JP2000146952A - Method and device for discriminating paper - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable the determination of which paper among a database the paper has a high degree of similarity with by subjecting the image of the paper to frequency analysis and identifying the characteristics of a predetermined wavelength range, for example, a wavelength range including periodic components derived from paper manufacturing processes. SOLUTION: In a image pickup part 2, image data is obtained form light transmitted through or reflected by at least part of a paper sample and is subjected to frequency analysis at a frequency analyzing part 18. Integration data on each of a plurality of wavelength ranges from the data after the frequency analysis is obtained. The correlation of the integration data is obtained to the corresponding integration data of the same wavelength range of a contrast standard paper sample to determine whether the sample paper is identical to the standard paper sample.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and an apparatus for identifying the history of paper, which aids criminal investigation by identifying when a sample paper is from which paper machine and from which paper machine. It is. For example, the present invention can be applied to paper used for creating a counterfeit note, paper left at a crime scene, and the like.

[0002]

2. Description of the Related Art As a method for discriminating paper differences, a chemical method has been mainly proposed so far. Examples include pulp fiber comparison methods, filler analysis, and coating agent analysis.

[0003]

However, since these methods are identification methods for destroying a sample, development of a non-destructive method is desired.

[0004]

According to a paper identification method of the present invention, a frequency analysis is performed on image data obtained by light transmitted or reflected by at least a part of a sample paper sample, and the frequency analysis after the frequency analysis is performed. The integrated data for each of the plurality of wavelength ranges is obtained from the data, and the correlation of the integrated data is obtained for the integrated data corresponding to the same wavelength range of the reference paper sample, and the sample paper sample is identical to the standard paper sample. It is determined whether or not there is.

Further, there is a plurality of data of the standard paper samples, and the correlation is obtained for each standard paper sample, and it may be determined that the standard paper sample having the highest correlation is the same paper sample as the sample paper sample. . Further, the plurality of wavelength ranges include a papermaking wire in a papermaking process, a wavelength at least including a periodic component derived from at least one of a fabric joint used for suction dehydration and a fabric joint used for drying. More preferably, it is within the range.

Further, it is more preferable that the periodic component derived from the papermaking process is a periodic component derived from a papermaking wire. Specifically, the frequency component can be set to 0.6 mm or less. The paper identification device according to the present invention is an image data acquisition unit that obtains image data by light transmitted or reflected by at least a part of a sample paper sample, a frequency analysis unit that performs frequency analysis on the obtained image data, A calculation unit for obtaining each integrated data for a plurality of wavelength ranges from the data from the analysis unit, and obtaining a correlation of the integrated data for the corresponding integrated data of the same wavelength range of the reference paper sample; and A determination unit for determining whether the sample paper sample is the same as the standard paper sample.

[0007] In the paper identification method according to the present invention, a frequency analysis is performed on image data obtained by light transmitted or reflected through at least a part of the paper to extract a periodic component derived from the papermaking process, The paper is identified using the pattern. The periodic component derived from the papermaking process may be considered to be a periodic component derived from at least one of a papermaking wire, a roll joint used for suction dehydration, and a fabric joint used for drying.

The case where the periodic component derived from the papermaking process is the periodic component derived from the papermaking wire is most easily discriminated. The paper identification device according to the present invention includes an image data acquisition unit that obtains image data by light transmitted or reflected through at least a part of paper, a frequency analysis unit that performs frequency analysis on the obtained image data, It may have a collating and judging section for differentiating and discriminating by focusing on the periodic component based on the output, and a database section for reference paper.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention utilizes image processing in comparison with the above-mentioned prior art, and the periodicity of a pattern of a paper formation and a paper-making wire formed in a paper-making process, which the prior art did not pay attention to. The identification is performed by paying attention to. This will not destroy the sample. In commercial paper such as newsprint or printing paper, it is almost inevitable that a repeated pattern is formed on the paper during the papermaking process due to a papermaking net (hereinafter referred to as a papermaking wire), a fabric in a suction dehydration and drying step, and the like. Then, when a transmission or reflection image of the paper is obtained and a frequency analysis such as Fourier transform is performed, a repetition pattern thereof is clearly extracted.

