change in transparency of the part where the toner, etc. on the paper is attached is not random (change unique to each paper), and the part where the change in transparency is not random is set as the reference area However, if the reference data obtained by reading the reference area is used for authenticity determination, it becomes vulnerable to forgery. Therefore, even when the reference area is read after printing on the recording paper 22 is performed. , Standard Frequency is preferably set to within a range in which toner or the like on the paper is not attached.

  When the reference area is read after printing on the recording paper 22, it is possible to determine the range where the toner on the recording paper 22 is not attached by using the print data as described above. The portion of the recording paper 22 to which the toner or the like is attached has a clear contrast compared with the portion to which the toner or the like is not attached, so that instead of using the print data as described above, the recording is performed. By reading the paper 22 and obtaining the contrast (difference between the maximum value and the minimum value of the gradation value (brightness value or density value)) for each part on the recording paper 22 based on the data obtained by the reading, It is also possible to determine a range where the toner or the like on the recording paper 22 is not attached.

  In general, as the size of the region to be read (specifically, the region for which the correlation value is calculated in the authenticity determination) increases, the accuracy of the determination of authenticity improves (at least one of FAR and FRR decreases). However, since it is necessary to make the area of the recording paper 22 to which the toner or the like is not attached even if printing is performed on a larger area, the degree of freedom of printing is reduced, and the processing such as authenticity determination is complicated. Problems arise. For this reason, in this embodiment, the size of the reference area at a reading resolution of 400 dpi is set to 32 × 32 dots (about 2 mm × about 2 mm). As will be apparent from the experimental results described later, if the reference area is made smaller than the above size, the accuracy of determination of authenticity decreases, but even if the reference area is made larger than the above size, the degree of improvement in the judgment accuracy is slight. It is. Therefore, it is not necessary to use an expensive and troublesome microscope for reading, and a reading device (such as the reading unit 28 built in the color printer 10 or an inexpensive commercially available scanner) capable of reading at a resolution of about 400 dpi is used. It is practical to use.

  Further, in reading the reference area, if the output signal of the light receiver 28B is saturated due to excessive light quantity incident on the light receiver 28B, the transparency of the reference area represented by the reference data obtained by the reading is reduced. Since reference data that accurately represents the change in transparency in the reference area cannot be obtained, such as when the change is partially whitened, it is desirable to moderate the exposure when reading the reference area. Further, when reading is performed using a scanner provided with a photo mode / document mode or the like as a reading mode instead of the reading unit 28 built in the color printer 10, the change in the transparency of the paper is further increased. It is desirable to perform reading by selecting a reading mode (for example, a photo mode) capable of fine reading.

  When the reference area is read as described above, in step 104, the reference data obtained by the reading is compressed by applying a discrete cosine transform or the like, and in the next step 106, based on the compressed data. Bitmap data for printing the data on the recording paper (original) 22 as a code (for example, a two-dimensional barcode) in a format that can be automatically read by the machine is generated. Note that data compression in step 104 is not essential, and encoding may be performed without performing data compression. Further, when the position of the reference area is changed depending on the document, it is preferable to perform compression / coding after adding information indicating the position of the reference area to the reference data obtained by reading. Data encryption may also be performed.

  In the next step 108, the bitmap data to be printed (the print data received by the color printer 10 from the PC is converted into bitmap data so that the code representing the reference data is printed at a predetermined position on the recording paper (original) 22. The bitmap data generated in step 106 is added to (obtained by developing). In step 110, the bitmap data is output to the light beam scanning device 16 when printing on the recording paper (original) 22. As a result, a document that the user desires to print as an original is printed on the recording paper (original) 22 with a code representing the reference data added to a predetermined position.

  It should be noted that if the recording paper 22 on which the original document is printed has a stain such as ink adhering to the area that has been read as the reference area, the determination accuracy in the authenticity determination described below will be reduced. There is a problem of doing. For this reason, when printing a document as an original, for example, by simultaneously printing a mark or the like that clearly indicates an area that has been read as a reference area, the user is warned to prevent the area from being contaminated. Is preferred. On the other hand, since it is effective for preventing counterfeiting to not explicitly indicate the read area as the reference area, the area may not be intentionally specified for the purpose of preventing forgery.

  In addition, in order to avoid a decrease in the accuracy of determination of authenticity even when dirt or the like is attached to the area read as the reference area, a plurality of reference areas are set, and each individual reference area is read, It is preferable to store a plurality of reference data obtained by reading. As a result, even when dirt or the like adheres to a part of the plurality of areas read as the reference area, this area is excluded, and authenticity determination is performed using another area to which no dirt or the like is attached. It is possible to avoid a decrease in the accuracy of the true / false determination.

  Next, the authenticity determination process executed by the PC 32 when determining the authenticity of paper (document) on which a code is printed at a predetermined position will be described with reference to the flowchart of FIG. In this authenticity determination process, for example, when a user who wishes to confirm the authenticity of the document is instructed to execute authenticity determination, the authenticity determination program is read from the HDD of the PC 32 and read. This is realized by executing the issued authenticity determination program by the CPU of the PC 32.

  In step 120, a message requesting to set the document for authenticity determination on the scanner 34 (placed on the document table) is displayed on the display, so that the document for authenticity determination is set on the scanner 34. In step 122, it is determined whether the document setting is completed, and step 122 is repeated until the determination is affirmed. When the document for authenticity determination is set in the scanner 34, the determination in step 122 is affirmed and the process proceeds to step 124, and the scanner 34 is instructed to read the document placed on the platen.

