JPH1114557A - Coin recognition system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To recognize a contaminated, rusted or deformed coin, as well as a true coin, by determining an optimal threshold value for binarization from an image data being read out depending on the surface conditions, e.g. contamination, rust or deformation. SOLUTION: When a coin 1 is thrown in, a light source 4 is lighted and a reflected light from a half mirror 6 is further reflected on the coin surface to enter into a reading section 7. A coin type decision section 9 decides the coin type through comparison with a reference outside diameter data. A binarization section 10 binarizes the image of the coin 1. An appropriate threshold value is set such that the fluctuation of output due to contamination, rust and deformation of the coin 1 has no effect. A data binarized based on the threshold value is outputted to a reference data extracting section 11. A true/ false decision section 12 takes out a relevant reference image from several reference images prestored in a registration pattern storing section 13 and makes a comparison. In addition to the decision of the true coin and the type thereof, a decision is made depending on the surface conditions, e.g. contamination, rust or deformation of the coin.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a coin which is mounted in an automated machine such as an automatic transaction machine (ATM) installed in a financial institution or the like and which is inserted by a customer in an automated process such as a deposit and deposit or a transfer process. The present invention relates to a coin recognizing device for recognizing types and authenticity.

[0002]

2. Description of the Related Art Conventional coin recognition devices include, for example,
Japanese Patent Application Laid-Open Nos. Hei 4-195577 and Hei 6-12548
Japanese Patent Application Laid-Open Publication No. H10-163, etc. are known. Japanese Unexamined Patent Publication No. Hei 4
Japanese Patent Application Laid-Open No. 195,577 discloses a method of obtaining a center point in image data of a coin, providing a radial window, extracting an image, and comparing the extracted image with reference data of a pattern prepared for each window of a coin in advance. Thus, coins are recognized. In addition, Japanese Unexamined Patent Application Publication No.
Japanese Patent Application Laid-Open No. 12548 discloses a pattern of a small area cut out in an annular shape in image data of a coin. The type, the front and the back are identified, and thereafter, the coin is recognized by discriminating the authenticity from the pattern having an area larger than the small area cut out in the annular shape.

[0003]

However, coins actually circulating in the world gradually become deformed or worn out due to dirt, rust, scratches and the like. For this reason, a coin recognizing device mounted on a cash (coin) automatic depositing / dispensing machine at a financial institution not only discriminates the authenticity and denomination of a coin when recognizing the coin, but also uses a true coin. Even if the coin is dirty, rusted, scratched, etc., it will automatically detect irregularities and irregularities, and wear of irregularities, etc., so that even bad deposits such as dirt, rust, discoloration, deformation, and wear will not be dispensed even if deposited. Function is required.

However, in any of the conventional techniques described above, coins are recognized only based on the discrimination criteria of discriminating the authenticity and denomination of coins, so that even true coins can be stained, rusted, discolored and deformed. In addition, there is a problem that it is not possible to make an accurate determination of an object having severe wear. That is, even if
Even if the coin is dirty, rusted, discolored, or severely worn, if it is determined to be a true coin, such coins will be dispensed. Therefore, there is a problem that the circuit scale and the device become large. Also, if such a coin is determined to be something other than a true coin, such a coin cannot be deposited and will be refunded as a reject, so the customer must bring it to the counter and burden the customer. There is a problem to put on.

Therefore, in addition to authenticity, the degree of dirt, rust, discoloration, deformation, and the like is determined, and a function of preventing coins of inferior degree from being dispensed is provided. Foreign coins, counterfeit coins, and game coins are completely removed. There has been a demand for a coin recognition device that does not allow deposits.

[0006]

SUMMARY OF THE INVENTION Accordingly, the present invention provides a coin recognition apparatus for recognizing coins by discriminating the authenticity or dirt of coins from binary data of images of surface patterns of various coins. Experimentally read out, from the read image data, extract a frequency distribution representing the frequency of output according to the surface condition such as dirt, rust, discoloration, etc., according to this frequency distribution for binarization An optimal threshold was determined. In addition, it is divided into each area for each threshold distribution range according to the surface condition such as dirt, rust, discoloration, etc., at the time of coin recognition,
It is preferable to determine which area the read image data belongs to to recognize the condition of the surface. In addition, the binary data of the read image is compared with reference data prepared in advance by matching, and correlation data is output based on the correlation between the number of matched pixels and the number of unmatched pixels to determine whether the image is true or false. preferable. Further, as reference data for comparison, it is preferable that an image rotated by one rotation at a predetermined step angle is prepared in advance and stored.

