JP2003233851A - Method for judging authenticity/soundness of coin from picture and its device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method and device for discriminating the authenticity/soundness of a coin from a picture. <P>SOLUTION: The picture signal of a target coin is digitally processed, and sampling data obtained by extracting the pattern of the calculated coin picture for 360° at each fixed angle of extraction on a circumference and a basic template are relatively rotated to calculated angular displacement angle whose matching level is the most satisfactory so that the denomination and surface/back face of the target coin can be decided, and that the kind of the decided denomination and surface/back face can be confirmed from the pattern of the matching level when the sampling data and the basic template deviate by a prescribed angle of rotation with the angular displacement whose matching level is the most satisfactory as a start point to judge authenticity of the coin. Then, the total sum of pixel values is calculated for each prescribed small area in the prescribed angle pitch of ring data with the angular displacement whose matching level is the most satisfactory as a start point so that a density histogram can be prepared, and this density histogram is compared with the density histogram corresponding to the basic template so that soundness of the coin can be judged. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention discriminates / determines the denomination, authenticity, etc. based on an image of a coin that is conveyed freely on the front and back sides and the rotational angle position, and further determines a similar coin or a modified / altered coin.
The identification / judgment of the case of including counterfeit coins is performed, and the authenticity of coins /
The present invention relates to a method and an apparatus for determining a fitness.

[0002]

2. Description of the Related Art The denomination of coins to be identified is determined by the outer diameter dimension and hole diameter dimension measured based on the detection data of a magnetic sensor and the imaged image data. However, at this stage, the front and back of coins cannot be determined. For the determination of the front and back sides, it is necessary to identify the image. As a first conventional technique related to image identification and authenticity determination, for example, a coin is identified by matching with a standard image of a regular coin, and then authenticity is determined by further matching with a template of similar coins. There is something I did. The problem with this method is that each time a similar coin appears, a new template must be prepared, so there is a labor cost and other costs for adding a new template to the commercialized identification device. However, it is not preferable.

A second conventional technique is, for example, Japanese Patent Laid-Open No.
No. 236472. Although this method selects another optimal ring image, it has a problem in that when a new coin is issued, the ring that gives a remarkable feature value must be reselected.

As a third conventional technique, Japanese Patent Laid-Open No. 6-2 is known.
There is one described in No. 36471. Similar to the second conventional technique, this method also uses the same image data at the place selected as the detailed image, and there is a new coin or a modified / counterfeit coin with a more characteristic pattern at another place. Has a problem that its discriminating power is reduced.

[0005]

Problems to be solved by the invention are those which identify / determine the denomination, authenticity, and the like of coins by means of the images described above.
The third conventional technique can be cited as an example. As described above, in the first conventional technique, it is necessary to prepare a dedicated template for game coins or foreign coins in advance, so that the template is used every time a similar coin appears. Has to be prepared, which has the drawback of incurring personnel costs for adding a new template to the commercialized identification device, and that it is not practically applicable to new types of counterfeit coins. There's a problem. Further, in the second and third conventional techniques, the image data of the place selected as the detailed image is the same,
If there is a new similar coin or a counterfeit coin with a more characteristic pattern in another place, there is a problem that the identification result becomes extremely unreliable.

Further, recently, in addition to identifying the authenticity of coins, coins such as stains on coins, scratches on the forged surface, and missing pattern portions are excluded as spoiled coins to distinguish them from normal genuine coins without stains. There is a strong demand for separate processing. Nevertheless, no coin discriminator having such a fitness judgment function has appeared.

The present invention has been made in view of the above-mentioned circumstances, and an object of the present invention is to realize a new similar coin or a modified / counterfeit coin without using a new template or the like. It is possible to determine whether the coins are true or false, to identify similar coins effectively with less basic template data, and to determine whether the coins are right or wrong. It is to provide a loss determination method and apparatus.

[0008]

According to the present invention, an image of a target coin is taken in, the obtained image signal is digitally processed, and the calculated coin image pattern is extracted on the circumference by 360 degrees at constant extraction angles. The true / false judgment method for calculating the similarity between the sampled data and the basic template, which is the standard image data of the identification candidate coin, and identifying the coin and determining whether the coin is correct, Is the relative rotation of the sampling data and the basic template, the density difference between corresponding pixels is integrated to calculate the matching degree, the rotation angle position with the best matching degree is obtained, and the best rotation angle is obtained. The denomination and the front and back are determined based on the matching degree at the position, and the sampling data and the basic template are set with the rotation angle position having the best matching degree as a starting point. By the pattern of the degree of matching when a predetermined rotation angle is shifted with respect to the table, the authenticity is determined by checking the confirmed denomination and the type of the front and back, and the degree of matching is the best rotation angle position as a starting point. A density histogram is created by summing the pixel values for each area within a predetermined angular pitch of the ring data, and the fitness of the target coin is determined by comparing the density histogram with the corresponding density histogram of the basic template. To be achieved.