  As a result, the entire surface of the document subject to authenticity determination is read by the scanner 34 with the same resolution (400 dpi) and the same gradation (8-bit gray scale) as when reading the reference area, and the image data obtained by the reading is read. Is input from the scanner 34 to the PC 32. Even in this reading, it is desirable to moderate the exposure so as to obtain image data that accurately represents the change in transparency of the document to be judged as authenticity, particularly in the collation area. When a photographic mode / document mode or the like is provided as a reading mode of the scanner 34, a reading mode (for example, a photographic mode) capable of reading a change in transparency of paper with higher definition can be selected as the reading mode. desirable.

When the image data is input from the scanner 34, in the next step 126, data of an area where a code representing the reference data is printed is extracted from the input image data. Further, in step 128, based on the data extracted in step 126, the data represented by the code printed on the authenticity determination target document is recognized, and the recognized data is decompressed (decrypted if encrypted). The reference data is restored by performing a process such as Note that the above-described steps 124 to 128 correspond to the “acquisition of reference data” step and the acquisition means described in claim 8 together with the scanner 34 that actually reads.

In the next step 130, from the image data input from the scanner 34, the collation area (therefore, this area center position matches the center position of the reference area and has a larger area (64 × 64 dots) than the reference area). The matching area includes the reference area). When the position of the reference area is changed depending on the document, the position of the reference area can be recognized based on information indicating the position of the reference area added to the reference data, for example. This step 130 corresponds to the step of “obtaining collation data” according to the present invention and the reading means described in claim 8 together with step 124 described above and the scanner 34 that actually performs reading.

  Also, instead of recognizing the position of the reference area based on the information added to the reference data, after printing some mark in the vicinity of the reference area at the time of printing and reading for authenticity determination, The position of the reference area may be automatically recognized by searching for the mark on the image data obtained by reading. As a result, even if there is a slight misalignment in the authenticity determination target document placed on the document table at the time of reading for authenticity determination, the reference region is not affected by this misalignment. The position can be recognized accurately.

  The mark may be a point shape, for example. Also, if a plurality of marks are printed at positions where they do not overlap (the number of marks is preferably as small as possible, the optimum number is two), the positional relationship between each mark and the reference area is known. If so, the position and orientation (angle) of the reference area can be specified from the positions of the plurality of marks. The mark can be detected as follows, for example.

  That is, as a result of searching for a mark on the image data, for example, when one point that can be regarded as a mark is detected, the detection failure or the reading of the reference area is not performed (not a document printed as an original) to decide. For example, when two points that can be regarded as marks are detected, the Euclidean distance between the two marks is obtained, and if it is within the allowable range, it is determined that the mark indicates the reference region, and if it is out of the allowable range Judged as detection failure. When three or more points that can be regarded as marks are detected, the Euclidean distance between the marks is obtained, and if there is one pair of marks within the allowable range, the mark pair is a mark indicating the reference region. Judge. If there are 0 and 2 or more pairs of marks whose distance is within the allowable range, it may be determined that the detection has failed, or a pair whose distance is close to the allowable range may be used as a candidate. In the present invention, since the FAR can be made extremely low by appropriately setting a threshold value for authenticity determination, even if a point that is not actually a mark indicating the reference area is erroneously determined as a mark indicating the reference area, the processing time However, the accuracy of the true / false judgment is hardly adversely affected.

  By the way, in the true / false determination process according to the present embodiment, an area having the same size as the reference area (corresponding to the first area according to the present invention) from the data of the collation area (calculation target area: equivalent to the second area according to the present invention). The data corresponding to () is taken out and the correlation value between the data and the reference data is calculated, while the position of the calculation target region is moved. Therefore, in the next step 132, the data extraction position (the position of the calculation target area) in the collation area is initialized.

  In step 134, data (collation data) of a region having the same size as the reference region located at the set data extraction position is extracted from the collation region data. In step 136, the correlation value between the reference data restored in step 128 and the collation data extracted in step 134 is calculated by the normalized correlation method according to the above equation (1), and the correlation value obtained by the calculation is stored in the RAM. And so on.

In the next step 138, it is determined whether or not the calculation target area has scanned the entire surface of the collation area. If the determination is negative, the process proceeds to step 140, the data extraction position is moved vertically or horizontally by 1 dot, and then the process returns to step 134. Thereby, step 134 to step 140 are repeated until the determination in step 138 is affirmed. In this embodiment, since the reference area is 32 × 32 dots and the collation area is 64 × 64 dots, the correlation value is calculated (64−32 + 1) × (64−32 + 1) = 1089 times, and 1089 correlation values are obtained. Will be obtained. Steps 132 to 140 described above correspond to the “repeating correlation value calculation” step according to the present invention and the calculation means described in claim 8 .

  When the calculation of the correlation value is completed, the determination in step 138 is affirmed and the process proceeds to step 142, and the maximum value is extracted from the large number of correlation values obtained by the above calculation. In the next step 144, after calculating the standard deviation and average value of a large number of correlation values, the calculated standard deviation / average value and the maximum value of the correlation value obtained in step 142 are expressed by the above equation (2). By substituting for each, the normalized score of the maximum correlation value is calculated.

In step 146, it is determined whether or not the maximum correlation value obtained in step 142 is equal to or greater than a threshold value and the normalized score calculated in step 144 is equal to or greater than the threshold value. For example, “0.3” can be used as the threshold value of the maximum correlation value, and “5.0” can be used as the threshold value of the normalized score. If the determination in step 146 is affirmed, the process proceeds to step 148, and the determination result is output by displaying a message indicating that the document to be determined is “true” on the display. The process ends. If the determination in step 146 is negative, the process proceeds to step 150, and the determination result is output by displaying a message indicating that the document to be determined is “fake” on the display. The process ends. Thereby, the authenticity of the document (paper) subject to authenticity determination can be determined with high accuracy by simple processing. Steps 142 to 150 described above correspond to the “determining the authenticity of the determination target solid” according to the present invention and the determination means according to claim 8 .