Further, it is preferable to determine the authenticity, dirt, rust, discoloration and the like from the combination of the extracted threshold value judgment data and the correlation data, and to control the deposit and withdrawal of coins. By the way, in the above coin recognition device, the denomination is determined from the read image data, and the binarization means of the image data may be changed according to the denomination. In addition,
The denomination can be determined based on the outer diameter data extracted from the image data, and the center position data is extracted from the read image data, and this center position comes to the center point position of the read image. The image data may be corrected as described above. Furthermore, the range for reading coins may be set to be equal to or larger than the diameter of the coin having the largest diameter among coins to be recognized.

[0008]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is an explanatory diagram of the configuration of the embodiment. FIG. 1A shows a state in which the coin 1 being transported is viewed from above, and FIG. 1B shows a state in which the coin 1 is viewed from the side.

In FIG. 1, reference numeral 2 denotes a conveyance path, which is provided with a slit-shaped window 3. The coin 1 is transported on the transport path 2 in the direction of arrow A in the figure, for example. Reference numeral 4 denotes a light source, for example, a line light source such as a fluorescent lamp or a line filament lamp. The light from the light source 4 is used through the slit 5. Reference numeral 6 denotes a half mirror, which is installed so that the direction of light irradiation from the slit 5 is perpendicular to the surface of the coin 1.

Reference numeral 7 denotes a reading unit which is constituted by, for example, a line image sensor such as a CCD, and is set on a vertical axis of the coin surface so as to receive specularly reflected light from the coin surface through the half mirror 6. Note that the specularly reflected light is reflected light received by the reading sensor when the light source 4 and the reading unit 7 are installed symmetrically with respect to a vertical axis with respect to the surface of the coin 1.

Each of the above components is irradiated with light from the light source 4 through the slit 5, the irradiation direction of the light is in the direction of the vertical axis of the coin surface by the half mirror 6, and the reading unit 7 is positioned on this vertical axis. And is arranged to receive the reflected light from the coin surface. In addition, the above-described components such as the window 3 and the reading unit 7 are set to have a length equal to or longer than the diameter of the coin having the largest diameter among the coins to be recognized, and can scan all of the coin types to be recognized. There is.

Reference numeral 8 denotes an image capturing unit, and a reading unit 7
1 which is analog image data read by
The data of the reflected light intensity is captured, A / D converted to create digital image data, and this is written and stored in an image storage unit (not shown) such as a frame memory. The stored digital image data is sent to a denomination determining unit 9 and a binarizing unit 10 described below. The image capturing unit 8 can also extract the outer diameter of the coin 1 and its center point.

The denomination judging unit 9 determines a denomination corresponding to the outer diameter data of the coin 1 output from the image capturing unit 8, and outputs the denomination data to a subsequent processing unit as described later. have. The binarizing unit 10 sets an optimal threshold for performing binarization from the digital image data, and outputs binarized image data by comparing the threshold with the image data. In particular, it has a function of switching the binarization means based on the denomination data obtained by the image capturing unit 8. In addition, the binarizing unit 10 has a function of detecting the degree of dirt, rust, discoloration, and the like of the coin 1 based on the magnitude of a threshold value for performing binarization.

Reference numeral 11 denotes a reference data extraction unit which outputs a boundary line component from the binary image output from the binarization unit 10 so as to obtain a boundary between a white image and a black image. It has the function of outputting a binary image that has been compressed as much as possible. Reference numeral 12 denotes a true / false judgment unit.
A compressed binary image from the reference data extraction unit 11,
It has a function of comparing the corresponding registered pattern stored in a registered pattern storage unit 13 to be described later to determine the authenticity of the coin 1, and uses reference data of the coin surface and outer diameter data of the coin. Then, a corresponding one of the reference data stored in advance in the registered pattern storage unit 13 is selected, and the reference data of the coin surface is compared. In addition, the authenticity judgment unit 12 outputs the judgment data of the compressed binary image from the reference data extraction unit 11 and the judgment data such as dirt, rust, and discoloration from the binarization unit 10 from the output. It has a function to make all judgments such as true / false, dirt, rust, and discoloration.

