JPH0644437A - Recognizing device for paper sheet etc - Google Patents

Abstract

PURPOSE:To process data of the whole paper sheet, etc., in a short time, and to execute recognition at a high speed and with high reliability. CONSTITUTION:By plural light sources whose wavelength is different, image data to each wavelength is read by a scanner 2, and written in a buffer memory 14. Subsequently, in these image data, an area being effective for recognizing a paper sheet, etc., 1 is determined by an effective area determining part 3. Next, from a correlation of the image data of these effective areas, a feature quantity is extracted, and this feature quantity is replaced with a frequency distribution of the feature quantity. Thereafter, this frequency distribution is compared with plural frequency distribution shapes which become a reference, prepared in advance, and the kind and truth of the paper sheet, etc., 1 are recognized.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a paper sheet recognition apparatus used for recognizing the type and authenticity of paper sheets such as banknotes, securities and bonds.

[0002]

2. Description of the Related Art Conventionally, automatic deposit / withdrawal machines (AT) have been used to automate deposit / saving and transfer processing in financial institutions.
M) and automatic transaction equipment are installed. In this automatic depositing / dispensing machine, when a customer inserts a bill, the type and authenticity of the bill is recognized. This processing is performed by the bill validator, but here, first, an optical or magnetic pattern is read along one or more lines parallel to the bill transport direction. The optical pattern has certain characteristics for each denomination due to the pattern of the bill.
The magnetic pattern also has the same characteristics. The detected pattern, which is an analog signal, is compared with a predetermined threshold value and binarized at many points on the read line. The pulse train thus obtained is counted by the counter circuit, and the count value is compared with the dictionary data. Then, for all the detected lines, if the count value is close to any one of the Nakano in the dictionary data, it is possible to recognize the type of banknote and the authenticity or the front and back.

[0003]

However, the above-mentioned conventional technique has the following problems. That is, usually, the feature of a partial area along a specific line on a bill is extracted and recognized. Therefore, when a small-area forgery exists outside the detection area, it may be recognized as a genuine note. This occurs, for example, when the banknotes are partially pasted together. In addition, in order to increase the reliability of recognition, the external dimensions and thickness of the banknotes are detected, and the types of banknotes from various viewpoints,
There is also a way to recognize the truth. However, in such a method, if even one parameter is out of the standard range, it is determined to be a fake note. Therefore, in such a case, the rate of determination as a fake bill increases, which is a problem in practical use. Of course, such a problem is not limited to the appraisal of banknotes, and is similarly applicable to the appraisal of securities and other various paper sheets.

To solve these problems, first,
It is preferable that the widest possible area of the paper sheet is read at a high resolution and used as a reference for recognition so that forgery or bonding of a small area can be detected. In addition, it is preferable to recognize the type and authenticity by using one criterion having the highest reliability for improving the recognition rate. However, if the amount of data to be read is large, it takes a long time to process the data, making it difficult to perform collation at high speed. Therefore, it is not practical to use it in an automatic transaction device provided in a bank or the like. Further, in the method of reading a part of the paper sheet, if the paper sheet is misaligned in the conveyance, the detection data may fluctuate, and erroneous recognition may occur. In order to solve this, a recognition process is desired that is not affected by any positional deviation of paper sheets. The present invention has been made in view of the above points, and an object thereof is to provide a paper sheet recognizing device that is resistant to positional deviation, skew deviation, and the like, and has high speed and high reliability.

[0005]

In the paper sheet recognition apparatus of the present invention, a plurality of light sources having different wavelengths are used, and light of each wavelength is used for the paper sheets to be recognized by the plurality of light sources. A scanner that irradiates and reads image data including the entire paper sheet, and an effective area determination unit that extracts an effective area occupied by the paper sheet from the image data read by the scanner and writes the effective area in a buffer memory, The image data of the effective area is read from the buffer memory, the characteristic data is extracted from the correlation of the grayscale image for each wavelength, and the characteristic extraction unit outputs the frequency distribution of the characteristic data. A feature is provided that includes a recognition unit that compares the output frequency distribution shape with a plurality of prepared reference frequency distribution shapes to recognize the type and authenticity of the paper sheet. It is intended to.