  Further, in the above authenticity determination process, when a plurality of reference areas are set, the processes in steps 130 to 150 may be performed for each reference area. In addition, since there is a possibility that dirt due to ink or the like may adhere to the plurality of reference areas, for example, the data obtained by reading the collation areas corresponding to the individual reference areas, When the maximum value and the minimum value of the gradation value are obtained, and the difference between the maximum value and the minimum value of the gradation value exceeds a predetermined threshold value or a threshold value dynamically calculated from data obtained by reading. May issue a warning or exclude the corresponding reference area and collation area and perform the authenticity determination.

  In the above description, the correlation value is calculated using the data of all pixels in the calculation target area cut out from the reference area and the matching area. However, the present invention is not limited to this, and the original document is printed. As shown in FIG. 13, considering the possibility that there is a dark and dark part due to the adhesion of ink (or toner) or dirt in the read area of the recording paper 22 as the reference area. Based on the gradation value distribution of the reference data, a gradation value range that is estimated to include a noise component corresponding to the above-described shade portion is set, and data of pixels belonging to the set gradation value range is set After the above is excluded from the reference data, the correlation value may be calculated.

  The only difference between the authenticity determination process shown in FIG. 13 and the authenticity determination process shown in FIG. 12 will be described. In the authenticity determination process of FIG. The distribution of gradation values of the reference data is analyzed by creating a histogram of gradation values. To create a histogram of gradation values, for example, a plurality of gradation value ranges are set, data of a single pixel is extracted from the reference data, and the gradation value of the pixel is set to any of a plurality of gradation value ranges set in advance. It is possible to repeat incrementing the count value of the pixel corresponding to the gradation value range determined to belong to 1 for all the pixels of the reference data. An example of the distribution of gradation values generated by performing this processing is shown in Table 1 below.

In Table 1, the range of 0 to 255 gradation values (lightness values) represented by 8-bit data per pixel is divided into 29 gradation value ranges, and the number of pixels for each gradation value range is counted. An example of the result is shown.

  In the next step 162, based on the distribution of gradation values of the reference data obtained in step 160, the maximum gradation in which the cumulative frequency from the minimum value of the gradation value (lightness value) in the reference data does not exceed a predetermined value. A value threshold is extracted. For example, in the gradation value distribution shown in Table 1, when the predetermined value is 3%, a gradation value = 183 where the cumulative frequency from the minimum value of the gradation value is 1.99% is extracted as the gradation value threshold, When the value is 5%, a gradation value = 188 at which the cumulative frequency from the minimum value of the gradation value is 4.66% is extracted as the gradation value threshold value. In step 164, the pixel data belonging to the gradation value range from the minimum gradation value to the gradation value threshold obtained in step 162 in the reference data is excluded from the correlation value calculation target. Mark. Thus, for example, in the gradation value distribution shown in Table 1, when the predetermined value is 3%, the data of 326 pixels with the gradation value ≦ 183 is excluded from the correlation value calculation target, and the predetermined value is 5%. In this case, data of 763 pixels having a gradation value ≦ 188 is excluded from the correlation value calculation target.

  In the authenticity determination process of FIG. 13, instead of simply calculating the correlation value (step 136) as in the authenticity determination process of FIG. 12, in step 166, the correlation value is calculated from the reference data. Non-marking pixel data not marked for exclusion and non-marking data (non-marking data extracted from the matching area data and having the same size as the reference area located at the set data extraction position) A correlation value is calculated by a normalized correlation method using only pixel data corresponding to the marking pixel, and the correlation value obtained by the calculation is stored in a RAM or the like. In step 166, the correlation value can be obtained by substituting only the data of the pixel for which the correlation value is to be calculated into the equation (1), but more simply, the total number of non-marking pixels in the reference data is expressed as n. The sum of the gradation values of the non-marking pixels is msum, the sum of the squares of the gradation values of the non-marking pixels is msum2, and the sum of the gradation values of the pixels corresponding to the non-marking pixels in the collation data is tsum, When the sum of squares of the gradation values of the corresponding pixels is tsum2, and the sum of the products of the gradation values lum1 of the non-marking pixels and the gradation value lum2 of the pixels corresponding to the non-marking pixels in the matching data is mtsum, It can also be obtained by performing the calculation of equation (2).

As described above, the tone value range (the tone value from the minimum tone value to the tone value threshold value) that is estimated to contain noise components based on the tone value distribution with respect to the reference data set the range), to exclude the data of the pixels corresponding to the non-marking pixels in the data and verification data of the pixels belonging to the tone value range set (unmarked pixels) from the calculation target of the correlation values, claim 5 This corresponds to the invention described in ( 6 ). As a result, even when the recording paper 22 on which the original document is printed has a light and dark part due to adhesion of ink (or toner) or dirt in the area read as the reference area. Correlation value even though the document to be determined is “fake” due to the presence of a similar shaded part in the collation area of the document to be determined whether it is “fake” Since it is possible to prevent the normalized score of the maximum value of correlation and the maximum value of correlation values from exceeding a threshold value, it is possible to perform true / false determination with high accuracy.

  Also, the gradation value range in which the noise component is distributed in the reference data changes due to the influence of a change in illumination conditions at the time of reading the reference area, but in the above, the gradation value (lightness value) in the reference data Extracts the maximum gradation value threshold value whose cumulative frequency from the minimum value does not exceed the predetermined value, and calculates the correlation value for the data of the pixels belonging to the gradation value range from the minimum gradation value value to the gradation value threshold value Since it is excluded from the target, it is possible to appropriately set the gradation value range that is estimated to contain noise components without being affected by fluctuations in lighting conditions during reading of the reference area, and so on. Corresponding data can be accurately excluded from the calculation target of the correlation value.