The registration pattern storage unit 13 stores, for example, an RO
M and a storage means such as a magnetic disk device. Further, the registration pattern as the reference data is, for example, 500 yen, 100 yen, 50 yen,
When 10 yen, 5 yen, and 1 yen are to be accepted, each coin is stored as data that has been subjected to space conversion described below in advance. It should be noted that the type of coin is not limited to the case described above, and can correspond to coins to be issued.

Next, the operation of the coin recognition device having the above configuration will be described. When the coin 1 is inserted into the coin recognition device, the light source 4 is turned on, and the light passing through the slit 5 is reflected by the half mirror 6 to irradiate the lower surface of the coin 1. Light specularly reflected from the coin surface passes through the half mirror 6 and enters the reading unit 7. Here, the concavo-convex pattern on the surface forming the pattern of the coin 1 is clearly extracted as described below.

FIG. 2 is an explanatory view of the extraction of the concavo-convex pattern, showing the concavo-convex portions on the surface forming the pattern of the coin surface. In FIG. 1B, the light reflected by the half mirror 6 illuminates the lower surface of the coin so that the light is irradiated upward from below. However, in FIG. Is expressed. That is, in FIG. 2, the irradiation light 14 from the half mirror 6 is reflected on the surface of the coin 1 and is reflected as the reflected light 15 from the coin 1.

First, in FIG. 2A, the irradiation light 14 irradiates the flat portion of the coin 1, so that the reflected light 15a goes in the opposite direction of the irradiation light 14. As a result, the reflected light 15
“a” passes through the half mirror 6 and enters the reading unit 7 as specularly reflected light as shown in FIG. On the other hand, in the case shown in FIG. 2B, the irradiation light 14 irradiates the upwardly raised portion of FIG.
b goes in a direction different from the direction of the specularly reflected light. Therefore, the reflected light 15b does not go to the half mirror 6, and does not enter the reading unit 7.

In FIG. 2C, since the irradiation light 14 irradiates a flat portion on the convex portion on the coin surface, the reflected light 15c has the same value as in FIG. The light passes through the half mirror 6 as regular reflection light and enters the reading unit 7. Further, in the case shown in FIG. 2D, the irradiation light 14 irradiates the lower left portion in FIG. 2 on the coin surface, so that the reflected light 15d goes in a direction different from the direction of the regular reflection light. Therefore, the reflected light 15d does not enter the reading unit 7 as in the case of FIG.

Thus, the pattern on the surface of the coin 1 is
In a part where the state of the coin is flat, the specular reflection light makes the coin brighter, while in an obliquely inclined part, it appears as a shadow and darkens.
The uneven pattern on the surface can be read. less than,
The calculation of the center point coordinates ( _Xc , _Yc ) will be described with reference to the drawing.

FIG. 3 is an explanatory diagram for explaining the extraction of the center point array in the x direction, and FIG. 4 is an explanatory diagram for explaining the extraction of the center point array in the y direction. The figure shows the x and y axes,
This corresponds to the x-axis end point coordinates X _s and X _e of the coin 1 is determined, end point coordinates Y _s and element Y _e coin 1 corresponding to the y-axis is determined. The center point coordinates ( _Xc , _Yc ) are calculated from the end point coordinates. In each of the figures, reference numeral 16 denotes image data stored in the storage unit. An image 17 of a white portion in the image data 16 is an image of the coin 1 and is represented as if an uneven pattern exists in the image.

In the x direction, as shown in FIG. 3, a line is drawn in the horizontal direction and the left and right end points X of the image 17 of the coin 1 are drawn.
_s (1) and X _e (1) are detected. For even two line-similarly seek X _s (2) and X _e (2), obtaining the N lines all right endpoints sequence X _s (n) and X _e (n). After the left and right end point arrays are extracted in this way, a center point array X _c (n) is extracted from each of the left and right end point arrays. That is, for the first line, X _c (1) =
Determined by _{(X s (1) + X} e (1) / 2), similarly extracts the center point sequence X _c (n) for all N lines.

