JP2012084175A - Coin classification device and coin classification method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve coin image collation accuracy and efficiently collate detailed features of coins.SOLUTION: A denomination discrimination processing part acquires the denomination of a coin and a first authenticity identification processing part which receives the acquired denomination executes coin authenticity identification processing. An era identification processing part identifies an issuance era of the coin, a second authenticity identification processing part acquires a feature part template image and a feature range, by retrieving a feature table using the identified denomination and issuance era, and collates the feature part template image with the corresponding part of the input image. An authenticity determination part determines the authenticity of the coin based on processing results of the first authenticity identification processing, the era identification processing and the second authenticity identification processing.

Description

  The present invention relates to a coin classification device and a coin classification method for efficiently classifying coins while taking into account the age at which the coins were issued.

  2. Description of the Related Art Conventionally, there has been known a currency identification device that compares an input image obtained by photographing a deposited money with a CCD (Charge Coupled Device) camera, an image sensor, or the like and a template image registered in advance to determine the authenticity of the money. It has been.

  For example, in Patent Document 1, a correlation value is calculated by comparing an input image of coins with a template image of coins, and the correlation value exceeds a threshold in a portion of the entire image to be compared that exceeds a predetermined area. In such a case, a coin identification technique for determining a coin related to an input image as a genuine coin (hereinafter referred to as “true coin”) is disclosed.

JP 2003-187289 A

  However, when this conventional technology is used, the issue year portion of the coin varies depending on the coin casting year, so the correlation value of the issue year portion is low, and as a result, the issue year portion is usually excluded from the target for verification. It is. For this reason, it is possible to accurately identify counterfeit coins (hereinafter referred to as “fake coins”) that have imitated the design of true coins but are not engraved in issue year, and fake coins that have been engraved in a specific issue year. There was a problem that I could not.

  In recent years, as a measure against counterfeit coins, true coins often have various features. For example, attempts have been made to give detailed and fine security features such as fine dots and fine lines to the design of coins and to change such security features every few years. In the future, it is expected that the frequency of changing security features will increase in order to deal with counterfeit coins that are skillfully imitating true coins. However, if all such detailed and fine security features are to be collated, the collation process is overloaded and it is difficult to perform efficient collation.

  Such a problem is not only a problem that occurs only in the coin identification process, but also a problem that occurs in the identification process of banknotes and the identification process of security signs attached to commodities distributed in the market.

  The present invention has been made in view of the above problems (problems), and improves the image matching accuracy of coins and classifies coins with high accuracy by efficiently collating minute features of coins. An object of the present invention is to provide a coin classification device and a coin classification method.

  In order to solve the above-described problem, the present invention provides a coin classification device that performs the coin classification process by collating an input image of coins with a pre-registered character template image, and converting the input image to a polar coordinate Polar coordinate conversion image acquisition means, rotation angle acquisition means for acquiring a rotation angle by collating with a pre-registered reference image, rotation of the coin input image and the character template based on the acquired rotation angle The image rotation means for matching the corners, the partial image including the character part indicating the issue year that forms part of the input image of the coin, and the character template image are collated with the rotation angles of both images being aligned. , An issue year identifying means for identifying the issue year of the coin based on the collation value, and a classification processing means for classifying the coin based on the identification result of the issue year identifying means. Characterized in that was.

  Moreover, the present invention is characterized in that, in the above-mentioned invention, the classification processing means classifies coins of a predetermined issue year identified by the issue year identification means with coins of other issue years.

  Further, the present invention is characterized in that, in the above-mentioned invention, the classification processing means classifies the coin as a subject of rejection when the issue year of the coin is not identified by the issue year identification means.

  Further, the present invention is the above invention, wherein the classification processing means includes a feature part template storage means for storing a plurality of feature part templates respectively indicating feature parts of coins of a plurality of issue years, and a coin by the issue year identification means. When the issue year of the coin is identified, a search means for searching for a feature template of the coin corresponding to the issue year from the feature template storage means, and a feature portion of the coin that forms a part of the input image of the coin A true / false identification processing means for identifying whether the coin is a true coin or a false coin based on a collation value between the partial image indicating the feature portion template and the feature template searched by the search means It is characterized by.

Further, the present invention is characterized in that, in the above-mentioned invention, the classification processing means classifies the issued coins after a predetermined issue year and issued coins of the old year until then.
Further, the present invention is a coin classification method of a coin classification device that performs a classification process of the coins by collating an input image of coins with a pre-registered character template image, wherein the input image is subjected to polar coordinate conversion. And a rotation angle acquisition step of acquiring a rotation angle by collating with a reference image registered in advance, and an image for matching the rotation angle of the input image of the coin and the character template based on the acquired rotation angle Rotating step, collating the character template image with a partial image including a character part indicating a publication year forming a part of the input image of the coin in a state where the rotation angle is matched, and based on the collation value, An issue year identifying step for identifying the issue year of coins and a classification processing step for classifying the coins based on the identification result of the issue year identifying step are included.