  Furthermore, in the above description, an example in which the gradation value range estimated to contain noise components and the exclusion of data belonging to the set gradation value range is performed only on the reference data has been described. However, it may be performed only on the collation data, or may be performed on each of the reference data and the collation data. However, when the gradation value range is set and the data belonging to the gradation value range is excluded from the collation data, each time the position of the calculation target area for extracting data as the collation data is moved from the collation area, Since it is necessary to repeat the above processing for the area to be calculated after movement, the calculation load is reduced by setting the gradation value range and excluding data belonging to the gradation value range to the reference data. So desirable. The setting of the gradation value range for the reference data and the exclusion of the data belonging to the gradation value range may be performed when the reference data is registered, but the registered reference data is also used for purposes other than authenticity determination. When there is a possibility, it is desirable to perform the above processing after reading the registered reference data.

  In the above description, the maximum gradation value threshold that the cumulative frequency from the minimum value of the gradation value (lightness value) does not exceed the predetermined value is used as the gradation value threshold. However, the present invention is not limited to this. The minimum gradation value threshold value in which the cumulative frequency from the maximum value of the tone value (lightness value) exceeds a predetermined value may be applied, the average value of the gradation values is AVE, and the standard deviation of the distribution of gradation values Where σ is a predetermined value and n is a predetermined value (for example, 2), the gradation value threshold may be AVE-nσ, and the gradation value range excluded from the correlation value calculation target may be AVE-nσ or less. In the above description, an example in which the gradation value range from the minimum value of the gradation value to the gradation value threshold is set as the gradation value range estimated to include the noise component has been described. For example, when the background of the recording paper 22 on which the document is printed is black, the gradation value threshold value (for example, the maximum value of the gradation value (brightness value)) is changed from the maximum gradation value. Is the minimum (or maximum) gradation value threshold that does not exceed the predetermined value, or the average value of the gradation values is AVE, the standard deviation of the gradation value distribution is σ, and the predetermined value is The gradation value range up to AVE + nσ) when n is set may be set as a gradation value range estimated to contain a noise component.

  Moreover, although the example which made the reference | standard area | region and the collation area | region rectangular shape (specifically square shape) was demonstrated above, it is not limited to this, A rectangle, a trapezoid, a triangle, a circle | round | yen, an ellipse, linear shape (for example, Any shape such as a very flat rectangular shape having a width of 1 to several dots can be employed. Moreover, the said area | region does not need to exist continuously and may exist in discrete steps. However, even if the shape of the area is complicated, it does not contribute to improving the accuracy of the authenticity determination.Therefore, unless there are special circumstances such as many printed parts on the paper, the shapes of the reference area and the collation area are A simple rectangle or circle is desirable.

  In general, paper is rectangular in documents, securities, etc., and when the reference area and the collation area are rectangular, if one side of the area is parallel to one side of the paper, the paper and the area are all Since the sides of the reference area are parallel or perpendicular to each other, the reference area is set so that any one side of the reference area is parallel to any one side of the paper when registering the reference data. If the collation area is set so that any one side is parallel to any one side of the paper, rotation of the collation area with respect to the reference area can be avoided, so that the determination accuracy of the authenticity determination can be improved. .

  When the reference area and the collation area are circular, if the center of the collation area can be overlapped with the center of the reference area by any method, the same processing as that of the rectangular shape can be performed by converting to the polar coordinate format. However, in general, a scanner has a structure in which a line sensor and an original are relatively moved in a sub-scanning direction (a direction orthogonal to the sensor array of the line sensor) to perform two-dimensional reading, and the output order of data from the scanner is also a rectangular area. Since the order is suitable for taking in the data, there is little merit of making the reference area and the collation area both circular, and the process of superimposing the centers of both areas is not a simple process. It is appropriate that the reference area and the collation area are rectangular.

  In addition, if the features with randomness that are inherent to the object of authenticity determination are features with color changes, it is effective to separate and read the features of the solid into a plurality of color components. However, as a characteristic characteristic of paper as a solid, color information is not necessary when using a random change in the transparency of the paper due to the randomness of the entanglement of the fibrous material forming the paper. It is sufficient to read with a single color gray scale. Further, the resolution of gradation in reading may be 256 gradations (8 bits) if the object to be read is paper. Even if the gradation resolution in reading is further increased, the determination accuracy of authenticity determination is hardly improved.

  Commercially available inexpensive scanners can usually read with a gradation resolution of at least 8 bits, but if for some reason compression of the number of bits of the gradation value is required, for example, the gradation value between the shadow and the highlight is changed. Instead of allocating bits evenly (linearly) with respect to changes, more lightness ranges (lightness ranges close to highlights) where gradation value changes corresponding to random changes in paper transparency are distributed If bits are assigned, even if the number of bits of the gradation value is set to 6 bits or 4 bits, the number of bits of the gradation value is set to 8 bits, and the bits are assigned evenly with respect to the change of the gradation value. Equivalent true / false determination accuracy can be obtained. In addition, lossy compression such as JPEG may be applied to data obtained by reading.

  In the above description, the reference data is encoded and recorded on paper (original). However, the present invention is not limited to this, and the reference data is stored in a database connected to the PC 32 that performs authenticity determination via a network. Reference data may be registered. In this case, if the information for identifying the corresponding reference data registered in the database is recorded on the paper (original), the corresponding reference data is retrieved from the database via the network based on the information. It can be easily obtained.

  Further, the example in which the size of the collation area is made larger than the size of the reference area has been described above. However, the present invention is not limited to this, and the size of the reference area is made larger than the size of the collation area (in this case, The area corresponds to the first area according to the present invention), and the correlation value between the collation area and the partial area (corresponding to the second area according to the present invention) of the same size as the collation area in the reference area is calculated. Alternatively, the authenticity determination may be performed repeatedly while shifting the position of the partial area in the reference area. However, in this case, there is a disadvantage that the reference data increases in capacity, and it is preferable to make the size of the collation area larger than the size of the reference area because the storage capacity for storing the reference data can be saved.