The center point X _c is obtained from the center point X _c (n) of the image 17 of the coin 1 in the x direction. That is, X _c
= (ΣX _c (n)) / N. In the y direction, a line is drawn in the vertical direction as shown in FIG.
The upper and lower points Y _s (1) and Y _e (1) of the image 17 of the coin 1 are detected. Similarly, Y _s (2) and Y _e (2) are obtained for the second line, and upper and lower end point arrays Y _s (m) and Y _e (m) are obtained for all M lines. After the upper and lower end point arrays have been extracted in this way, they are obtained from the upper and lower end point arrays and the center point array Y _c (m).

The center point _Yc is obtained from the center point _Yc (m) of the image 17 of the coin 1 in the y direction. That is, Y _c
= (ΣY _c (m)) / M. Returning to FIG. 1, the image capturing unit 8 sets the coin 1 obtained in this manner.
The center point coordinates (Y _c , Y _c ) of the image 17 are output to the reference data extraction unit 11, the image written in the storage unit (not shown) is output to the binarization unit 10, and the image 1 of the coin 1 is output.
7 and outputs the outer diameter data to the denomination determining unit 9.

The denomination discriminating section 9 performs denomination discrimination by comparing with reference outer diameter data stored in advance for each denomination from the image capturing section 8. That is, the minimum value of Y _s (m) and Y
The difference between the maximum value of _e (m) or the minimum value of X _s (n) and X
The average value of the difference between the maximum values of _e (n) is used as the outer diameter data, and this is compared with the reference outer diameter data stored in advance. If data other than the outer diameter data of these domestic coins is extracted from foreign coins, an error signal is output and coin 1 is excluded.

The denomination judging section 9 determines the binarizing means of the binarizing section 10 according to the determined denomination. There are various methods for binarization such as a method using a fixed threshold and a method using automatic threshold determination such as Ptile. When the state of the surface of the coin 1 is dirty, rusted, or discolored, the fixed binarizing means usually cannot extract the uneven pattern. In addition, the amount of information of the concavo-convex pattern on the surface of the coin 1 may be different between the front and the back. Therefore, in the authenticity determination, the reading configuration reads at least one side of the coin 1. If either one of the front and back sides of the coin 1 is true, it is sufficient to be able to determine that the coin is true. , Rust,
A binarizing means which is not affected by the discoloration and the difference in the amount of information on the front and back uneven patterns is selected. Depending on the denomination, there are those made of rust-resistant white copper and those made of rust-resistant brass, and the optimum binarization means differs depending on the material. As described above, it is necessary to previously determine the binarizing means for each denomination based on parameters such as rust, dirt, and discoloration, the information amount of the uneven pattern, the area of the flat portion, and the number of pixels. For this reason, in the present embodiment, the denomination judging unit 9 is made to determine the denomination and to determine the binarization means.

The denomination discriminating section 9 discriminates the denomination by comparing it with the reference outer diameter data, and outputs the denomination data to the binarization section 10. According to the binarization means output from the denomination discriminating section 9, the binarization section 10 uses the binarization means to read the image taking section 8
The image of the coin 1 obtained in the above is binarized. Hereinafter, an example of the binarization operation will be described.

The binarizing section 10 firstly receives the image taking section 8
The image data obtained as described above is taken out, for example, by reading it from the image storage unit 9, and the occurrence frequency distribution of the density representing the gradation value is extracted. In order to binarize the image of the coin 1, it is necessary to determine an appropriate threshold value so as not to be affected by fluctuations in output due to dirt, rust, discoloration and the like of each coin 1 and fluctuations of the reading unit 7 with time. Good. For example, calculating the average value of the density distribution and determining the optimum threshold value based on the calculated average value.

FIG. 5 is an explanatory diagram for determining a threshold value. In the figure, the horizontal axis indicates the density of the target image, and the vertical axis indicates the frequency of occurrence of this density. The figure shows the case where the average value of the density is set as the threshold value 18. However, in addition to this, for example, when the distribution shape of the density occurrence frequency has two peaks, the valley of the peak is set as the threshold value. There is a way to decide. From these, a binarization method most suitable for the properties of the coin to be processed is selected, and the threshold value is determined. In order to determine the most suitable binarization method for each of the denominations, it is preferable to empirically determine the binarization method from the coin. The binarized data binarized based on this threshold is output to the reference data extraction unit 11.
Output to

The binarizing unit 10 determines the degree of dirt, rust, discoloration and the like of the target coin 1 based on the determined threshold value. If the determined threshold is significantly higher or lower than usual from the density occurrence distribution extracted from the image data of the coin 1, it is regarded as a counterfeit, falsified or foreign coin, and is determined to be excluded. .