The present invention also relates to an apparatus for determining whether a coin is true or false based on an image. The above object of the present invention is to provide an image pickup means for picking up an object coin and a digital signal for digitally processing an image signal from the image pickup means. Outer diameter measurement center position calculating means for calculating the outer diameter value and center position of the image, and the sampling data by extracting the pattern of the coin image centered on the center position for 360 degrees at a constant extraction angle on the circumference. A ring cutting means for cutting out as, an edge emphasizing means for emphasizing a portion where the density changes drastically with respect to the sampling data, and the sampling data so as to match with a basic template which is standard image data of the denomination candidate. Data compression means for compressing data, density conversion means for converting density values of the sampling data into standard density values of the denomination, and edge enhancement Stage, the sampling data processed by the data compression unit and the density conversion unit and the basic template are relatively rotated, and the matching level is calculated by integrating the density differences between the corresponding pixels, and the matching level is calculated. Is a rotation matching unit that determines the best rotation angle position and determines the denomination and the front and back based on the matching degree at the best rotation angle position; and the matching degree is processed with the best rotation position as the starting point. By a pattern of the degree of matching when the sampling data and the basic template are deviated by a predetermined rotation angle, while confirming the confirmed denomination and the type of front and back, and authenticity determining means for determining the authenticity of the target coin. , Starting from the rotation angle position with the best matching degree, the sum of pixel values is calculated for each region within a predetermined angular pitch of ring data. What creates a density histogram, said is achieved by providing a fitness determining means for determining a positive loss of the subject coin by comparing the corresponding density histogram of the density histogram and the base template.

[0010]

BEST MODE FOR CARRYING OUT THE INVENTION According to the present invention, a similarity is calculated by taking in an image of the entire surface of a coin and rotationally collating an image cut out in a ring shape (ring-shaped sampling data) and a basic template (standard image data). In addition to determining the type and authenticity of coins, it is possible to determine genuine coins and non-coins by matching density histograms. In the present invention, a 360-degree image cut out in a ring shape is divided into a predetermined number (for each rotation angle position), and the most similar portion (angle position).
Based on, the degree of matching with the basic template is calculated when the image is shifted by a predetermined angle, and a graph of the degree of matching with respect to the rotation angle position is created. Whether or not the pattern of this graph matches the regular one Check the denomination, front and back of the identification candidate, and determine whether it is true or false.
A true coin and a loss coin are efficiently determined by pattern matching of partial density histograms. The graphed data becomes waveform data specific to that coin,
If the characteristic amount is extracted from the waveform data of a regular coin and registered, it is possible to perform various checks including true / false determination.

Therefore, similar foreign currency or altered coin can be eliminated without using a dedicated template, and new data can be registered as long as the pattern of the graphed data does not match even if a new foreign currency or altered coin appears. Can be eliminated without doing. Further, by judging whether the coin is correct or not, it is possible to classify coins having stains on the forged surface, scratches on the forged surface, coins lacking the pattern portion, and the like as damaged coins.

Embodiments of the present invention will be described in detail below with reference to the drawings.

FIG. 1 shows an example of the structure of a coin authenticity determination device to which the present invention is applied. An image pickup means 21 such as a CCD camera or a line sensor picks up an image of the entire surface of a coin and displays it on the xy coordinates. The two-dimensional image data is fetched in, and the image data is digitized and stored in the frame memory 31b. The outer diameter measurement center position calculating means 22 calculates the outer diameter of the coin and calculates the center position, and creates a density histogram of xy coordinates on the hist memory 31a from the image data taken in the frame memory 31b. The outer diameter value and the center coordinates of the coin are calculated based on the density histogram, and the density data of each pixel is binarized to calculate the hole diameter value (0 for no hole).

The ring cutting means 23 is a frame memory 3
The image data of 1b is cut out in a ring shape with the center coordinates from the outer diameter measurement center position calculating means 22 as the center, and the work memory 31c is used as two-dimensional array data in the x direction (rotation angle θ direction) and the y direction (radius R direction). Remember. The ring cutout unit 23 also cuts out a ring image for coin damage determination and stores it in the work memory 31c. The ring image for judging the fitness includes a place where the coin is easily damaged as a characteristic portion. The edge emphasizing means 24 performs differentiation processing on the image data cut out by the ring cutting means 23 to emphasize the edge portion where the density change is abrupt, and the data compressing means 25 the image data edge-emphasized by the edge emphasizing means 24. Is compressed to have the same size as the basic template (standard image data) of the denomination candidate. The basic template is stored in advance in the basic template storage unit 31e for each denomination, and the data compression rate by the data compression unit 25 is determined according to the denomination.

Further, the density converting means 26 is an image pickup means 21.
In order to reduce the effects of variations in lighting conditions and changes in lighting conditions so that data can be captured under the same conditions, the density values of each pixel are converted to standard density values, and each pixel data after compression is converted using the density conversion template. The density value of is converted into the density value of the basic template in correspondence with the density order. Rotation verification means 27
Calculates the degree of similarity with the standard image by performing rotational matching between the density-converted image data (ring-shaped sampling data) from the working memory 31c and the basic template from the basic template storage unit 31e, and calculates the degree of similarity with the standard image. A pattern matching value at each rotation angle is calculated for both the front and back of each identification candidate, and the rotation matching result is stored in the storage memory 31.
Store in f.