One of the particularly strong cycles derived from the papermaking process remaining on the paper extracted by the frequency analysis is derived from the papermaking wire, and the cycle is 0.6 mm or less. In order to identify differences between papers, a capturing device and an identification algorithm are required.
The capture device has the following configuration. First, the paper is fed into the apparatus by using a roller or the like to feed the paper to the next imaging site.

Next, a portion where the inserted paper is irradiated with light to obtain a transmitted image. This is not limited to a transmission image, but may be a reflection image. Next, the part that captures the transmitted or reflected image. An imaging tube, a CCD camera, a flatbed scanner, or the like can be considered, but the method is not limited. The computer may need an image board to take the signal into the computer.

Although the signal taken into the computer is stored in the memory, the signal may be stored in an external storage device such as a hard disk, a floppy (registered trademark) disk, a magneto-optical disk, and a CD-ROM. Subsequently, as an identification method, first, a frequency analysis is performed on the image signal in the memory. As a method of the frequency analysis, a Fourier transform, a wavelet transform, or the like can be considered, but the method is not limited. The same applies to whether these conversions are performed one-dimensionally for linear signals or two-dimensionally for planar signals.

An example of a specific matching method is performed by comparing the values of the power spectrum of Fourier transform, but the present invention is not limited to this method. Subsequently, the paper in the database to be compared is analyzed in the same manner, and both are collated with the previous sample. The data to be accumulated in the database may be not only a transmission or reflection image of paper but also a result of frequency analysis performed on the sample under the same method and conditions as those of the sample. One example is a power spectrum of Fourier transform.

A transmission or reflection image of these commercial papers is stored in a database, and it is checked which one of the papers of the sample matches. A mathematical method such as a cross-correlation method or a shortest distance method can be used for the matching algorithm, but the present invention does not limit the method.
In general, the periodic component remaining on the paper itself as described above has an extremely strong periodicity as compared with a printed portion or wrinkles, and thus is less affected by printing or wrinkles. However, the printed portion may have a halftone dot or a ripple-like pattern, which may affect the frequency analysis for extracting the repetition period derived from the manufacturing process included in the paper. Therefore, in order to improve the accuracy of the evaluation, it is more desirable to analyze the frequency of the portion without printing or wrinkles.

[0015]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a block diagram schematically showing the configuration of an embodiment of a paper identifying apparatus according to the present invention. This example is
A transmission image of paper as a sample is captured, two-dimensional discrete Fourier transform is performed on the transmission image, and the same operation is performed on reference paper data in a database. Indicating similarity is sequentially performed for each of a group of reference papers in a database, and it is determined that the one having the highest similarity among the groups of reference papers has the same history as the sample. Things.

The imaging section 2 is for inputting a transmission image, and includes a transmission platen (not shown) on which a paper sample sent by the paper take-in driving device 4 is placed, and a paper sample placed on the transmission platen. And a CCD camera 8 for taking a transmission image of the illuminated paper sample. The input of the image of the paper sample is not limited to transmission, but may be a reflection image or the like.

As the image pickup device, in addition to the CCD camera 8, an image pickup tube, a drum scanner, an image scanner, or the like can be used. Here, as the imaging unit 2, a transparent image of paper is converted into a color image scanner (GT-9).
000 WIN and transparent document unit GTA4FLU:
Both were prepared using Seiko Epson Corporation).
The capture conditions are black and white photography mode, resolution: 400 dpi, 475 pixels vertically (about 30.16 mm) x 640 pixels horizontally (about 4 pixels).
0.64 mm). This image data was used for analysis with a luminance resolution of 8 bits (256 gradations of black and white).