FIG. 6 is an explanatory diagram for explaining an operation example of the threshold value judgment. In the figure, the horizontal axis indicates the density of the target image, and the vertical axis indicates the frequency of occurrence of this density. On the horizontal axis, for example, the left side has a bright value, that is, a low density value, and the right side has a dark value, that is, a high density value. In the density generation distribution extracted from the image data of the coin 1, the average distribution of the coins in circulation has a waveform as shown by, for example, 20a in the figure, and the threshold value is determined to be 20b.

However, when the intensity of the reflected light increases like a new coin, the distribution has a waveform like 19a in the figure, and the threshold value is determined to be 19b accordingly.
Also, coins with many stains, rust, discoloration, etc.
The intensity of the reflected light decreases, the distribution becomes a waveform like 21a in the figure, and the threshold is determined to 21b accordingly.

Here, the optimum threshold value to be determined is coin 1
Fluctuates depending on the state of the surface, but in the present embodiment,
It has a function of determining the state of the coin surface based on the determined threshold value. That is, when the determined threshold value belongs to the area A1 or the area A3 in the drawing, it is determined that the surface of the coin 1 is different from a normal coin. For example, when the coin 1 does not have an uneven pattern and the entire surface is specularly reflected, the determined threshold value belongs to the area A1.
When the surface of the coin 1 is extremely dirty, rusted, or discolored, the property that the determined threshold value belongs to the area A3 is used.

As described above, the threshold determination data (A1, A
In order to accurately output (2, A3), a large number of coins are read for each coin, and the coins are determined by experiments. In addition, the binarization unit 10 determines that the threshold determination result of the target coin 1 is A1
Alternatively, when the value is A3, the result is output to the authenticity determination unit 12 described later instead of excluding the coin 1 on the spot, and the authenticity determination unit 12 is referred to as being excluded. The coin 1 is finally determined. Further, the binarization unit 10 outputs such threshold determination data to a true / false determination unit 12 described below.

Next, the configuration of the reference data extraction unit 11 will be described. FIG. 7 is a block diagram of the reference data extraction unit.
Reference numeral 22 denotes a boundary component extraction unit, which converts the binarized image 17
The white / black boundary line is extracted. The boundary line component extraction unit 22 performs the extraction using a template described later, for example.

FIG. 8 is an explanatory diagram of an example of a template for extracting a boundary line component. As shown in the figure, a template 25 composed of 3 × 3 is exemplified here. The template 25 has a white pixel 26 and a black pixel 27, and is configured so that a white / black boundary passes through the center of the template. If the data to be recognized matches any one of the templates in the figure, it is determined that the data is a white / black boundary line, and a black pixel is output in correspondence with the boundary line component. Get an image.

Returning to FIG. 7, reference numeral 23 denotes a center position correction unit which converts the image output from the boundary line component extraction unit 22 into
The center position is corrected based on the center point data output from the image capturing unit 8. A compression unit 24 reduces the size of the binary image output from the boundary line component extraction unit 22 using a generally known compression method.

Hereinafter, an operation example of the reference data extraction unit 11 will be described with reference to the drawings. The reference data extraction unit 11 first extracts a white / black boundary line of the image binarized by the boundary line component extraction unit 22. Then, as shown in FIG.
A 3 × 3 template 25 has white pixels 26 and black pixels 27, and is configured such that a white / black boundary passes through the center of the template. If it matches any of the templates in the figure, it is a white / black boundary line and a black pixel is output.

The center position correcting unit 23 converts the image of the coin 1 in the binary image output from the boundary line component extracting unit 22 into
Correct to a fixed position. That is, in the image reading unit 8, since the position of the coin 1 is not always constant and the outer diameter varies depending on the type of the coin 1, the coin reading is corrected so as to unify them all at the center position. By doing so, a highly accurate discrimination result can be obtained by rotation matching or the like.