  In the above description, paper is used as an example of the solid according to the present invention, but the inventors of the present application also represented a non-printing area on the back surface of a prepaid card having a silver back surface, represented by a highway card or a passnet card. Each card's unique features (random irregularities that cannot be controlled during manufacture) are distributed, and it has been confirmed by experiments that this feature can be read and used for authenticity determination. It is also possible to apply these cards. For example, the authenticity determination of a highway card is specifically performed by, for example, reading random irregularities distributed in a partial area (reference area) of the non-printing area on the back side as a card-specific feature at the time of card manufacture, An electronic signature is attached to the reference data obtained by reading and recorded as magnetic data on the card. When reading and writing the magnetic data to the card at the toll collection point, the reference data on the back side of the card is used. Optically reading a collation area that includes the range and wider than the reference range, calculates the correlation value between the collation data obtained by reading and the reference data included in the magnetic data read from the card, and the maximum correlation value and The authenticity of the card can be determined by comparing each normalized score of the maximum correlation value with a threshold value.

  Next, an experiment conducted by the inventors of the present application in order to confirm the determination accuracy of the authenticity determination when the present invention is applied will be described. In this experiment, continuous 10 sheets were extracted from A4 500-sheet A4 office paper (Fuji Xerox Office Supply Co., Ltd. C2 paper product code V436) and used as a sample.

[Experiment for FRR confirmation]
In this experiment, 40 reading areas were set on the entire surface of the A4 sample at substantially equal intervals (see FIG. 14). The center coordinates of each reading area when the upper left corner toward the reading surface of each sample is the origin in the state where the longitudinal direction of the sample is the vertical direction are as follows in terms of the number of dots of 400 dpi.
(500,500), (500,1000), (500,1500), (500,2000), (500,2500), (500,3000), (500,3500), (500,4000), (1000,500 ), (1000, 1000), (1000, 1500), (1000, 2000), (1000, 2500), (1000, 3000), (1000, 3500), (1000, 4000), (1500, 500), (1500,1000), (1500,1500), (1500,2000), (1500,2500), (1500,3000), (1500,3500), (1500,4000), (2000,500), (2000 , 1000), (2000, 1500), (2000, 2000), (2000, 2500), (2000, 3000), (2000, 3500), (2000, 4000), (2500, 500), (2500, 1000 ), (2500, 1500), (2500, 2000), (2500, 2500), (2500, 3000), (2500, 3500), (2500, 4000)
Each sample was read using a FUJITSU fi-4010CU (flatbed scanner) with a resolution of 400 dpi and a gradation of 8 bit gray scale.

  The reading area size is 16 × 16 dots (about 1 mm × about 1 mm), 32 × 32 dots (about 2 mm × about 2 mm), 64 × 64 dots (about 4 mm × about 4 mm), 128 × 128 dots (about 8 mm × Four types (about 8 mm) were set. In this experiment, each reading area is used as both a reference area and a collation area, the entire surface of the sample is read by a scanner in order to reduce the number of readings, and data corresponding to each reading area (reference reference) is obtained from the image data obtained by reading. Region data and data used as collation region data) were cut out and used for authenticity determination. The combinations shown in the following Table 2 were used as the combinations of the sizes of the reference area and the collation area so that the size of the collation area with respect to the reference area was doubled or quadrupled as the ratio of the side lengths.

  Also, taking advantage of the fact that the scanner's platen is slightly larger than the A4 size, samples can be placed on the platen using the upper right butting (normal placement) and the lower left butting (upper right The sample position is about 2 mm in the longitudinal direction and about 10 mm in the short direction with respect to the contact), clockwise rotation right alignment (clockwise rotation about 1 degree), counterclockwise rotation left alignment (counterclockwise rotation about 1 degree) 4 were determined, and each sample was read with each sample placed on the document table.

  In this experiment, as the combination of the reference area data and the collation area data used for the authenticity determination, the authenticity determination was performed by combining data obtained by reading the samples in different states. Since there are three different placement methods that can be combined with a certain placement method, a single sample can be used in a single combination of the reference region and collation region size combinations shown in Table 2. 4 × 3 = 12 true / false judgments are performed for a single reading area, and there are 40 reading areas in a single sample, and the number of samples is 10, so that the reference area In addition, 12 × 40 × 10 = 4800 times of authenticity determination was performed for each combination of the size of the collation area.

  As mentioned above, the data of the reference area and verification area used for authenticity determination is combined with the data obtained by reading the samples in different ways, so it corresponds to the reading area from the image data. When the data to be extracted is cut out, the position of the reading area is corrected so that the center position of the reference area and the center position of the collation area substantially coincide.

  That is, when the image data for cutting out the data in the reading area is data obtained by reading the sample placed on the “upper right abutment”, the position correction of the reading area is not particularly performed. In the case of “bottom left abutment”, based on the image data obtained by reading with the scanner, the amount of positional deviation at the end of the sample was calculated to correct the position of the reading area. As for “clockwise right justification” and “counterclockwise left justification”, the position of the corner of the sample is detected based on the image data, and the actual position after rotating and moving the sample based on the detected position of the corner. The position of the reading area was calculated, and the position of the reading area cut out as data from the image data was corrected (only the center position was corrected, and the rotational distortion was not corrected).

  As in the above-described embodiment, the true / false determination is performed by calculating the correlation value between the reference area and the reference area in the reference area and the reference area using the normalized correlation method. (M−n + 1) × (m−n + 1) correlation values are obtained by repeating while moving one dot at a time in the region (however, the reference region is m × m dots and the matching region is n × n dots) Obtains the maximum correlation value and the normalized score of the maximum correlation value, and determines whether the maximum correlation value is 0.3 or higher and the normalized score of the maximum correlation value is 5.0 or higher It was done by doing. The results of the experiment are shown in Table 3 below.