[0006]

In the paper sheet recognition apparatus of the present invention, the minimum necessary grayscale image data of the read image is cut out by the effective area determination unit and is processed. In addition, reading is performed using a plurality of light sources having different wavelengths to obtain mutually different image data. Then, by extracting the feature of the correlation of each image data, the specific feature data corresponding to the pattern on the paper surface of the paper sheet can be obtained. Next, a frequency distribution is output from this characteristic data, and this is compared with a frequency distribution shape prepared in advance for identification to recognize the type and authenticity of the paper sheet. In this way, since the recognition is performed by the correlation between the grayscale images having different wavelengths, the recognition rate becomes high. Further, since minimum necessary data is cut out and collation is performed, high speed processing is possible.

[0007]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 1 is a block diagram of an embodiment of the paper sheet recognition apparatus of the present invention. The illustrated apparatus includes a scanner 2 that reads an image of a paper sheet 1, an effective area determination unit 3, a buffer memory 4, a feature extraction unit 5, and a recognition unit 6.
The paper sheet 1 is a recognition target of the illustrated apparatus, and is, for example, a bill, a securities, a bond, or the like. Optical or magnetic data is recorded on such paper sheets.
The illustrated apparatus is configured to read such data using the scanner 2. In the following description, an example is shown in which the paper sheet is a banknote and the optical information printed on the banknote is read using the scanner 2.

The scanner 2 is composed of, for example, an image line sensor, and is electrically scanned in the main scanning direction.
It is mechanically scanned in the sub-scanning direction. Effective area determination unit 3
Writes the image data read by the scanner 2 in the buffer memory 4. The buffer memory 4 is composed of a RAM (random access memory) or the like, and stores the grayscale image data obtained by reading the paper sheets. In this case, immediately after being read by the scanner 2, an image such as a background arranged around the paper sheet is also read at the same time. The image such as the background is removed by the effective area determining unit 3. As a result, the minimum necessary data is taken out and the effective image data area is determined. The feature extraction unit 5 converts the effective grayscale image data stored in the buffer memory 4 into a feature amount by parameter calculation. This feature amount has a certain feature depending on the type of banknote, front and back, and the like.

The recognizing section 6 compares the reference value prepared in advance with the characteristic data output from the characteristic extracting section 5, and recognizes the type, authenticity, front / back, etc. of the paper sheet based on the result. Hereinafter, the specific configuration and operation of each unit of the apparatus shown in FIG. 1 will be described in order. FIG. 2 shows the scanner 2 and the effective area determination unit 3.
3 is a block diagram showing a more specific configuration of FIG. In this figure, an image line sensor 21, an A / D conversion unit 22, a buffer memory 4, and an effective area determination unit 3 which are arranged on a paper sheet conveyance path are shown.

The image line sensor 21 is composed of a well-known optical image reading element group such as a CCD. Then, the outputs of the image line sensor when three light sources having different wavelengths irradiate the different lights are digitized by the A / D converter 22 and stored in the buffer memory 4. Light sources with different wavelengths are, for example, white light sources with red filters, green filters, and blue filters attached,
Alternatively, a light source dedicated to infrared light, a light source dedicated to ultraviolet light, a light source dedicated to visible light such as green, or the like is used. Image line sensor 2
Reference numeral 1 reads an optical analog pattern of reflected light or passing light of the paper sheet 1. This signal is A /
The D conversion unit 22 converts the digital grayscale image data. However, the grayscale image data is composed of three types corresponding to each of the three light sources.

In this way, each of the grayscale image data read into the buffer memory 4 includes the data around the paper sheet 1 and in the background portion thereof. If you try to perform subsequent processing including these data,
The processing speed cannot be increased. Therefore, first, the effective area determination unit 3 shown in FIG. 2 detects the area of the corresponding grayscale image data in the buffer memory 4. FIG. 3 is an explanatory diagram of the operation of the effective area determination unit 3. As illustrated, the buffer memory 4 stores the grayscale image data. Here, only one of the three types of grayscale image data is used to determine the effective region.

A white portion 32 at the center of the grayscale image data 31 is a grayscale image data region of the paper sheet 1 to be recognized. When writing the grayscale image data into the buffer memory 4, the effective area determination unit 3 shown in FIG. 2 addresses Xmin (y), Xma of the end position of the paper sheet 1 shown in FIG.
Detect x (y). Further, as this method, the effective area determination unit 3 obtains density waveforms based on the grayscale image data for the total width Xw in the main scanning direction and the total width Yw in the sub scanning direction of the buffer memory 4. The graph shown on the lower side of FIG. 3 is this concentration waveform. And based on these waveforms,
The threshold Mt is set. The threshold Mt is set to an intermediate value between the maximum density and the minimum density. Here, the background of the table or the like on which the banknotes are placed is colored so that the maximum density is obtained.