The external denomination determining means 28 presets and registers reference data of coins to be subjected to denomination identification, authenticity determination, etc. for each denomination and each front and back sides, and determines the outer diameter size, presence / absence of holes, The denomination concerned is decided from the material etc. and registered. In this embodiment, a density conversion template (standard density data) and a matching basic template (standard image data) are set for each denomination and for each front and back side, and are stored in the standard density data storage means 31d and the basic template storage means 31e, respectively. Remember. Further, the basic template storage means 31e also stores a basic template for determining whether a coin is in a normal state or not. Further, the external denomination determining unit 28 considers an external coin (for example, a foreign coin or a game coin) similar to the target coin as the determination value (the characteristic amount of the target coin) used in the authenticity determination based on the rotation pattern waveform. Then, it is determined for each denomination, for each front and back, and stored in the basic template storage means 31e. The denomination determining means 29 determines the denomination of the target coin based on the detection data of the magnetic sensor for detecting the material and the image data,
In the present invention, the denomination and the front and back are determined based on the matching degree at the best rotation angle position based on the rotation matching result obtained by the rotation matching means 27, and then the trajectory drawn during the rotation matching (the matching degree is plotted. , "Rotation pattern waveform"
Call) and determine the authenticity of the coin according to the trajectory.

On the other hand, the mask section 40 has ring data from the working memory 31c and basic template storage means 31e.
The masks of the fitness judgment basic template stored in the above are masked at corresponding positions, and each masked fitness judgment data is sent to the fitness judgment unit 41. The fitness judgment unit 41 generates a density histogram of the unmasked area of the input ring data by taking the sum of the pixel values for each predetermined section, and the basic template for fitness judgment is not masked. The corresponding density histogram of the part is read from the predetermined part. The corresponding density histogram corresponds to the density histogram of the unmasked area of the ring data, and may be generated from the fitness judgment basic template each time.
The fitness determination unit 41 determines the fitness of the coin by comparing the density histogram with the corresponding density histogram.

An example of the processing in the above configuration will be described in detail with reference to the flowchart of FIG.

First, the coins that have been conveyed pass through the image sensor section of the image pickup means 21 with the front side or the back side facing upward, and x
The image 1 of the coin is captured on the y-coordinate, for example, with 256 × 240 pixels as one screen (step S1). This image 1 is a high-density image in which one pixel is about 0.13 mm square, and the brightness is divided into 256 gradations by A / D conversion to be multi-valued, and the brightness (density or brightness) is Z-axis. Is stored in the frame memory 31b as three-dimensional data (step S1).

The outer diameter measurement center position calculation means 22 searches the image data stored in the frame memory 31b for each coordinate value at which the coin image detection boundaries are maximum and minimum in the xy directions, and the average value thereof is calculated. Center coordinates of coin image (C
x, Cy), and the difference between the maximum coordinate value and the minimum coordinate value is calculated as the outer diameter value D of the coin image. At that time, in order to avoid the influence of background noise (reflection of adjacent coins or a conveyor belt), the center coordinates (Cx, Cy) of the coins and the outer diameter value D are obtained by the following method.

In the image 1 shown in FIG. 3A, when the direction perpendicular to the coin transport direction A is the horizontal direction x, the region of the transport guide and the region of the transport belt are defined with respect to the horizontal direction x. setting the excluded section _V L -V _R as a search range. Further, in the vertical direction y, the regions at both ends (for example, 1/3 of the radius of the smallest coin are prevented so that the influence of the contact edge of the coins successively conveyed does not occur.
Setting the excluded period _H T -H _B of approximately ~ 1/4) as the search range. Then, as shown in FIGS. 3B and 3C, image density data in each search range is used to create vertical and horizontal density histograms and store them in the hist memory 31a. Here, the density histogram Vhist in the vertical direction refers to all horizontal lines.
1 determined as a horizontal line concentration sum of each pixel value in the interval V _L -V _R in the horizontal line (concentration) is obtained by arranging the respective horizontal line concentration determined in numerical order of the horizontal lines.
The horizontal density histogram Hhist is a vertical line density obtained by calculating the sum of pixel values in the section H _T −H _B in one vertical line for all vertical lines, and calculating the calculated vertical line density. They are arranged in the order of the vertical line numbers.

The outer diameter measurement center position calculating means 22 searches the edge portions of the coin image 2 from the vicinity of the center of the horizontal density histogram Hhist and the vertical density histogram Vhist toward both sides, and detects the detected edge portions. Point (midpoint of LEFT and RIGHT in Fig. 3 (C) and TO in Fig. 3 (B)
The coordinates of the middle point of P and TAIL are the center coordinates of the coin image 2 (C
x, Cy). The outer diameter value D of the coin is the outer diameter value in the horizontal direction (see FIG. 3) so as not to be affected by the reflection of the belt.
The difference between the coordinate values of LEFT and RIGHT in (C) is used.

The hole diameter value of coins is obtained as follows. The outer diameter measurement center position calculation means 22 first binarizes all the pixels of the image 1 by comparing them with a threshold value Sk shown in the following equation (1).

Sk = I _MAX V _MIN + Offset (1) where I _MAX V _MIN is the maximum density (line within one line) having the minimum horizontal line density (the line having the darkest part in the background part). The brightest pixel value),
Offset is an adjustment value.

For example, if the density of the pixel is less than or equal to the threshold value Sk, it is "1" (hole pixel), and if the density exceeds the threshold value, it is "0" (non-hole pixel), and all the pixels of the image 1 are binarized.
Next, in the image 1 after binarization, as shown in FIG. 4, with the center coordinates (Cx, Cy) of the coin image 2 as the center and the radius Ir (Ir
= Maximum hole diameter of target coin + α) is set as a hole pixel evaluation area, the number of pixels having the value “1” existing in the evaluation area is counted, and this count value is set as the hole diameter value Ir ( Step S2).