  As can be seen from Table 3, under the conditions of 400 dpi resolution and 8-bit gray scale reading, if the reference area size is 32 × 32 dots and the matching area size is 64 × 64 dots, then it is practical It can be understood that the FRR is low enough to cause no problem. It is also clear that the above reading conditions are sufficiently realizable with a commercially available inexpensive scanner, and it is not necessary to use an expensive reading device such as a microscope for reading.

  In addition, the inventors of the present application analyzed the case where an erroneous determination (an erroneous determination that a genuine article is a fake) occurred in the above-described experiment, and in particular, the sample was rotated clockwise or counterclockwise. It has become clear that misjudgment tends to occur when Therefore, for example, the rotational distortion is detected and corrected, care is taken so that the paper to be read does not rotate when the paper to be read is placed on the original plate of the scanner, and the original plate of the scanner is made difficult to rotate. It seems that the improvement of FRR can be easily achieved if measures such as prevention or reduction of rotational distortion are taken.

[Experiment for FAR confirmation]
Similar to the FRR experiment, the data corresponding to the reference area and the data corresponding to the collation area were cut out from the image data obtained by reading the entire surface of the A4 sample at a resolution of 400 dpi and an 8-bit gray scale gradation. . Since the FAR is a probability that a fake is erroneously determined as a real thing, in the experiment for confirming the FAR, all areas on the sample can be used as collation areas. In this experiment, if the correlation value with the reference area is calculated for all areas except the reference area on the entire surface of A4, it is the same if it is determined as a fake from the normalized score of the maximum correlation value and the maximum correlation value. Since it is self-evident that any collation area on this sample is also determined to be fake, the collation area is a 3307 × 4676 dot area obtained by reading the entire scanning area of the scanner including the entire surface of the A4 sample at 400 dpi. .

  In this experiment, the number of samples was five, and four reading areas were set on the entire surface at approximately equal intervals. The center coordinates of each reading area are (500, 500), (500, 3500), (2500, 500), (2500, 3500) in terms of the number of dots of 400 dpi. In addition, the reference area has four sizes of 16 × 16 dots, 32 × 32 dots, 64 × 64 dots, and 128 × 128 dots.

  For four reference areas per sample, the authenticity determination is performed over the entire surface of the other four samples. Therefore, four locations x 4 per sample = 16 times authenticity determination. become. Since this is carried out for five samples, the total number of true / false determinations is 5 × 16 = 80. Although it seems that the number of times for the experiment for FRR confirmation is small, as described above, since the entire reading area of the scanner including the entire surface of the A4 sample is set as the collation area, it is only seen so. If it is divided into areas, it is equivalent to having made the authenticity judgment 10 million times or more. The experimental results are shown in Table 4 below.

  As can be seen from Table 4, when the size of the reference area is other than 16 × 16 dots, FAR = 0.000%. Therefore, the verification area is divided into small areas of arbitrary size, and authenticity determination is performed. Even if done, the FAR is guaranteed to be 0.0000%. On the other hand, when the size of the reference area is 16 × 16 dots, FAR = 31.250%, which is an unsuitable value. This is the worst value, and there is a possibility that FAR can be improved if the collation area is divided into small areas. However, even in the experiment for FRR confirmation described above, when the size of the reference area is 16 × 16 dots, The determination accuracy of authenticity determination is lower than that in the case where the size of the region is made larger. Therefore, it has been clarified that the reference area size should be 32 × 32 dots as the lower limit at a resolution of 400 dpi.

  Subsequently, when a gradation value range that is estimated to contain noise components is set with respect to the reference data, and processing that excludes data belonging to the set gradation value range from the correlation value calculation target is performed. The results of an experiment conducted by the present inventors in order to confirm the effect will be described.

  In this experiment, the same A4 paper as in Example 1 is passed through the printer process of a color multifunction device (Ducu Center Color 400 manufactured by Fuji Xerox Co., Ltd.), so that the black spot noise corresponding to the toner scattering is completely removed. Samples attached to each location were created. The entire surface of this sample was read with a scanner of the same machine at a resolution of 600 dpi, and the image data obtained by reading was read into a PC and image processing was performed to obtain an 8-bit grayscale image of 4967 × 7020 dots. For this image, 40 reference points (5 horizontal positions × 8 vertical positions) are set every 750 dots in both vertical and horizontal directions, and data of 128 × 128 dot square reference areas (reference data) centered on this reference point. ) Was extracted by image processing. In the range corresponding to each reference area on the sample, as shown in FIG. 15, for example, black spot noise exists, and the reference data extracted from the 40 reference areas is accompanied by black spots. Noise components corresponding to noise are mixed.

  Next, using the position of the sample on the scanner platen when reading for the above reference data acquisition as a reference, the same sample on the scanner platen is (1) several mm in length and width To acquire reference data while performing various movements and rotations of (2) rotate clockwise about 1 degree, (3) rotate counterclockwise about 1 degree, and (4) return to the reference position Under the same reading conditions as when reading the image, the entire surface is read repeatedly (four times) by the scanner, and the image data obtained by the reading is taken into the PC and image processing is performed, whereby an 8-bit gray scale of 4967 × 7020 dots is obtained. An image was obtained, and data of a matching area of 256 × 256 dots and 512 × 512 dots centered on the above-described 40 reference points was extracted by image processing. As described above, since black spot noise exists in each of the 40 reference areas, black spot noise also exists in each matching area including any reference area as shown in FIG. Accordingly, noise data corresponding to black spot noise is mixed in the collation data extracted from each collation area.