  According to the present invention, the input image is subjected to polar coordinate conversion, the rotation angle is obtained by collation with a reference image registered in advance, and the coin input image and the character template rotation angle are matched based on the obtained rotation angle. The partial image including the character part indicating the year of issue that forms part of the input image of the coin and the character template image are collated in a state where the rotation angles of both images are aligned, and based on the collation value, Since the issue year is identified and the coins are classified based on the identification result, the coins can be efficiently classified in consideration of the issue year in which the coins were issued.

It is a functional block diagram which shows the structure of the money identification device which concerns on Example 1. FIG. It is a flowchart which shows the outline | summary of the process sequence of the money identification device which concerns on Example 1. FIG. It is explanatory drawing for demonstrating the outline | summary of the 1st authenticity identification process shown in FIG. It is explanatory drawing for demonstrating the outline | summary of the age identification process shown in FIG. It is explanatory drawing for demonstrating the outline | summary of the 2nd authenticity identification process shown in FIG. It is a flowchart which shows the process sequence of the 1st authenticity identification process shown in FIG. It is explanatory drawing for demonstrating the process outline | summary of the image cutting-out process shown in FIG. It is explanatory drawing for demonstrating the Sobel operator used in the edge extraction process shown in FIG. It is explanatory drawing for demonstrating the process outline | summary of the edge extraction process shown in FIG. FIG. 6 is an explanatory diagram for explaining a processing outline of polar coordinate conversion processing according to the first embodiment. It is explanatory drawing for demonstrating the process outline | summary of the rotation angle detection process shown in FIG. It is a flowchart which shows the process sequence of the collation process by the positive / negative separation correlation shown in FIG. FIG. 6 is an explanatory diagram for explaining each region image according to the first embodiment. It is a flowchart which shows the process sequence of the normalization correlation value image positive / negative separation process contained in the positive / negative separation correlation image generation process shown in FIG. It is a flowchart which shows the process sequence of the template image positive / negative separation process included in the positive / negative separation correlation image generation process shown in FIG. It is a flowchart which shows the process sequence of the area | region image generation process included in the positive / negative separated correlation image generation process shown in FIG. It is explanatory drawing for demonstrating the image generation procedure corresponding to each area | region shown in FIG. It is explanatory drawing for demonstrating the image mask used in the expansion process shown in FIG. It is a flowchart which shows the process sequence of the expansion process shown in FIG. It is explanatory drawing for demonstrating the image produced | generated by the expansion process shown in FIG. It is explanatory drawing for demonstrating the block division of the image used in the collation value calculation process shown in FIG. It is a flowchart which shows the process sequence of the age identification process shown in FIG. It is explanatory drawing for demonstrating the character collation table used in the age identification process shown in FIG. It is a flowchart which shows the process sequence of the 2nd authenticity identification process shown in FIG. FIG. 25 is an explanatory diagram for describing a feature table used in the second authenticity identification process illustrated in FIG. 24. It is a functional block diagram which shows the structure of the money identification device which concerns on Example 2. FIG. It is a functional block diagram which shows the modification of the structure of the money identification device which concerns on Example 2. FIG. It is a flowchart which shows the process sequence of the 1st authenticity identification process which concerns on the modification shown in FIG.

  Exemplary embodiments of a currency identification device, a currency identification method, and a currency identification program according to the present invention will be described below with reference to the accompanying drawings. In the following, the case where the currency to be identified is a coin will be described. Further, in the following embodiments, since the identification target is a coin, the coordinate conversion process is included, but when the identification target is a non-circular object such as a banknote, the polar coordinate conversion process is not included. You can also

  In the following description, in the case where the denomination identification process of coins is performed using a sensor such as an optical sensor or a magnetic sensor in the first embodiment, an image obtained by capturing the denomination identification process using a CCD camera or the like is used in the second embodiment. Each case will be described.

  FIG. 1 is a functional block diagram illustrating the configuration of the currency identification device 1 according to the first embodiment. As shown in FIG. 1, the money identification device 1 includes an optical sensor 2, a magnetic sensor 3, a CCD camera 4, a control unit 10, and a storage unit 20. The control unit 10 further includes a denomination identification processing unit 11, a first authenticity identification processing unit 12, an age identification processing unit 13, a second authenticity identification processing unit 14, and a true / false determination unit 15, and stores them. The unit 20 further includes a registered image 21, a rotated input image 22, a character matching table 23, and a feature table 24.

  The optical sensor 2 is a sensor for detecting the presence or absence of a coin hole, and the magnetic sensor 3 is a sensor for determining the material and diameter of the coin. These sensors (the optical sensor 2 and the magnetic sensor 3) perform processing for outputting the detected information to the denomination identification processing unit 11 of the control unit 10. The denomination identification processing unit 11 identifies the denomination of the coin based on the sensor output. Note that the diameter of the coin may be detected by the optical sensor 2.

  The CCD camera 4 is a camera for capturing an image of a coin to be collated, and a camera for capturing minute features of a coin such as a double speed VGA (Video Graphics Array) camera is used. The coin image picked up by the CCD camera 4 is transferred to the first authenticity identification processing unit 12. The CCD camera 4 handles the input image as an aggregate of a predetermined number of pixels. For example, the input image is recognized as a gray scale image having a density value of 256 gradations, and is output to the first authenticity identification processing unit 12 as a rectangular image having a predetermined size. In the first embodiment, the CCD camera 4 is used to capture a coin image. However, the present invention is not limited thereto, and a coin image may be captured using a high-resolution image sensor or the like. .