The compression section 24 converts the binary image output as described above so as to reduce the size of the image by a generally known compression method. FIG. 9 shows an example of the compression method. In the figure, reference numeral 28 denotes an image output from the boundary line component extraction unit 22, and reference numeral 29 denotes a sub-region obtained by dividing the image output from the boundary line component extraction unit 22. When the image size is reduced to one-quarter in both the x and y directions, for example, both vertically and horizontally, the sub-region has a size of 4 × 4. If there is at least one black pixel among the 16 pixels in this sub-region, the output image is compressed so as to be black. In this embodiment, an example in which compression is performed to a quarter has been described.
It may be 1/8 or 1/8. The larger the compression ratio, the smaller the data amount, the smaller the processing amount, and the higher the speed. However, since the accuracy of the determination is affected, it is better to empirically determine how much compression can be performed here.

FIG. 10 is a block diagram of the authenticity judging unit 12. Hereinafter, an operation mode of the authenticity determination unit 12 will be described. 3
Reference numeral 0 denotes a matching unit that compares the binary image output from the reference data extraction unit 11 with a reference registration pattern stored in advance from the registration pattern storage unit 13 and outputs a matching degree and a mismatch degree described later. Things. The registration patterns stored in advance in the registration pattern storage unit 13 have data prepared by rotating the front and back surfaces of each denomination by a predetermined angle.

Numeral 31 denotes a counter section, which is arranged in order of data rotated by a predetermined step angle in the order of 1, 2, 3,..., N,.
There are two arrays numbered N. The first one assigns the degree of coincidence output from the matching unit 30 for each predetermined step angle. The other is to assign the degree of inconsistency output from the matching unit 30 for each predetermined step angle. It has a function of extracting the minimum value of the degree of mismatch from this array.

Numeral 32 denotes a correlation data output unit, based on the data output from the counter 31 and the threshold value judgment data output from the binarization unit 10, the authenticity, dirt, rust, etc. of the coin 1.
Discoloration and others are determined. Hereinafter, the authenticity determination unit 1
Operation 2 will be described with reference to the drawings. Based on the denomination data output from the image capturing unit 8, the authenticity discriminating unit 12 first stores a corresponding reference image among several reference standard images stored in the registration pattern storage unit 13 in advance. Are taken out and compared.

FIG. 11 is an explanatory diagram for explaining an operation example of the matching unit 30. In the figure, reference numeral 33 denotes a reference image output from the compression unit 24, and an image 34 of the coin 1 is included in the binary image. Z is a reference image stored in the registration pattern storage unit 13 in advance. As the reference image Z, an image rotated by a predetermined notch angle is prepared, and includes the respective images Z (1) to Z (N).

The matching section 30 includes a reference image 33 and each reference image Z (N) rotated by a predetermined step angle.
Is performed. The reference image Z stored in the registration pattern storage unit 13 is obtained by rotating the front and back surfaces of each denomination by one turn at intervals of, for example, 5 degrees, to obtain reference data. In other words, for each of the front and back sides of the denomination of each coin, one turn (5
0 °), rotate 1-N (in 5 ° increments, N is 7
There is reference data of 2).

The predetermined angle is not limited to 5 degrees, and may be, for example, 10 degrees. The larger the step angle, the smaller the amount of processing, but the accuracy of the recognition is reduced. Therefore, it is preferable to experimentally determine the optimum step angle according to the recognition target. The matching unit 30 first compares the binary image 33 output from the compression unit 24 with the reference image Z (1). In the image, compare the image output at the same position,
The coincidence degree I (1) and the disparity degree D (1) are counted. The coincidence is a value that is counted only when the image output of the binary image 33 is black and the image output of the reference image Z (1) is black, for example. Discrepancy is
For example, the value is counted only when the image output of the binary image 33 is white and the image output of the reference image Z (1) is black. The coincidence degree I (1) and the non-coincidence degree D (1) thus calculated are output to the counter unit 31.