Subsequently, in order to confirm FRR, correlation values between reference data extracted from a single reference region and data of four types of matching regions including the reference region and having different positions or angles are obtained as follows: No data to be excluded from (2) Exclude from the calculation data the gradation value range from the minimum gradation value to the cumulative frequency of 3% of the reference data, (3) Minimum gradation value of the reference data The data of the gradation value range from the value to the cumulative frequency of 5% is excluded from the calculation target, respectively, and the maximum value of the correlation value and the normalized score of the maximum value of the correlation value are calculated respectively. This is performed for each of the reference data of the 40 reference areas. As a result, four calculation results can be obtained per single reference area for a single calculation method, and 160 calculation results can be obtained from 40 reference areas. The combination of the size of the reference area and the size of the collation area for which the above calculation is performed is as follows.
Reference area size Collation area size 128 × 128 dots 256 × 256 dots 128 × 128 dots 512 × 512 dots For reference to FAR, reference data extracted from a single reference area differs from the reference area 39 Correlation values with the data of the collation area including the reference area (39 places × 4 types = 156 collation areas) are calculated by the above-described three types of calculation methods, respectively, and the maximum correlation value and the correlation value are calculated. The calculation of the maximum normalized score is performed for each of the reference data of the 40 reference areas. As a result, with respect to a single calculation method, 156 calculation results can be obtained per single reference area, and therefore 6240 calculation results can be obtained from 40 reference areas. The combination of the size of the reference area and the size of the collation area where the above calculation is performed is the same as described above.

  Then, the threshold value of the maximum value of the correlation value is changed in 101 increments in increments of 0.01 from 0.0 to 1.0, and the threshold of the normalized score of the maximum correlation value is changed in 101 increments of 0.1 increments from 0.0 to 10.0. For each of the 10201 combinations of the maximum correlation value and the normalized score of the maximum correlation value obtained, data for FRR confirmation (calculation result of maximum correlation value and normalized score) ) Is calculated for each calculation method, and FAR when applied to FAR confirmation data (maximum correlation value and calculation result of normalized score) is calculated for each calculation method. Asked.

The effect of performing the process of excluding the gradation value range data estimated to contain noise components from the reference data from the correlation value calculation target is FRR = 0.00% and FAR = 0.00%. This can be grasped from the change in the number of combinations of threshold values. The experimental results are shown below.
[No data excluded from calculation]
Size of reference area Size of collation area Number of threshold combinations 128 × 128 dots 256 × 256 dots 1008 ways 128 × 128 dots 512 × 512 dots 1316 ways [tone values from the minimum value of gradation values to 3% cumulative frequency Exclude range data from calculation target)
Size of reference area Size of collation area Number of threshold combinations 128 × 128 dots 256 × 256 dots 2093 ways 128 × 128 dots 512 × 512 dots 2800 ways [tone values from the minimum value of the tone value to 5% cumulative frequency Exclude range data from calculation target)
Reference area size Collation area size Number of threshold combinations 128 × 128 dots 256 × 256 dots 1904 ways 128 × 128 dots 512 × 512 dots 2581 ways In addition, the experimental results are shown in FIG. When the maximum correlation value is taken on the horizontal axis (0.00 at the left end and 1.00 at the right end), the normalized score of the maximum correlation value is taken on the vertical axis (upper end is 0.0, lower end is 10.0). Shows changes in FRR and FAR values with respect to changes in the maximum correlation value and the threshold value of the normalized score of the maximum correlation value. As is apparent from FIG. 17, when data of a gradation value range estimated to contain a noise component is excluded from the correlation value calculation target, a threshold region where FRR = 0.00% and FAR = 0.00%. The area of is increasing. Further, although FAR is more important than FRR in authenticity determination, it can also be confirmed that the region where FAR = 0.00% and FRR> 0.00% is expanded in the horizontal axis direction. Therefore, by excluding the data of the gradation value range that is estimated to contain noise components from the target of correlation value calculation, the accuracy of threshold setting required for accurately performing authenticity determination is eased. As a result, it can be understood that the accuracy of the authenticity determination is improved.

  Note that if the data of the gradation value range that is estimated to contain noise components is excluded from the correlation value calculation target, the number of data used for the calculation decreases, so that the ink (or toner) in the reference area or the collation area ) And shading due to the adhesion of dirt, the accuracy of the maximum correlation value and the normalized score of the maximum correlation value, and hence the accuracy of true / false judgments, are reduced. There are concerns that lead to For this reason, the inventors of the present application conducted the above-described experiment on a sample to which no black spot noise was added. The result of this experiment is shown in FIG. As is clear from FIG. 18, there are no shaded parts due to the adhesion of ink (or toner) or dirt in the reference area or the collation area, and noise components are mixed in the reference data or collation data. If it is not, the accuracy of the true / false determination will not deteriorate even if the data of the gradation value range estimated to contain the noise component is excluded from the correlation value calculation target.

  Therefore, the gradation value that is estimated to contain noise components regardless of whether or not there is a shaded part due to the adhesion of ink (or toner) or dirt in the reference area or the collation area Even if the data in the range is uniformly excluded from the correlation value calculation target, it does not adversely affect the accuracy of the true / false judgment when there is no shading in the reference area or the collation area. It can be understood that it is possible to improve the accuracy of the true / false determination in the case where a light and shade part exists in the verification region.