When reading the image shown on the upper side of FIG. 3, the image line sensor 21 is sequentially scanned in the main scanning direction in the order of the sub scanning direction. Then, in the middle of this, the pixel (x, y)
When the grayscale output value of the above becomes smaller than the threshold value Mt, when it occurs for the first time on the main scanning coordinates, the value of Xmin (y) is assigned the main scanning coordinate number at that time. Also, when the value finally becomes smaller than the threshold value Mt on the main scanning coordinate, Xmax
The main scanning coordinate number at that time is assigned to the value of (y).
By the above-described processing, when the effective area determination unit 3 has written all the grayscale image data in the buffer memory 4, the paper sheet area is extracted as the following section. Xmin (y), Xmax (y) where 0 ≦ y <Yw The grayscale image data cut out in this way is shown in FIG.
It is converted into a feature amount by the feature extraction unit 5 shown in FIG.

FIG. 4 is a block diagram showing the structure of the feature extraction unit 5. The feature extraction unit 5 includes a feature amount calculation unit 41 that calculates a feature amount based on the data read from the buffer memory 4, and a feature amount frequency distribution creation unit 42 that obtains the occurrence frequency of the feature amount. . Here, the buffer memory 4 stores Xw × Yw grayscale image data Mi, i = {1,2,3} arranged in a lattice form as shown in the following equation. M1 = {(x, y) | 0≤x <XW, 0≤y <Yw} M2 = {(x, y) | 0≤x <XW, 0≤y <Yw} M3 = {(x, y) | 0 ≦ x <XW, 0 ≦ y <Yw} The grayscale image data Mi, i = {1,2,3} is the reflected light output of each light source installed in the image line sensor 21 shown in FIG. It corresponds to. Feature quantity calculation unit 4
Reference numeral 1 reads out the grayscale image data of the above-mentioned effective area from the buffer memory 4 and calculates the characteristic amount. That is, the grayscale image data shown in the following equation is read out.

M1 = {(x, y) | 0≤Xs≤x <Xe≤XW, 0≤Ys≤y <Ye≤Yw} M2 = {(x, y) | 0≤Xs≤x <Xe≤XW , 0≤Ys≤y <Ye≤Yw} M3 = {(x, y) | 0≤Xs≤x <Xe≤XW, 0≤Ys≤y <Ye≤Yw} First, the grayscale image data read from the pixel is read. Mi (x, y), i =
As {1,2,3}, choose the lightest value among these and set this to mi
Let nMi (x, y), select the darkest value, and use this as maxMi (x,
y). The distribution of each of the read grayscale image data Mi (x, y), i = {1,2,3} of the pixel (x, y) is obtained by the calculation shown in the following equation. D1 (x, y) = {maxMi (x, y) -M1 (x, y)} / {maxMi (x, y) -minMi (x, y)} D2 (x, y) = {maxMi (x, y) -M2 (x, y)} / {maxMi (x, y) -minMi (x, y)} D3 (x, y) = {maxMi (x, y) -M3 (x, y)} / { maxMi (x, y) -minMi (x, y)}

For example, if M1 (x, y) is equal to the darkest value maxMi (x, y) in the read grayscale image data of the pixel (x, y), D1 (x, y) is 0.0. Becomes Conversely, M1 (x, y)
Is equal to the lightest value minMi (x, y) in the grayscale image data read from the pixel (x, y), D1 (x, y) becomes 1.0. Further, M1 (x, y) is not equal to the darkest value maxMi (x, y) in the read grayscale image data of the pixel (x, y) and the lightest value minMi (x, y). In some cases, 0.0 <
D <1.0. In this way, M2 (x, y), M3 (x, y)
Similarly, D2 (x, y) and D3 (x, y) are obtained.
From these data, the feature quantity H (x, y) at pixel (x, y)
Is calculated by the calculation as shown in FIG.