The outer diameter value D and the hole diameter value thus obtained
Specify the denomination based on the IR and the detection data of the magnetic sensor to identify the denomination, and select the specific denomination from the prepared basic templates (standard image data) for each denomination. Select as a basic template for matching. Although the denomination candidate as the identification result is determined at this stage, only the denomination has been specified and the front and back of the coin are not known at this stage, so the basic template for the denomination is the front side. Two basic templates on the back side are selected (step S2).

Next, the pattern of the coin image 2 is extracted on the circumference by 360 degrees at a constant extraction angle, and is composed of array data which is developed with the radius R direction as the Wy direction and the rotation angle θ direction as the Wx direction. Obtain three-dimensional striped sampling data. The ring cutting means 23 is a cutting parameter table TB.
As shown in FIG. 5, using 1, the center coordinates = (Cx, Cy), the base point for cutting (the starting position in the radial direction from the center coordinates).
= R, cutting width (radial cutting range) = Wy, cutting length (circumferential cutting range) = Wx,
The pattern of the coin image 2 is 3 on the circumference for every certain extraction angle.
Sixty degrees are extracted and cut out as two-dimensional striped sampling data, and as shown in FIG. 6, two-dimensional array data PDAT with the rotation angle θ direction as the X axis and the radius R direction as the Y axis.
Expand to A.

For example, when a coin of 500 yen is specified from the diameter D, the denomination 500 is selected from the cut-out parameter table TB1.
The ring-shaped image data SDATA is sampled in a stripe shape and cut out with the cutout starting point r = 65, the cutout width Wy = 20, and the cutout length Wx = 512 for the circle, and this is shown in FIG. 6 in the r (Wy) direction, θ. Set as two-dimensional array data PDATA in the (Wx) direction. The expansion from the ring-shaped image data SDATA to the two-dimensional array data PDATA is performed by converting the polar coordinates into Euclidean coordinates by the following equation (2) with the center coordinates of the coin image 2 as (Cx, Cy). To do. In addition,
When cutting out a ring, a 3 × 3 pixel spatial filter is used in the edge enhancement process to perform differentiation, so the range between (r -1) and (r + Wy +1) is targeted in the radial direction. As a result, extra pixels are cut out one by one.

PDATA (i, j) = SDATA (Px + Cx, Py + Cy) (2) where Px = ri × sin θj, Py = ri × cos θj, ri = i +
(r−1), θj = j × 2π / Wx, i = 0 to (Wy−1), j = 0
(Wy−1).

In this way, in the ring cutting means 23,
The ring-shaped image data up to a specific rotation angle position including 360 degrees is cut out into stripes and appropriately extracted as two-dimensional array data. In the example of the 500-yen coin, the cut-out image includes 22 pixels in the R direction and 512 pixels in the θ direction (step S3).

The edge emphasizing means 24 performs x-direction / y-direction differentiation in the vicinity region of 3 × 3 pixels in order to emphasize the feature of the pattern on the cut-out image. As a processing method of this edge enhancement, for example, there is a known technique "Prewit
As shown in FIGS. 7A and 7B, 3 × 3 spatial filtering by “t operator” is used, the pixel data before conversion is PDATA (i, j), and the pixel data after conversion is GDATA.
As (m, n), the edge portion is emphasized by correcting the density value according to the following equations (3) to (5). At that time, x
As the filter and the y filter, for example, a 3 × 3 filter is used.

ΔxDATA (i, j) = PDATA (i-1, j-1) + PDATA (i-1, j) + PDATA (i-1, j + 1)-(PDATA (i + 1, j- 1) + PDATA (i + 1, j) + PDATA (i + 1, j + 1)} (3) ΔyDATA (i, j) = PDATA (i-1, j-1) + PDATA (i, j -1) + PDATA (i + 1, j-1)-{PDATA (i-1, j + 1) + PDATA (i, j + 1) + PDATA (i + 1, j + 1)} (4 ) GDATA (m, n) = | ΔxDATA (i, j) | + | ΔyDATA (i, j) | ... (5) Note that the differential sum is calculated for each pixel in the 0th row and the (Wy + 1) th row. Since it cannot be calculated, the resulting GDATA (m,
The matrix of n) has Wy rows and Wx columns as shown in FIG. By the edge enhancement processing by the edge enhancement means 24, an image in which a portion where the density change is abruptly raised is obtained (step S4).

Next, the edge-enhanced data WDATA in the Wy row and Wx column is data-compressed into a matrix of the same size as the basic template for collation. This compression rate differs depending on the denomination, and in the case of a 500-yen coin, a compression rate of 1/8 is preferably applied. For example, an average density value is calculated for each data of 4 pixels in the x-axis direction and 2 pixels in the y-axis direction. Then, the data of 8 pixels is compressed into the data of 1 pixel. As a result, Wy = 20
The data GDATA in the row and Wx = 512 columns is compressed into the data CDATA in 10 rows and 128 columns (step S5).

Next, in order to reduce the influence of variations in the image pickup means 21 and the like, a density conversion process is performed on the compressed data CDATA using a density conversion template in which standard density data is set. 8A shows an example of the density distribution of the target image, and FIG. 8B shows an example of the density distribution of the basic template.

