JPH11306411A - Paper sheet identification device - Google Patents

Abstract

(57) [Problem] To provide a general and effective identification that can be applied to various applications by characterizing an image by setting a weight to the image to be identified, thereby enhancing the identification effect, and applying to various applications. It is an object of the present invention to provide a paper sheet identification device capable of performing the following. A sheet identification apparatus according to the present invention includes an image reading unit that reads an image of a sheet, a first storage unit that stores identification reference data serving as a reference for identification of the sheet, and a reading unit. Computing means for calculating the similarity between the image and the identification reference data using the weighting data, and identification determination means for determining the identification result from the similarity, weighting only the identification reference data side, or By weighting both the identification reference data and the read image data, the image can be characterized, and effective identification can be performed when obtaining a similarity with the identification reference data.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

TECHNICAL FIELD The present invention relates to, for example, banknotes,
The present invention relates to a paper sheet identification apparatus that performs authenticity identification or type identification on paper sheets having a print pattern such as securities and tickets.

[0002]

2. Description of the Related Art In a conventional paper sheet discriminating apparatus, for example, an discriminating apparatus for discriminating banknotes, an analysis tool for setting an item for performing a statistical analysis process, a neural network, and the like are used to develop an identification algorithm. , Fuzzy inference, etc. are used, but in the case of analysis tools, there are many items to be tuned (adjusted) and there is a problem that automation is difficult, and in the case of neural networks, automation is possible, However, there is a problem that the amount of calculation is large and it takes time, and in the case of fuzzy inference, the items to be tuned are not always intuitively understandable to humans, and the fuzzy inference may not be determined. General with problems and applicable to various applications,
At present there is nothing effective.

[0003] As a discrimination method, a method of determining the similarity by using a normal correlation method or a Euclidean distance may be considered. However, printing discrepancies in the production stage of paper sheets and stains in a distribution process are considered. , Large variations within the same type,
It is difficult to further increase the identification accuracy beyond a certain level, and it is difficult to increase the accuracy in identifying a very similar print pattern due to the influence of printing deviation and the like.

[0004]

SUMMARY OF THE INVENTION According to the present invention, an image is characterized by setting a weight for the image to be identified.
It is an object of the present invention to provide a general and effective paper sheet identification apparatus that can enhance the identification effect and can be applied to various applications.

[0005]

SUMMARY OF THE INVENTION The present invention provides an image reading means for reading an image of a paper sheet, a first storage means for storing identification reference data serving as a reference for identification of a paper sheet, and a read image. The apparatus is characterized in that the apparatus is a paper sheet identification apparatus including: an arithmetic unit that calculates a similarity between the data and the identification reference data using the weight data; and an identification determination unit that determines an identification result from the similarity. Further, a second storage means for storing weight data in which a weight is set for the image to be identified is provided. Further, the weight data is set by setting a higher weight to a pixel having high stability of a paper sheet image. Further, the weight data is set by increasing the weight of a pixel having a large difference in tone value of an image from another sheet. Further, the arithmetic means is configured to normalize the read image data, calculate the similarity by multiplying the read image data by weight data.

[0006]

According to the present invention, it is possible to characterize an image by weighting only the identification reference data, or by weighting both the identification reference data and the read image data. When determining the degree of similarity with the identification reference data, effective identification can be performed.

Further, since the weighted image can be obtained by setting without repeating the simulation, the development period of the identification algorithm can be shortened. Furthermore, the calculation formula (weight) of the weighted image can be changed if there is no need to change the basic concept.
Even if the object changes, it can be used in common, and versatility can be obtained.

Further, by changing the calculation formula (weight) of the weight image, a weight image suitable for the application can be obtained, and the identification accuracy can be improved. If printing misalignment occurs during the manufacturing process of paper sheets, or if dirt adheres during the distribution process of the paper sheets, there is a difference in gradation from the sample (identification reference data). The position (coordinates) of dirt and dirt has a smaller weight, and the influence on identification is reduced, and good identification can be performed even in the case of these paper sheets.

Further, in the identification of paper sheets having a very similar print pattern, the weight of pixels having different gradation values is increased, so that the identification can be performed with high accuracy.