The image capturing section 12 captures image data representing a grayscale image from the image capturing section 2, and can be realized by an image board or the like. The memory 14 stores the image data captured by the image capturing unit 12 and provides the data for calculation. This image signal can also be recorded and stored in a large-capacity external storage device 16 such as a floppy disk, hard disk, magneto-optical disk, or CD-ROM. The database unit 17 for storing reference paper data, which is a standard paper sample, is normally provided in the large-capacity external storage device 16, but may be stored in the memory 14.

The database unit 17 accumulates reference data created by the same method as that of the sample. One of them may integrate the transmitted image itself captured under the same conditions as the sample, for example, the image capturing device, image capturing conditions such as exposure and aperture, magnification, image resolution, luminance resolution, and the like. In another example, the image created in this manner may be a power spectrum created under the same conditions as the sample, for example, an analysis algorithm, an image size to be analyzed, and the like.

The frequency analysis section 18 obtains frequency analysis data from the image data stored in the memory 14. The frequency analysis unit 18 includes an image processing analysis device DA
-5000S and its optional software (manufactured by Oji Scientific Instruments), a two-dimensional fast Fourier transform
(FFT) 256 pixels each in the vertical and horizontal directions (40
A range of about 16.26 mm square at 0 dpi) was subjected to frequency analysis processing.

The memory 14, the frequency analysis unit 18 and the operation unit 20 constitute an image data processing unit. As a result of the frequency analysis processing, the power spectrum is output as a numerical value table and its diagram (power diagram). It is processed by the arithmetic unit 20 and a power diagram to be handled will be described.

FIG. 2 shows a paper sample 2 obtained by FFT processing.
An actual example of the power diagram of the dimensional fast Fourier transform is shown.
Hereinafter, it is abbreviated to a power diagram. FIG. 3 is an explanatory diagram for explaining information of a power diagram. The power diagram shows, as its properties, the wavelength of the periodic component, its repetitive direction, and the intensity of the period. The intensity is expressed by the intensity of the whiteness of the point. First, as for the intensity, the larger the power, the stronger the cycle. Next, as for the wavelength, since the distance from the center of the power diagram to that point is proportional to the reciprocal of the wavelength, a point closer to the center of the diagram indicates the existence of a long period, and indicates a shorter period as it goes outside. As for the direction,
The azimuth of the point as viewed from the center of the figure is the axis of the direction in which the cycle repeats. The analysis here focuses primarily on wavelength.

A paper such as a general printing paper always contains a periodic component derived from the manufacturing process. The cycle is mainly in the manufacturing process, paper wire (net), felt,
It is derived from a canvas (blanket). Among them, the strongest cycle is 0.5m
m. Therefore, in the present invention, the sample data as the sample paper sample and the reference data as the standard paper sample in the database are collated by collating the power spectrum. Here, the transparent image of paper is 4
Since 256 pixels in each of the vertical and horizontal directions are captured with the resolution of 00 dpi, the peak at a distance of 25 pixels from the center of the power diagram means a period of 0.65 mm. Therefore, the peak group located outside the concentric circle including the peak group means the existence of a period finer than 0.65 mm.

The following is a more general description. When the power spectrum of the FFT of 256 × 256 pixels in the vertical and horizontal directions is tabulated in the same manner as in the figure, the value of the DC component is located at the center, and the surrounding pixels are successively arranged from the long wavelength component to the short wavelength component. 25.4 / 400 × 25
The power at a wavelength of 6 / n (mm) is shown. The higher the power value, the stronger the wavelength component is.

In rare cases, printing, creases, stains, etc. on paper may have an effect, but they are often long wavelengths near the center of the power diagram, that is, at a point inside 25 pixels from the center of the power diagram. Become. Therefore, if the long wavelength region is omitted using a band-pass filter and subjected to the matching calculation, the matching accuracy can be further improved.