The density conversion processing will be described with reference to FIG. First, the compressed data (a) is rearranged in order of density, and a density order table (b) and a density order index table (c) are created. The element of the density order index table (c) is the position (index) in the table (a) of the corresponding element of the density order table (b). Next, the element of the density order table (d) corresponding to each element of the density order table (b) is referred to, and the value is indicated by the density order index table (c), that is, the compressed data table (a). The table (e) is prepared by arranging the table (e) at the position inside. As a result, the compressed data (a) is converted into the density value (higher order) of the basic template. FIG. 10 shows a concrete example corresponding to FIG. 9, and the compressed data (a) which is the data before conversion is the data (d) as a result.
Is converted to. If the density values are the same, the value corresponding to the density of the template is set (step S6).

In this way, the above steps S3 to S
For the two-dimensional striped sampling data obtained by performing the process of No. 6, the collation operation is performed with the collation basic template which is the standard image data of the identification candidate coin.
FIG. 11 shows the sampling data IDATA and the basic template MDATA after the processing for the raw image of the 500-yen coin. As shown in FIGS. 12 (A) and 12 (B), the calculation method related to collation is the sampling data IDAT after processing.
Corresponding pixel fn in A and the basic template MDATA
And the density difference between the pixels tn are calculated over the entire area, and the sum total value Pm of the absolute values of the calculated density differences is calculated by the following equation (6) to be used as the pattern matching value at the rotation angle position.

Pm = Σ | fn-tn | (6) where n = 1 to the total number of pixels (= the number of array data after compression).

Then, while the sampling data IDATA and the basic template MDATA are relatively rotated by one round, the pattern matching value Pm is obtained for each rotation angle.
The sampled data after processing in this example is array data IDATA having 10 rows in the x direction (= R direction) and 128 columns in the y direction (= θ direction), and the rotation pitch is 360/12.
Calculation of the pattern matching value Pm while rotating one image in the θ direction at 8 degrees, that is, while shifting the array data IDATA in the x direction (or the basic template MDATA in the −x direction) by one pixel (one column) at a time. Is repeated a number of times corresponding to one rotation (128 times in this example), and the pattern matching value Pm for each rotation angle position is stored in the storage memory 31f (step S7).

The pattern matching value Pm is the minimum value min (Pm)
Is the rotation angle position with the best degree of matching with the basic template image. The rotation matching unit 27 calculates the best rotation angle position for both the front and back of each denomination identification candidate of the denomination, and stores it in the storage memory 31f together with the pattern matching value PM for each rotation angle position (step S8).

The denomination determining means 29 determines that the minimum value min (Pm) of the pattern matching values has the highest degree of similarity with the basic template (hereinafter, "best matching value PM-Best").
The best matching value PM-B
Based on est, determine the denomination and the front and back of the target coin. In this example, the front / back of the denomination of the target coin is the front / back of the denomination of the basic template image corresponding to the smallest of the best matching values PM-Best obtained for each front and back of each identification candidate. It is confirmed (step S9).

FIG. 13 shows an example in which there are four identification candidates, and the processing of steps S6 to S9 when there are a plurality of identification candidates will be described with reference to FIG. The compressed sampling data CDATA is converted into the density value of the template using the density conversion template of each identification candidate. The rotation collating means 27 includes density conversion data IDATA1 to 4 and a basic template MDATA1 on the front and back.
4 to 4 are performed to obtain the best matching values PM-Best1 to PM-Best4 for each identification candidate and for each front and back. Then, each best matching value PM-Best1 to PM-Best4 is compared with the reference value Sb, and those having the best matching value equal to or less than the reference value Sb are determined as the same pattern, and the denomination and the front and back are determined based on the same pattern.

When the best matching value PM-Best exceeds the reference value Sb, the coin is determined as a reject coin. Originally, if the rotation angle position of the coin image and the rotation angle position of the rotation calculation match, the best matching value PM-Best
Should be 0, but it does not always become 0 in the case of matching matching with dirty coins or a coarse pitch, so it is set to a reference value Sb with an allowable range set in advance.

The denomination and the front / back determined as described above are determined based on the total value of the density differences from the basic template image. In the present invention, as the processing after step S10, the rotation angle position of the best matching value PM-Best (= shift position of the array data in the x direction) is further used as the starting point, and the calculation result of the equation (6) at the time of rotation matching is calculated. Using the pattern matching value Pm for each rotation angle position, the pattern of matching degree when the ring image (sampling data) and the basic template (standard image of a regular coin) are deviated by a predetermined rotation angle (hereinafter, "rotation Denomination), the denomination determining means 29 confirms the determined denomination and the legitimacy of the front and back, and determines the authenticity of the coin.

Hereinafter, the determination process of denomination, authenticity, etc. based on the rotation pattern after step S10 will be described in detail with a specific example.

An example of the case of a 500-yen coin will be shown. As a result of the above-mentioned rotational collation, 128 matching values Pm are obtained.
Are stored in the storage memory 31f. Of these 128 matching values Pm, the matching value Pm at the rotation angle position with the highest degree of similarity is the best matching value PM-Best, and the matching value Pm at each rotation angle position after the best matching value PM-Best is , The rotation pitch is 360 degrees / 128 = about 2.8 degrees, and is a value when one image is shifted by one pitch in the θ direction. For example, if rotational matching is performed using the same template as the target coin, the matching value at a certain rotation angle position that is offset from the best rotation angle position by a predetermined rotation angle becomes data specific to that coin, and that data is converted into a waveform. By doing so, the characteristics of the coin can be extracted. By utilizing this, by matching the waveform formed by each matching value Pm when rotating normal coins and the waveform formed by the target coin and the normal coin, even if the similar coin or the modified coin is high. It becomes possible to determine with accuracy. At that time, instead of checking all the waveforms formed by the target coin and the normal coin, for example, the distance from the best matching value PM-Best to the specific angle is calculated, and whether the distance is within the range of the normal coin or not By checking whether or not it is possible to make a true / false decision, it is possible to effectively distinguish similar coins with a small amount of data.