[0010]

An embodiment of the present invention will be described below with reference to the drawings. The drawing shows a paper sheet identification apparatus. In FIG. 1, a paper sheet, for example, a bill 10 is conveyed in a predetermined conveyance direction by conveyance rollers 11 which are appropriately driven and controlled, and is transmitted from a light source 12 in the conveyance path. The sensor 13 receives the light and reads a transmission image of the two-dimensional print pattern. The above-mentioned sensor 13 is formed in a line shape by, for example, a CCD array, and is provided so as to sufficiently cover the width of the bill 10 to be conveyed.

Note that the reading of the print pattern may be one-dimensional data, and the reading method may be configured so that the sensor 13 moves. Further, the reading is not a transmission image but a reflection image. Is also good.

FIG. 2 shows coordinate data (X axis 0 to N, Y axis 0 to M) of a read image. The data is a two-dimensional digital image, and each pixel is 0 (dark) to 255 (bright). )
Take the gradation values up to.

FIG. 3 shows a control configuration diagram. The CPU 20 controls the driving of the image reading section 23 and the transport section 24 based on a program stored in the ROM 21.
21 stores a weighted template (identification reference data) and a weighted image (weight data) for identification.

The RAM 22 stores data necessary for the operation, the image reading section 23 takes in the image data read by the sensor 13, and the transport section 24 drives and controls the transport rollers 11. Next, the identification processing of the sheet identification apparatus will be described with reference to the flowchart of FIG. The CPU 20 drives and controls the transport unit 24 to drive the transport rollers 11 in a predetermined transport direction, and further controls the image reading unit 23 to drive the light source 12.
Is turned on, the transmitted image of the banknote 10 is read by the sensor 13, and
Each gradation value (0 (dark)-) of each position (1 ≦ x ≦ N, 1 ≦ y ≦ M) of the coordinates (X axis 0 to N, Y axis 0 to M) of this transmission image
255 (bright)) as input image data [Equation 1] (step n1).

[0015]

## EQU1 ## Next, the input image data [expression 1] is normalized with respect to gradation by [expression 2] (step 2).

[0016]

## EQU2 ## Next, the set weight is applied to each pixel of the input image (step n3). This weight is set in advance for each class of the banknote 10 (for each denomination, 4 in the front and back, front and back directions for one denomination).
) And stored in the ROM 21 as a weighted image (weight data) [Equation 3].

[0017]

## EQU3 ## In the ROM 21, a weighted template (identification reference data) obtained by adding the set weights to the normalized image data of each class of the banknote 10 and renormalizing the data (Equation 4)
Is also stored.

[0018]

## EQU4 ## The weight setting of [Equation 3] described above is intended to identify pixels (coordinates) that are more important for identification by increasing the weight. For example, the setting method is as follows. is there.
As one of them, the weight of stable pixels is increased for the image of the banknote 10 (four types in the front and back, one in the front-back direction for one denomination).

Second, the weight of a pixel having a large difference in tone value from an image of a banknote 10 of another class (denomination) is increased. Third, the weight of a pixel having a large difference in tone value between classes of the banknote 10 that is difficult to classify (denomination) is increased.

As a fourth factor, weights are assigned so that erroneous identification hardly occurs in a class (for example, a large bill) having a high risk of erroneous identification.

In the above example of setting weights, 1 and 2
In the following, an example of a method of creating a weight image will be described. An average image is created for each class from a sample image of each class (for each denomination) for creating a template, and a variance value of the sample image in each class is further created. And a weighted image (weight data) [Equation 5] is created with these.

[0022]

## EQU5 ## Further, the above-mentioned weighted image is applied to the normalized sample input image for each pixel and renormalized to create a template image (identification reference data) [Equation 6]. The process of step n3 of applying the set weight to each pixel of the input image is performed by applying the weighted image of each class (each denomination) of the banknote 10 to the input image that has been normalized for each pixel, and normalizing again [Equation 7]. ]
Do

[0023]

Then, a normal correlation value with the weighted template image of [Equation 4] is obtained by [Equation 8] for each class of banknote 10 (calculation of similarity) (step n4).