(A) is a registered image and (B) is an image figure which shows an example of a collation image for demonstrating the experiment which this inventor etc. implemented. It is an image figure for demonstrating the calculation of the correlation value of the registration image and collation image in the said experiment. (A)-(C) are the diagrams which show the distribution of the correlation value on various conditions with the maximum value of a correlation value, and the normalized score. It is a diagram which shows distribution of the correlation value in a 1st comparative example among the experiments which this inventor etc. implemented. It is a diagram which shows distribution of the correlation value in the 2nd comparative example. It is a diagram which shows distribution of the correlation value in a 3rd comparative example. It is a diagram which shows distribution of the correlation value in the experiment assumed as a case where the probability that a fake is erroneously determined to be true is high. 1 is a schematic configuration diagram of a color printer according to an embodiment. 1 is an external view of a PC and a scanner that function as a true / false determination device. FIG. It is a flowchart which shows the content of the reference | standard data registration process performed with a color printer. It is the image figure which visualized an example of reference data. It is a flowchart which shows the content of the authenticity determination process performed with PC (authenticity determination apparatus). It is a flowchart which shows the content of the other example of authenticity determination processing. It is an image figure which shows the position of the reading area | region in the experiment for FRR confirmation. It is an image figure which shows an example of the reference | standard area | region where black spot noise exists. It is an image figure which shows an example of the collation area | region where black spot noise exists. It is an image figure which shows the threshold value of the normalized score of the maximum value of a correlation value and the maximum value of a correlation value, and FAR and FRR in the experiment using the reference | standard area | region and collation area | region with a black spot noise. It is an image figure which shows the threshold value of the normalized score of the maximum value of a correlation value and the maximum value of a correlation value, and FAR and FRR in the experiment using the reference | standard area | region and collation area | region with a black spot noise.

Explanation of symbols

10 Color Printer 28 Reading Unit 30 Printer Controller 32 PC
34 Scanner

Claims (9)

A true / false determination method for determining the authenticity of a solid in which readable and unique features having randomness are distributed along a surface,
Obtaining reference data representing features distributed on the true solid, obtained by pre-reading the features of the true solid,
By reading the characteristics of the determination target solid, the matching data representing the characteristics distributed on the determination target solid is obtained,
Based on the reference data and the collation data, data representing characteristics distributed in a first region of a predetermined size on one of the true solid and the solid to be determined, and the first region on the other solid And calculating the correlation value with the data representing the characteristics distributed in the second area of the same size, while moving the position of the second area on the other solid within the area larger than the predetermined size. repetition,
The maximum value of the plurality of correlation values obtained by the calculation is equal to or greater than the first predetermined value, and the value obtained by subtracting the average value of the correlation values from the maximum value of the correlation values is divided by the standard deviation of the correlation values. A true / false determination method, wherein the authenticity of an object to be determined is determined based on whether or not a normalized score of a maximum correlation value obtained is equal to or greater than a second predetermined value .   The feature of the solid is optically readable, and the reference data and the collation data are images obtained by irradiating light to the true solid or the solid to be judged and reading reflected light or transmitted light. The authenticity determination method according to claim 1, which is data.   3. The authenticity determination method according to claim 2, wherein the solid is a sheet-like or flat medium, and the characteristics of the solid are read with a flatbed scanner.   2. The authenticity determination method according to claim 1, wherein the correlation value is calculated by a normalized correlation method. For at least one of the reference data and the collation data, a gradation value range estimated to contain a noise component based on a gradation value distribution is set, and data belonging to the set gradation value range is 3. The authenticity determination method according to claim 2, wherein the correlation value is calculated after the exclusion. The reference data obtained by optically reading the characteristics of the true solid and recorded on the predetermined medium is read from the predetermined medium to acquire the reference data, and the gradation value of the acquired reference data A gradation value range estimated to contain a noise component with respect to the reference data based on the distribution of the reference data, and after excluding data belonging to the set gradation value range from the reference data, the correlation 6. The authenticity determination method according to claim 5, wherein a value is calculated. As the gradation value range estimated to contain the noise component, the range from the maximum value or the minimum value of the gradation value until the cumulative frequency reaches a predetermined value, or the average value of the gradation values is AVE, 7. The range of gradation values not less than AVE + nσ or not more than AVE−nσ, where σ is a standard deviation of gradation value distribution and n is a predetermined value. Authenticity determination method. A true / false determination device for determining the authenticity of a solid in which a unique readable characteristic having randomness is distributed along a surface,
Obtaining means for obtaining reference data representing features distributed on the true solid, obtained by reading the features of the true solid in advance;
Reading means for obtaining collation data representing features distributed on the determination target solid by reading the characteristics of the determination target solid;
Based on the reference data and the collation data, data representing characteristics distributed in a first region of a predetermined size on one of the true solid and the solid to be determined, and the first region on the other solid And calculating the correlation value with the data representing the characteristics distributed in the second area of the same size, while moving the position of the second area on the other solid within the area larger than the predetermined size. Repetitive computing means;
The maximum value of the plurality of calculated correlation values is equal to or greater than a first predetermined value, and the value obtained by subtracting the average value of the correlation values from the maximum value of the correlation values is obtained by dividing by the standard deviation of the correlation values. Determination means for determining the authenticity of the object to be determined based on whether the normalized score of the maximum correlation value is equal to or greater than a second predetermined value;
A true / false determining apparatus comprising: A computer connected to a reader capable of reading the solid-specific features distributed along the surface of the solid and having randomness;
Obtaining means for obtaining reference data representing features distributed on the true solid, obtained by pre-reading features of the true solid;
Read control means for obtaining collation data representing features distributed on the determination target solid by causing the reader to read the characteristics of the determination target solid;
Based on the reference data and the collation data, data representing characteristics distributed in a first region of a predetermined size on one of the true solid and the solid to be determined, and the first region on the other solid And calculating the correlation value with the data representing the characteristics distributed in the second area of the same size, while moving the position of the second area on the other solid within the area larger than the predetermined size. Repetitive computing means,
And the maximum value of the plurality of calculated correlation values is equal to or greater than the first predetermined value, and the value obtained by subtracting the average value of the correlation values from the maximum value of the correlation values is divided by the standard deviation of the correlation values. Determining means for determining the authenticity of the object to be determined based on whether or not the normalized score of the maximum correlation value obtained is greater than or equal to a second predetermined value
Program to function as.

Images