Next, the matching unit 30 compares the binary image 33 with the reference image Z (2), and as described above,
(2) and the degree of mismatch D (2) are output to the counter unit 31. In this way, N coincidences and non-coincidences are output for each step angle. FIG. 12 is an explanatory diagram illustrating an operation example of the counter unit 31. In the figure, reference numeral 35 denotes an array of the degree of coincidence, which is composed of I (1) to I (N). The array includes the coincidences that the matching unit 30 has compared and counted for each step angle. Reference numeral 36 denotes an array of the degree of mismatch, which is composed of D (1) to D (N). The array includes the degree of inconsistency that is compared and counted by the matching unit 30 for each step angle.

The counter section 31 calculates the minimum value D (min) from the values included in the mismatch degree array 36, and calculates the degree of coincidence I (min) at the step angle at that time. That is, when the degree of mismatch is the smallest, the binary image 33 matches the reference image Z rotated at the step angle.
Alternatively, it is indicated that the state is close to agreement. Also, the degree of coincidence I (min) at this step angle
Can be said to be the larger ones. In other words, the condition that the coin 1 is true is that the minimum degree of mismatch D (min) is small and the degree of coincidence I (min) is large. Then, the counter section 31 calculates the minimum value D of the degree of inconsistency.
(min) and the degree of coincidence I (min) are output to the correlation data output unit 32.

FIG. 13 shows an example of the operation of the correlation data output unit 32. In (a) of the figure, the horizontal axis indicates the degree of non-coincidence, and the vertical axis indicates the degree of coincidence. White part B in the figure
1 is a region where the matching rate is the highest, and the mismatch degree minimum value D (m
in) If the relationship of the degree of coincidence I (min) belongs to this area, the target coin 1 can be determined to be almost a true coin. In addition, it is an area where there are many mismatched elements and few matched elements, and foreign coins, counterfeit coins,
In the case of a game coin or the like, the coin enters the area B3.
In addition, coins such as coins that are very dirty, rusted, and severely discolored enter the area B2.

In the correlation diagram, the correlation data output unit 32 outputs the minimum discrepancy value D (m
in) and the relationship of the degree of coincidence I (min) at the notch angle, and outputs correlation determination data as to which of the area B1, the area B2, and the area B3. Also, the authenticity, dirt, rust, discoloration, etc. of the coin 1 is determined from the relationship between the threshold determination data output from the binarization unit 10 and the correlation determination data. A recognition result indicating whether it is possible or not is possible to output and withdraw money.

FIG. 9B shows an example of a template for outputting a recognition result from the relationship between the threshold value judgment data and the correlation judgment data. For example, when the threshold determination data is A2 and the correlation determination data is B1, a determination is output that the target coin 1 can be deposited and withdrawn. For example, when the threshold determination data is A2 and the correlation determination data is B2,
A determination is made that the target coin 1 is only allowed to be paid in and is not allowed to be withdrawn.

As described above, the combination of the threshold value determination data and the correlation determination data is set so as to output a recognition result indicating whether payment is possible, only payment is possible, and neither payment nor payment is possible. In the above description, the correlation determination data is classified into three stages, but may be classified into four stages. As described above, since the degree of dirt, rust, and discoloration of coins differs for each denomination, the respective results may be determined by experiments, and a combination may be set based on the results. In such a method, not only the authenticity of the coin 1 but also dirt and rust,
The degree of discoloration or the like is also determined, and the deposit and withdrawal of the coin 1 is controlled.

Further, an image copy of a coin described in a catalog of coins, or an image similar to a read image by the reading configuration according to the embodiment of the present invention in which a flat portion is bright and which is similar to such an image. Even if is affixed to the surface of the coin 1, this coin can be completely excluded by means for determining the threshold value of the binarization unit 10. According to the coin recognition device of the above embodiment, an image of the entire surface of a coin requiring recognition of a coin is read, the outer diameter of the coin is extracted from the image, and the discrimination data to be recognized is coins. Since the binarization means is changed so as not to be affected by local contamination, rust, discoloration, and the difference in the amount of information on the front and back surfaces of the uneven pattern, the effect of always performing stable recognition can be obtained.

The coordinates of the center point of the effect are calculated by extracting the outer diameter data, and the data of the image is corrected so that the calculated coordinates of the center point are read at the center position of the image. As a result, processing such as comparison with reference data can be performed accurately, and a coin that can accurately recognize coins can be obtained. Further, the binarizing means has an effect that the degree of dirt, rust, discoloration and the like on the coin surface can be determined based on the magnitude of the determined threshold value.