In practical use, the direction of the papermaking axis of a paper sample is generally unknown. Needless to say, even if the papers are the same, the two-dimensional power spectra almost never coincide if the directions of the papermaking axes are different. FIG.
In the power diagram as described above, even if the data of a certain sample and the data obtained by rotating the sample are simply compared, they do not match.

When the direction of the papermaking axis is not known, it is necessary to remove the azimuth from the collation conditions in order to collate the sample with the reference in the database. As a method therefor, there is a method in which integration is performed for each of the same wavelengths so that comparison can be performed only with the wavelength. Peaks having the same wavelength can be determined by coming on concentric circles in FIG. In other words, if the peaks on the concentric circles in FIG. 3 are integrated, it means that only the wavelength is focused and the repeated azimuth is removed. This method is called a ring mask method. Therefore, when the distance (unit: pixel) from the center of the power diagram is sequentially calculated, Fourier transform is performed for each of 256 pixels in the vertical and horizontal directions, so that a sequence of 127 data (vector) is obtained.

A cross-correlation method or the like can be considered as a collation method which is another task of the arithmetic unit 20, but the present invention is not limited to this method. If the papermaking axis is known, the power diagram can be collated as it is. If the axis of the paper making, which is almost the case in practice, is not known, a one-dimensional table is obtained by focusing only on the wavelength, but the collation method is the same. A given sample is sequentially collated against each group of reference papers in the database, and the one with the highest similarity among them is generally judged to be the same as the sample. can do.

The output unit 22 is an output unit such as a printer or a display that outputs the determined power value and the determination result of the similarity between the paper as the sample and the reference standard paper sample group in the database. 1 is a computer and a device controlled by the computer, which are collectively constituted by a personal computer or the like. A portion for creating a paper image is represented as an imaging unit 2.

Next, the calculation flowchart will be described. FIG. 4 shows a flowchart of the difference identification performed by the calculation unit. FIG. 4 is a general flow chart for determining which standard paper sample is the same as the paper sample. The sample paper sample 32 is a paper sample whose history is to be identified. In the next step 34, transmission image data is obtained. Then step 36
Then, at least a part of the transmission image data is subjected to a two-dimensional Fourier transform. In step 38, the result is captured as a power spectrum in the form of a diagram or a table.

A similar process is required for reference standard paper sample data in the database. However, the database may store either a paper transmission image or a power spectrum as a result of frequency analysis, if necessary. That is, the sample data and the reference standard paper sample data (in this example, _n of r _{1 to} rn
Any of the forms may be used as long as they are unified in each case. Hereinafter, the standard paper sample data for reference is omitted and referred to as reference.

Subsequently, at step 40, the reference and the sample are collated. The part is shown in FIG. FIG. 5 is a flowchart illustrating in detail the collation step 40 in FIG. The algorithm of the matching step 40 will be described. A loop 52 is formed for each of the sample and each of the reference r _{1 to} r _n , and the similarity between the two is sequentially calculated in step 54, and the reference data having the highest similarity is obtained in step 56 and subsequent steps. Judge as the same paper as the sample.

The details of the algorithm of step 54 for calculating the similarity will be described. Here, a cross-correlation method is given as one example. FIG. 6 is a flowchart for explaining the algorithm of the cross-correlation method for performing the similarity calculation in FIG. The algorithm is as shown in FIG. 6 for each of the sample paper sample 72 and the reference paper sample reference 74.
There is power spectrum data of two-dimensional fast Fourier transform of 256 pixels in each of vertical and horizontal directions. This power spectrum is integrated with the peak at the same distance from the center of the power spectrum diagram by the ring mask method, and the number of data of only the wavelength becomes 12
Replace with 7 sequences. Processing for improving the accuracy of calculation, such as removing the DC component and the long-wavelength component, respectively, is then performed in steps 76 and 78, respectively. For example, in the case of FFT at 400 dpi and 256 pixels, the peak at the position of 16 pixels from the center of the power diagram is 25.4 / 400 × 256/16 = 1.016
Since a wavelength of 1 mm is shown, if a value from 1 to 16 pixels is removed and the calculation is performed, a long wavelength of 1.016 mm or more is removed from the comparison target. At this time, only the short wavelength of about 1 mm or less was used as a comparison target here, but for example, 0.2 mm
Wavelength bands that are distant from each other, such as 0.4 mm and 0.5 mm to 0.6 mm, may be selected.