Here, it is desirable to select the wavelength of the light source from the image data M1, M2, M3 corresponding to different wavelengths so that the feature quantity H to be obtained greatly differs depending on whether the target paper sheet is true or false. . FIG. 5 is a flowchart illustrating a procedure for calculating the feature amount. First, the grayscale image data M
1 (x, y) is compared with the darkest value maxMi (x, y) (step S
1) If both are equal, then the grayscale image data M2
(x, y) is compared with the lightest value minMi (x, y) (step S2). As a result, when both are equal, the feature quantity H (x,
y) = {5 + D3 (x, y)} / 6 (step S3).
On the other hand, when they are not equal, the feature quantity H (x, y) = {1
-D2 (x, y)} / 6 (step S4).

As a result of comparing the grayscale image data M1 (x, y) with the darkest value maxMi (x, y), when they are not equal, the grayscale image data M2 (x, y) is the darkest. Value maxM
It is compared with i (x, y) (step S5). As a result, when they are equal, the grayscale image data M3 (x, y) is compared with the lightest value minMi (x, y) (step S6). As a result, when both are equal, the feature quantity H (x, y) = {1 + D1
(x, y)} / 6 (step S7). On the other hand, when they are not equal, the feature quantity H (x, y) = {3-D3 (x, y)} /
6 (step S8). Furthermore, the grayscale image data M2
As a result of comparing (x, y) with the darkest value maxMi (x, y), if they are not equal, then compare the grayscale image data M1 (x, y) with the darkest value minMi (x, y). Yes (step S9).
As a result, when both are equal, the feature quantity H (x, y) = {3
+ D2 (x, y)} / 6 (step S10). On the other hand, when they are not equal, the feature quantity H (x, y) = {5-D1 (x,
y)} / 6 (step S11). Feature amount calculation unit 41
Calculates the feature amount H (x, y) for each pixel (x, y) as described above, and outputs the feature amount H (x, y) to the feature amount frequency distribution creating unit 42.

FIG. 6 is an explanatory diagram of creating a frequency distribution of feature quantities. In this figure, the vertical axis represents the frequency of occurrence of the characteristic amount, and the horizontal axis represents the characteristic amount. Feature quantity frequency distribution creation unit 42
Reads the feature amount output from the feature amount calculation unit 41 for each pixel and creates a feature amount frequency distribution. The feature quantity frequency creation unit 42 outputs the feature quantity frequency distribution to the recognition unit 6. The recognition unit 6 compares the feature amount frequency distribution obtained from the feature amount frequency distribution creating unit as shown in FIG. 6 with a pre-stored frequency distribution shape, and based on the result, the type of paper sheet, authenticity, front / back Etc.

[0020]

As described above, according to the paper sheet recognizing apparatus of the present invention, the images of the entire region which requires the recognition of the paper sheets are read for each of a plurality of light sources, and the images between the respective images are read. Since the feature quantity representing the correlation is extracted and its frequency distribution is created, the data targeted for recognition will not be affected by the paper room or print density of the paper sheet, and stable recognition will always be possible. Become. Further, by setting the size of the area read by the scanner to the area of the target paper sheet, all necessary images of the target paper sheet can be extracted. As a result, the reliability of recognition can be improved.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of a paper sheet recognition apparatus of the present invention.

FIG. 2 is a block diagram showing a scanner and an effective amount determination unit.

FIG. 3 is an explanatory diagram of an operation of an effective area determination unit.

FIG. 4 is a block diagram showing a configuration of a feature quantity extraction unit.

FIG. 5 is a flowchart illustrating a procedure for calculating a feature amount.

FIG. 6 is a graph showing a frequency distribution of feature quantities.

[Explanation of symbols]

 1 paper sheet 2 scanner 3 effective area determination unit 4 buffer memory 5 feature extraction unit 6 recognition unit

Claims (1)

[Claims] 1. A plurality of light sources having different wavelengths, and a plurality of light sources irradiate the paper sheets to be recognized with light of each wavelength in turn, and image data including the entire paper sheets is obtained. A scanner for reading, extracting an effective area occupied by the paper sheet from the image data read by the scanner, and writing the effective area in the buffer memory, an effective area determination unit, and image data of the effective area from the buffer memory. A characteristic extraction unit that reads out the characteristic data from the correlation of the grayscale image for each wavelength and outputs the frequency distribution of the characteristic data, the frequency distribution shape output by the characteristic extraction unit, and the reference prepared in advance A paper sheet recognition apparatus, comprising: a recognition unit that compares a plurality of frequency distribution shapes to recognize the type and authenticity of the paper sheet.