Further, by performing the maximum compression of the binarized image in accordance with the minimum necessary concavo-convex pattern information on the coin, the image size can be reduced, the processing amount can be reduced, and the processing time can be reduced. A coin that can be shortened is obtained. Therefore, an effect of suppressing variations in coin transport speed can be expected.

Further, since the reference data registered based on the outer diameter data of the coin to be processed is selected,
The effect of omitting unnecessary determination processing and shortening the processing time is obtained. Further, in the comparison between the target coin image and the reference image, not only the authenticity is determined from the correlation between the degree of coincidence and the degree of disagreement, but also the dirt, rust, discoloration and other determination data are combined to determine. As a result, an effect is obtained that even a malicious coin on which a copy or the like is pasted can be accurately recognized.

[0057]

As described above, according to the coin recognition device of the present invention, in addition to the discrimination criteria of discriminating the authenticity and denomination of coins, the surface of stains, rust, discoloration, deformation, wear, etc. Since the discrimination according to the condition is performed, it is possible to obtain an effect of accurately recognizing a coin even if it is a true coin, which is very dirty, rusted, discolored, deformed, or worn. For this reason, it is possible to expect an effect that the size of the device can be reduced and the burden on the customer can be reduced. further,
Since the effect of the target can be accurately recognized, an effect of reliably distinguishing the coin from foreign coins other than the target coin, counterfeit coins, and game coins is obtained.

[Brief description of the drawings]

FIG. 1 is a configuration explanatory view of an embodiment.

FIG. 2 is an explanatory diagram of extraction of a concavo-convex pattern

FIG. 3 is an explanatory diagram illustrating extraction of a center point array in the x direction.

FIG. 4 is an explanatory diagram illustrating extraction of a center point array in the y direction.

FIG. 5 is an explanatory diagram of threshold determination.

FIG. 6 is an explanatory diagram illustrating an operation example of threshold value determination.

FIG. 7 is a block diagram of a reference data extraction unit.

FIG. 8 is an explanatory diagram of a template for extracting a boundary line component.

FIG. 9 is an explanatory diagram of a compression method.

FIG. 10 is a block diagram of a true / false judgment unit;

FIG. 11 is an explanatory diagram illustrating an operation example of a matching unit.

FIG. 12 is an explanatory diagram illustrating an operation example of a counter unit.

FIG. 13 is an explanatory diagram illustrating an operation example of a correlation data output unit.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 coin 4 light source 7 reading unit 8 image capturing unit 9 denomination determination unit 10 binarization unit 11 reference data extraction unit 12 authenticity determination unit 13 registration pattern storage unit

Claims (5)

[Claims] 1. A coin recognition device for recognizing a coin by discriminating the authenticity or dirt of a coin from binary data of an image of a surface pattern of various coins. From the image data obtained, a frequency distribution representing the frequency of output according to the surface condition such as dirt, rust, discoloration, etc. is extracted, and an optimal threshold value for binarization is determined according to this frequency distribution. A coin recognition device characterized by the above-mentioned. 2. A method according to claim 1, wherein each area is divided into respective areas according to a distribution range of a threshold value according to surface conditions such as dirt, rust, and discoloration. A coin recognizing device characterized in that it is determined whether or not it belongs to recognize the condition of the surface. 3. The method according to claim 1, wherein the binary data of the read image is compared with reference data prepared in advance by matching, and correlation data is output by correlation between the number of matched pixels and the number of unmatched pixels. A coin recognition device for determining whether the coin is true or false. 4. The coin recognition device according to claim 3, wherein the reference data for comparison is prepared and stored in advance as an image rotated by one turn at a predetermined step angle. 5. The method according to claim 1, wherein each of the areas is divided into respective areas according to a distribution range of a threshold value according to the condition of the surface such as dirt, rust, discoloration, etc. The state of the surface is recognized by judging whether the image data belongs to the image, the judgment result is generated as threshold judgment data, the binary data of the read image is compared with reference data prepared in advance by matching, and the number of matched pixels is determined. Correlation data is output based on the correlation of the number of pixels that did not match, and the authenticity is determined.The combination of the extracted threshold determination data and the correlation data is used to determine true / false, dirt, rust, discoloration, etc. A coin recognizing device characterized by controlling gold.