FIG. 14 is a graph showing a waveform (a locus of the matching value Pm at each rotation angle position) on one surface of a 500-yen coin with the matching value Pm on the vertical axis and the rotation angle position Rn on the horizontal axis. . In this example, the rotation pitch is 360 degrees / 128 =
It is a graph showing the locus formed by the matching value Pm when the basic template images on the backs of the 500 yen coins are rotated 128 times (1 rotation) relative to about 2.8 degrees. The matching value Pm at the angular position = 0) is the best matching value PM-Best. As for the matching value on the vertical axis, the allowable range is + α (in this example α
≅30000). Hereinafter, a locus drawn by coins of the denomination at the time of rotational matching is referred to as a “matching locus” or a “rotation pattern waveform”.

In this way, if the characteristic amount (judgment value) at the specific rotation angle position is checked (for example, the distance from the best matching value is checked) by the locus drawn by the self denomination at the time of rotational matching, the authenticity of the coin is judged. It can be performed.
The feature amount is the feature amount of the distributed coins to be handled, and it is not necessary to register the feature data of similar coins or the like. For example, taking a specific surface of a 500-yen coin as an example, FIG.
A plurality of specific angles are provided as check points as shown in to, and as a judgment value at each check point, for example, the distance from the best matching value PM-Best to (matching value Pm + allowable value α) at the check point. Pl1 to PL7 are set, and true / false determination and final denomination and front / back determination are performed based on these determination values PL1 to PL7.

An example of the back side of a 500 yen coin is as follows:
With the number of rotations indicating the position of the specific angle as the Zone value, the judgment value (distance PL) of each Zone value is shown in table TB-B500 as shown in FIG.
Set to. Then, in the denomination determination means 29, the Zone
Value = XX, the matching value between the sampling data and the basic template is ZoneXX, and the best matching value is PM-B.
When est satisfies all of the following judgment formulas (7) to (13) corresponding to the check points of to in FIG. 14, it is judged that the coin is a normal coin.

Zone10-PM-Best> 42000 ... (7) Zone32-PM-Best> 6300 ... (8) Zone56-PM-Best> 41000 ... (9) Zone64-PM-Best> 1900 ... (10) Zone73-PM -Best> 41000 (11) Zone97-PM-Best> 5000 (12) Zone120-PM-Best> 41000 (13) The table shown in FIG. 15 is prepared for each front and back of the denomination,
Fig. 16 (A) is a matching locus on the table of 10-yen coins, Fig. 1
6 (B) shows a matching locus on the back of a 10-yen coin. In the case of a 10-yen coin as an example, the judgment formulas in the table of 10-yen coins corresponding to the check points of-in Fig. 16 (A) are the following formulas (14)-(20), and in Fig. 16 (B). The following equations (21) to (27) are the determination equations on the back of the 10-yen coins corresponding to the check points of No.

Zone9-PM-Best> 9000 (14) Zone16-PM-Best> 12,500 (15) Zone28-PM-Best> 15000 (16) Zone64-PM-Best> 9000 (17) Zone83-PM -Best> 12000 (18) Zone101-PM-Best> 15500 (19) (Zone120-PM-Best) + (Zone9-PM-Best)> 2000 ... (20) Zone4-PM-Best> 3000 ... (21) ) Zone19-PM-Best> 11000 ... (22) Zone35-PM-Best> 5000 ... (23) Zone64-PM-Best> 9000 ... (24) Zone95-PM-Best> 5000 ... (25) Zone105-PM-Best > 9000 (26) (Zone124-PM-Best) + (Zone4-PM-Best)> 5000 ... (27) As described above, since similar coins of different countries have different rotation pattern waveforms, each predetermined angular position Each map obtained when rotated By performing pattern matching of the pattern of ring value it can be carried out without specially prepared conventional such templates, a determination of the authenticity, including checking of falsification / counterfeit coin with high accuracy.

The denomination determining means 29 obtains the pattern matching value Pm and the best matching value PM-Best for each rotation angle position in step S9, and after determining the denomination and the front / back of the target coin, Process.

First, with the best matching value PM-Best,
Calculate the phase shift between the captured image and the basic template image. In this example, after the captured image is made into stripe-shaped sampling data, it is used as array data of the pattern matching value Pm for each rotation angle position, and by using the best matching value PM-Best as the base point in this array data to be judged. , The data sequence of the regular array data of the denomination determined in step S9 is matched (step S
10).

Subsequently, the locus of the rotation pattern is extracted with the position of the best matching value PM-Best as the base point. The regular array data is the array data of the pattern matching value Pm obtained by rotating the basic template of the regular coin, and in this example, the determination table for extracting the feature amount as shown in FIG. 15 corresponds to the array data. . In this case, the trajectory of the rotation pattern is extracted by the pattern matching value Pm corresponding to each Zone value (rotation angle position of the determination point) shown in the determination table.
Is extracted from the sequence data to be determined (step S11).