Next, in steps 80 and 82, the respective integrated data are proportionally calculated and normalized so that the sum of the squares becomes 1. Next, as a step 84 for performing the cross-correlation method, the values of the same coordinates of the two are multiplied, and the integration is performed. This calculation is performed in the same manner as the inner product of vectors. The similarity E is obtained in step 86 as the value. The maximum of the similarity E is 1 and the minimum is 0. When the similarity is 1, s and r are the same.

Examples of calculation are shown in Tables 1 and 2. Table 1
Is a collation example between papers s and r with different histories, and Table 2 is a collation example between papers s and r 'made from the same papermaking wire. For each paper sample, the power value is first integrated for each distance pixel from the center of the power diagram corresponding to steps 73 and 75, and then steps 74 and 7
6, a long wavelength component of 1 mm or more from the center, that is, 16
Pixels are removed, and then proportional calculation is performed so that the power values of the distance pixels 17 to 127 are equal to 1 by normalization corresponding to steps 80 and 82, and mutual calculation corresponding to step 83 is performed. Calculation of correlation,
Here, s × r is written, and the sum is calculated as “similarity”
I have been asked. In the example of Table 1, the similarity is low as 0.024 because the papers have different histories, and in the example of Table 2, the similarity is high as 0.88 since the papers are of the same history. I have.

[0036]

[Table 1]

[0037]

[Table 2]

FIG. 7 shows a part of these power values. FIG. 7 is a diagram in which normalized power values of paper s, r, and r ′ are plotted with respect to distance pixels from the center of the power diagram. The papers s and r ′ having a high degree of similarity and the two upper lines pass through a place that is close to the appearance. Papers with low similarity
It can be seen that the one above r ′ and the one below r ′ are apart.

FIG. 8 shows the result of the collation. FIG.
FIG. 9 is a diagram showing an example of calculating the degree of similarity between a plurality of samples. a, b ,,, f is the name of the sample paper sample, A,
B,, and F are the names of the standard paper samples prepared for reference, obtained from the original wire from which the sample samples were placed in the database. That is, A and a are paper samples obtained by making paper with the same wire, and B and b, and C and c are paper samples obtained by making paper with the same wire. The height of each peak indicates the degree of similarity, and it can be seen that only papers having the same history show a high degree of similarity and are well collated.

In the above embodiment, the wavelength range of 0.65 mm or less and 0.127 mm or more is used as the cycle derived from the papermaking process. These values were determined for the following reasons. In the case of handmade papermaking wire, the cycle is 0.42 in the case of the coarsest 60 mesh in practical use.
mm, which is practically the finest 150 mesh and 0.17 mm. Therefore, if attention is focused on those ranges, a range from about 0.1 mm to about 0.6 mm may be set as the determination range. Although the weaving method of the papermaking wire of the actual machine is more complicated, it may be determined according to the study of the inventors that the wire falls within this range. The lower limit of 0.127 mm in the embodiment is derived from the resolution of image processing, and is 25 at 400 dpi.
When analyzing six pixels, the minimum resolution is naturally 0.127 mm. This value can be made finer if the resolution of image capture is made finer (400 dpi in the embodiment) or the number of pixels taken up is increased (256 pixels in the embodiment), but even if the period of the papermaking wire or felt is actually the minimum. Since it is about 0.17 mm, there is no need to make it so fine, and this value is sufficient.