[0024]

When the normal correlation value (similarity) NC is calculated, the value NC obtained for each class (each denomination) satisfies [Equation 9] −1 ≦ NC ≦ 1 and the normal correlation value ( It is determined that the closer the NC is to “1”, the more similar the input image is to the corresponding class, and the class closest to “1” among the values NC obtained for each class is obtained.

Next, it is determined whether or not 1 / | 1-NC | of the class closest to the above-mentioned "1" is equal to or greater than a specified value for determining the corresponding class (step n6). It is determined that the class is the class (step n7),
If it is below, it is determined that identification is impossible (step n8).

According to the above-described embodiment, as is apparent from the equation shown in [Equation 8], the comparison of the similarity based on the normal correlation value depends on the difference between the interior angles of the vectors. The similarity can be determined by the Euclidean distance method for comparing the degrees.

As described above, according to the above-described embodiment, the template image as the identification reference data,
By applying weights to the read image data, it is possible to add a feature to the image, and to determine the similarity with the template image that is the identification reference data, effective identification can be performed.

Further, since the weighted image can be obtained by setting without repeating the simulation, the development period of the identification algorithm can be shortened.

Further, when it is not necessary to change the basic concept, the calculation formula (weight) of the weight image can be used in common even if the object changes, and versatility can be obtained. Further, by changing the calculation formula (weight) of the weight image, a weight image suitable for the application can be obtained, and the identification accuracy can be improved.

If printing misalignment occurs during the manufacturing process of the banknote 10 or other paper sheets, or if dirt adheres in the distribution process of the banknote 10 or other paper sheets, the sample (identification reference data) may be used. , The position (coordinates) of printing misalignment or dirt is reduced in weight, and the influence on identification is reduced, and good identification can be performed even for these banknotes 10 and other paper sheets.

Further, in the identification of paper sheets having very similar print patterns, the weight of pixels having different gradation values is increased, so that identification can be performed with high accuracy. In the above embodiment, both the template image of the identification reference data and the read image data are weighted. However, good identification can be obtained by weighting only the identification reference data.

In correspondence between the configuration of the present invention and the above-described embodiment, the paper sheet of the present invention corresponds to the paper sheet having a print pattern such as a bill 10, a security, a ticket, etc. of the embodiment. Similarly, the image reading unit corresponds to the sensor 13 and the image reading unit 23, and the first and second storage units correspond to the ROM 21 storing the weighted template (identification reference data) and the weighted image (weight data). The calculation means and the identification determination means correspond to the CPU 20 which performs calculation processing and determination processing. However, the present invention can be applied based on the technical idea described in the claims.
The configuration is not limited only to the configuration of the embodiment.

[Brief description of the drawings]

FIG. 1 is a schematic side view of a paper sheet identification device.

FIG. 2 is a diagram showing coordinates of image data.

FIG. 3 is a control circuit configuration diagram of the paper sheet identification device.

FIG. 4 is a flowchart of an identification process.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 10 ... Banknote 13 ... Sensor 20 ... CPU 21 ... ROM23 ... Image reading part

(Equation 1)

(Equation 2)

(Equation 3)

(Equation 4)

(Equation 5)

(Equation 6)

(Equation 7)

(Equation 8)

Claims (5)

[Claims] An image reading means for reading an image of a paper sheet;
First storage means for storing identification reference data serving as a reference for identification of paper sheets; computing means for calculating the similarity between the read image and the identification reference data using weight data; A paper sheet identification device comprising: an identification determination unit that determines a result. 2. An image reading means for reading an image of a paper sheet,
First storage means for storing identification reference data serving as a reference for identification of paper sheets; second storage means for storing weight data in which weights are set for images to be identified; And a discriminating means for judging the discrimination result based on the similarity. 3. The paper sheet identification device according to claim 1, wherein the weight data is set by increasing the weight of pixels having high stability in the image of the paper sheet. 4. The paper sheet identification device according to claim 1, wherein the weight data is set by increasing the weight of a pixel having a large difference in tone value of an image from another paper sheet. 5. The paper sheet discriminating apparatus according to claim 1, wherein said calculating means calculates the similarity by normalizing the read image data and applying weight data to the read image data.