Next, the distance from the best matching value PM-Best to the specific angle position is calculated. For example, in the example of FIG. 14, the distances PL 1 to P 1 from the best matching value PM-Best to the (matching value Pm + allowable value α) at the check location.
L7 is calculated (step S12). Then, by comparing the calculated distance to the specific angular position with a regular distance (the judgment value shown in the judgment table),
Check the front and back and determine the authenticity. In the process of step S9, if there are a plurality of fixed denominations and front and back candidates that cannot be identified, then if the coin is a regular coin, the denomination and front and back are identified at this point ( Step S13).

After that, in the present invention, another ring for judging the fitness is read from the working memory 31c (step S14), the corresponding basic template for judging the fitness is read from the basic template storing means 31e, and the mask unit 40 is used to display it in FIG. As shown, mask processing of the corresponding place and size is performed (step S15). FIG. 17A shows a masked image of a ring image, and FIG. 17B shows a masked image of a basic template for fitness judgment, both of which are areas where a white portion is masked. After such masking processing, the fitness judgment unit 41 executes the fitness judgment processing shown in FIG. 18. That is, first, the density histograms Hal, Hb of the unmasked regions al, bl, cl of the ring image are obtained.
l and Hcl are generated by calculating the sum of pixel values for each predetermined small section (steps S21 to S23). Since the denomination and the front and back have already been determined, the basic template for determining whether the denomination and the front and the back are correct can be read from the basic template storage means 31e, and the area A corresponding thereto can be read.
l, Bl, Cl corresponding concentration histograms HAl, HB
l and HCl also become known values and a predetermined memory (not shown)
It is stored in. Therefore, it is determined whether or not the difference between the density histogram Hal of the area al and the corresponding density histogram HAl of the area Al is smaller than the predetermined value α (step S2
4) If the difference is equal to or greater than the predetermined value α, it is determined as a loss currency. If the difference is smaller than the predetermined value α, it is further determined whether or not the difference between the density histogram Hbl of the area bl and the corresponding density histogram HB1 of the area Bl is smaller than the predetermined value β (step S25). If is equal to or more than the predetermined value β, it is determined as a loss currency. When the difference is smaller than the predetermined value β, the density histogram Hcl of the area cl and the corresponding density histogram HC of the area Cl are further calculated.
It is determined whether or not the difference from l is smaller than a predetermined value γ (step S26), and if the difference is equal to or larger than the predetermined value γ, it is determined as a loss currency.

The difference in this case is calculated for each predetermined small section in which each frequency in the section where the density histogram is generated is calculated, and | Ha1 _i −HA1 _i | <predetermined value β _i (where i is the region number ) May be determined, or the density values (frequency) for several sections may be added or averaged. If the area not to be masked is small, the entire area (in the above example, areas al, bl, cl
It is also possible to calculate the sum or average value of the pixel values for each region (1) and compare it with the corresponding predetermined value. Alternatively, if the number of predetermined small section locations deviating from the predetermined value is equal to or more than a predetermined number, then the money may be treated as a loss money.

In this way, by comparing the density histograms, it is possible to determine whether the coin is right or wrong. The mask processing for the ring image can be efficiently determined by leaving a place where the coin is easily damaged. Further, the shape of the region not to be masked may be ring-shaped or rectangular, and it is preferable that a region including a feature that is easily obtained as a loss coin is included. This method can also be used for detailed authenticity determination of the true coin by using a threshold different from that of the loss coin determination.

In the above-described embodiment, each processing procedure is described in time series using a flowchart, but the order of processing is not limited to the order of the flowchart. Further, although the Japanese distributed coins have been described as the identification target, the same applies to the foreign distributed coins as the identification target coins. Regarding similar coins and counterfeit coins, the case where the feature amount (judgment value) of the rotation pattern is not prepared has been described as an example.
For example, if the feature amount of the found rotating pattern of the counterfeit coins is set, it is possible to notify the criminal activity by an alarm or the like when the counterfeit coins are recognized. Further, the mode in which one surface is imaged and processed by the imaging means has been described as an example, but a mode in which both front and back surfaces of a coin are imaged and processed may be used. In that case, for example, in a transport path formed of a transparent member. Image pickup parts of the imaging hand throw are installed opposite to each other above and below the surface, and processing such as data processing is performed in parallel based on the input image data from each imaging hand throw, and both the front and back sides are inspected to determine whether the image is true or false. It becomes the form which judges such as.

[0059]

According to the present invention, after the candidate denomination is found by the value having the best matching degree found by the rotational matching with the basic template, the total sum of the pixel values of the predetermined area for each predetermined rotation angle position is found. Since the density histogram obtained by plotting is obtained and the pattern matching of the density histogram is performed within a predetermined range, it is possible to discriminate between a damaged coin and a damaged coin. Furthermore, for new similar coins or new types of forged / counterfeit coins, the authenticity of the coins can be determined as long as the patterns of the density histogram do not match.

Since the present invention also determines whether coins are right or wrong, it is possible to exclude coins that are damaged by use and coins that are partially soiled as damaged coins. In the judgment of fitness, the normal coin is taken out by masking only the position where the probability of being a bad coin is high, and the fitness is judged by the difference in the density histogram of the obtained region. Since the denomination, the front and back, and the characteristic area of the coin are limited, there is a feature that the coin can be accurately determined.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration example of a coin authenticity / damage determination device according to the present invention.

FIG. 2 is a flowchart showing a processing example of the present invention.