The resolution of 400 dpi and 256 pixels in the embodiment is determined as follows. In other words, the period derived from the papermaking process of interest is 0.1 mm to 0.6 mm.
In the case of about mm, the minimum unit of a pixel must be smaller than those values. Then, since one pixel is 0.127 mm at 200 dpi, it must be finer than that. Therefore, it is considered that 400 dpi to 600 dpi is sufficient. The 256 pixels are defined as a range in which a period derived from the paper making process is several tens of times when the image is taken in at the resolution (here, 400 dpi).

When the resolution is further reduced, for example, to 1200 dpi, the range of taking in an image with the same number of pixels is reduced, and the number of times that a cycle derived from the papermaking process is inserted is reduced, thereby lowering the accuracy of the determination. When the number of captured pixels is increased, the amount of calculation increases.
Therefore, 400 dpi and 256 pixels were considered to be appropriate values.

[0043]

By paying attention to the periodicity of the fibers of the paper constituting the paper, the frequency of the image of the paper is analyzed, and the characteristics of a specific wavelength range, for example, a wavelength range including a periodic component derived from the papermaking process are identified. This makes it possible to determine which paper in the database has a higher similarity to the paper.

[Brief description of the drawings]

FIG. 1 is a block diagram schematically showing a configuration of an embodiment of a paper identification device of the present invention.

FIG. 2 is an actual example of a power diagram of a two-dimensional fast Fourier transform obtained by performing FFT processing on a paper sample.

FIG. 3 is an explanatory diagram for explaining power diagram information.

FIG. 4 is a general flow chart for judging which standard paper sample is the same as a paper sample.

FIG. 5 is a flowchart for explaining in detail a collation step 40 in FIG. 4;

FIG. 6 is a flowchart illustrating an algorithm of a cross-correlation method for performing similarity calculation in FIG. 5;

FIG. 7 is a diagram in which normalized power values of paper s, r, and r ′ are plotted with respect to distance pixels from the center of the power diagram.

FIG. 8 is a diagram showing an example of calculating the degree of similarity between a plurality of samples.

[Explanation of symbols]

 Reference Signs List 2 imaging unit 4 paper capture drive 6 transmission light source 8 CCD camera 10 computer 12 image capture unit 14 memory 16 external storage device 17 database unit 18 frequency analysis unit 20 calculation unit 22 output unit

Claims (6)

[Claims] An image data obtained by transmitting or reflecting at least a part of a sample paper sample is subjected to a frequency analysis, and each integrated data for a plurality of wavelength ranges is obtained from the data after the frequency analysis. A paper identification method for determining the correlation between the integrated data with respect to the integrated data corresponding to the same wavelength range of the control standard paper sample and determining whether the sample paper sample is the same as the standard paper sample. 2. A plurality of data of the standard paper sample, wherein the correlation is obtained for each standard paper sample, and it is determined that the standard paper sample having the highest correlation is the same paper sample as the sample paper sample. The paper identification method described. 3. The periodic component according to claim 1, wherein said plurality of wavelength ranges are derived from at least one of a papermaking wire in a papermaking process, a fabric joint used for suction dehydration, and a fabric joint used for drying. 3. The paper identification method according to claim 1, wherein the wavelength range includes at least the wavelength range included. 4. The paper identification method according to claim 3, wherein the periodic component derived from the papermaking step is a periodic component derived from a papermaking wire. 5. The paper identification method according to claim 4, wherein said frequency component is 0.6 mm or less. 6. An image data acquisition section for obtaining image data by light transmitted or reflected by at least a part of a sample paper sample, a frequency analysis section for performing frequency analysis on the obtained image data, and data from the frequency analysis section. A calculation unit that calculates each integrated data for a plurality of wavelength ranges from the integrated data, and calculates a correlation of the integrated data with respect to the integrated data corresponding to the same wavelength range of the reference paper sample. A paper discriminating device having a judgment unit for judging whether or not it is the same as a standard paper sample.