FIG. 3 is a diagram for explaining a method of calculating an outer diameter value and a center position of a coin.

FIG. 4 is a diagram for explaining a method of calculating a hole diameter value of a coin.

FIG. 5 is a diagram showing an example of a cutout parameter table of a ring image.

FIG. 6 is a diagram showing a data array after cutting out a ring image.

FIG. 7 is a diagram for explaining edge enhancement processing of a characteristic portion of a coin image.

FIG. 8 is a diagram for explaining a density conversion process to a standard density.

FIG. 9 is a diagram for explaining a density conversion process to a standard density.

FIG. 10 is a diagram for explaining a density conversion process to a standard density.

FIG. 11 is a diagram showing an example of a processed coin image and a basic template.

FIG. 12 is a diagram for explaining a rotational matching process with a basic template.

FIG. 13 is a diagram illustrating a method of determining a denomination and front and back when there are a plurality of identification candidates.

FIG. 14 is a diagram for explaining a method of determining a denomination and authenticity by inspecting a rotation pattern waveform.

FIG. 15 is a diagram showing an example of a table used for inspection of a rotation pattern waveform.

FIG. 16 is a diagram showing another example of a waveform formed by each matching value when circulating coins of the same denomination are rotated.

FIG. 17 is a diagram for explaining mask processing.

FIG. 18 is a flowchart showing a processing example of a fitness judgment.

[Explanation of symbols]

Image of 1 coin 21 Imaging means 22 Outline measurement center position calculation means 23 Ring cutting means 24 Edge enhancement means 25 Data compression means 26 Concentration conversion means 27 Rotation verification means 28 External denomination determination means 29 Kind determination means 31a Hist memory 31b frame memory 31c working memory 31d Standard concentration data storage means 31e Basic template storage means 31f storage memory 40 Mask 41 Judgment section

Continued front page    F-term (reference) 3E002 AA06 AA17 BD01 CA02 CA06                       CA15 DA04 EA05                 5B057 BA02 CA08 CA12 CA16 CD03                       CE03 CE09 DA12 DB02 DB09                       DC05 DC16 DC23 DC32                 5L096 AA06 BA03 BA18 EA16 EA17                       EA35 FA06 FA37 FA62 JA03                       JA09

Claims (5)

[Claims] 1. Sampling data obtained by capturing an image of a target coin, digitally processing the obtained image signal, and extracting a calculated coin image pattern for 360 degrees at a constant extraction angle on the circumference, and an identification candidate. In a true / false / incorrectness determination method of calculating similarity between a basic template which is standard image data of coins and identifying the coin and determining whether the coin is correct, the sampling data and the basic template are relatively Rotate, calculate the matching degree by integrating the density difference between the corresponding pixels, find the rotation angle position with the best matching degree, and determine the denomination and the front and back based on the matching degree at the best rotation angle position. , The matching degree when the sampling data and the basic template deviate from each other by a predetermined rotation angle with the rotation angle position having the best matching degree as a starting point. Degree pattern, the authenticity is determined by confirming the determined denomination and the type of the front and back sides, and the matching angle is set as the starting point at the rotation angle position, and the pixel is determined for each area within a predetermined angular pitch of the ring data. By creating a density histogram by summing the values and comparing the density histogram with the corresponding density histogram of the basic template, it is determined whether the target coin is right or wrong. Coin authenticity / damage determination method. 2. Sampling data used for obtaining the rotational angle position of the coin and ring data for generating the density histogram are different from each other, and the corresponding density histogram is preset according to denomination and front / back. The coin authenticity / damage determination method using the image according to claim 1. 3. The image according to claim 1, wherein a plurality of the ring data are used according to denominations and front and back sides, and a plurality of the corresponding density histograms are set corresponding to the plurality of ring data. How to judge the authenticity / incorrectness of coins. 4. The coin authenticity / damage determination method according to claim 1, wherein only characteristic portions that are prone to coin damage are used in the generation of the density histogram. 5. An image pickup means for picking up an object coin, an outer diameter measurement center position calculation means for digitally processing an image signal from the image pickup means to calculate an outer diameter value and a center position of a coin image, and the center position. A ring cutting means for extracting the pattern of the coin image on the circumference for 360 degrees at a constant extraction angle and cutting it out as sampling data, and a portion where the density change is abruptly emphasized with respect to the sampling data. An edge enhancing means for processing, a data compressing means for compressing the sampling data so as to match the basic template which is the standard image data of the denomination candidate, and a density value of the sampling data to a standard density value of the denomination. Density converting means for converting, the edge enhancing means, the data compressing means, and the sampling processed by the density converting means And the basic template are rotated relative to each other, the matching degree is calculated by integrating the density differences between the corresponding pixels, the rotation angle position with the best matching degree is obtained, and the matching angle at the best rotation angle position is calculated. Rotation collating means for determining the denomination and the front and back based on the matching degree, and when the processed sampling data and the basic template deviate by a predetermined rotation angle from the rotation angle position where the matching degree is the best as a starting point. A matching degree pattern confirms the determined denomination and the front and back types, and an authenticity determination unit that determines the authenticity of the target coin, and the matching degree is the best rotation angle position as a starting point, and a ring. A density histogram is created by summing pixel values for each area within a predetermined angular pitch of data, and the density histogram and the basic template are created. Corresponding density histogram and the target coin coin authenticity / soil detection device according to the features and image by comprising a determining fitness determining means positive losses by comparing the.