  The denomination identification processing unit 11 is a processing unit that identifies the denomination of a coin based on information output from the optical sensor 2 and the magnetic sensor 4. Specifically, the denomination identification processing unit 11 identifies the denomination of the coin based on the material and diameter detected by the magnetic sensor 3 and the presence / absence of a coin hole detected by the optical sensor 2. As described above, the diameter of the coin may be detected by the optical sensor 2. However, in this case, a method different from the above-described optical sensor that detects the presence or absence of a hole, for example, an optical sensor capable of detecting an area wider than the diameter of a coin, or a plurality of optical sensors is used. The diameter of the coin is detected by the method.

  Based on the denomination received from the denomination identification processing unit 11, the first authenticity identification processing unit 12 selects the registered image 21 stored in the storage unit 20 in advance, and selects the registered image 21 and the CCD camera 4 from the selected registration image 21. It is a processing unit that identifies the authenticity of a coin by collating it with the acquired input image. In addition, since the first true / false identification process uses an image matching process based on the positive / negative separation correlation described later, a highly reliable identification process can be performed. In the first embodiment, the first true / false discrimination process uses a collation method based on a positive / negative separation correlation described later, but other collation methods (for example, a general normalized correlation method) may be used. Good.

  The first true / false identification processing unit 12 performs true / false identification processing of a coin based on a coin pattern excluding a portion unique to each individual, such as a coin issuance year. Specifically, the template image used as the registered image 21 is an average image of the same type of coins having different issue years, thereby identifying the true / false of the coin by eliminating the influence of the portion unique to each individual. By such authenticity identification processing, it is possible to identify denominations such as 100-yen coins and 500-yen coins, and fake coins obtained by modifying foreign currencies with similar weights and metal compositions.

  The first authenticity identification processing unit 12 cuts out an image of a coin part from the input image, emphasizes the edge part by edge extraction processing, normalizes it to absorb individual differences, and then polar coordinates. Perform the conversion. Then, the rotation angle is calculated by collating the input image after the polar coordinate conversion with the registered image 21 registered through the same processing while being translated. Therefore, when the identification result is a true coin, a rotated input image 22 having the same rotation angle as that of the registered image 21 is generated. The rotated input image 22 is used in the age identification processing unit 13 and the second authenticity identification processing unit 14 described later.

  The age identification processing unit 13 is a processing unit that identifies an issue year (year) of a coin to be verified using the rotated input image 22 and the character verification table 23 of the storage unit 20. Specifically, the age identification processing unit 13 receives the denomination identified by the first authenticity identification processing unit 22 and acquires a character template image corresponding to the received denomination from the character verification table 23. To do. Then, the character portion representing the age is cut out from the rotated input image 22, and the age of the coin to be collated is identified by collating each cut out character with the character template image corresponding to the denomination. Then, the identified age is passed to the second authenticity identification processing unit 14.

  In addition, in the age identification processing unit 13, when the recognition of the characters constituting the age has failed, or when the character recognition has succeeded but it is an unfair age (for example, when the coin has not been issued) If the age identification fails as in (), the coin to be collated is determined to be “rejected”. In this way, when the age identification fails, it is determined not to be a fake coin immediately but to be “rejected” for the following reason.

  In other words, among coins that failed to identify the age, the font shape of the characters changed due to the real coins (dirty coins) where some characters could not be recognized due to dirt, the true coins that were not stamped with the age itself, and wear. This is because it is improper to immediately make a fake coin because it contains a true coin. In addition, it is good also as handling as a fake coin, when possibility that it is a fake coin among the "reject object" is high, for example, when the age itself is not stamped. In the first embodiment, a collation method based on positive / negative separation correlation, which will be described later, is used for this age identification process, but another collation method (for example, a general normalized correlation method) may be used.

  The second authenticity identification processing unit 14 searches the feature table 24 using the age passed from the age identification processing unit 13 to obtain a feature part template image representing a characteristic part peculiar to the coin of the age. To do. And this characteristic part template image is collated with the corresponding part of the rotated input image 22 of the memory | storage part 20. FIG. In the first embodiment, a collation method using positive / negative separation correlation described later is used for the second authenticity identification process, but other collation methods (for example, a general normalized correlation method) may be used. Good.

  The authenticity determination unit 15 receives the processing results of the first authenticity identification processing unit 12, the age identification processing unit 13, and the second authenticity identification processing unit 14, and finally determines the authenticity of the coin to be verified. It is a processing unit that performs. For example, when the identification result of the first authenticity identification processing unit 12 is “false coins”, the determination result is output without performing the age identification process or the second authenticity identification process. In addition, when all of the first authenticity identification processing unit 12, the age identification processing unit 13 and the second authenticity identification processing unit 14 indicate that they are “true coins”, the result is output. Will do.

  As described above, in the money identifying apparatus 1 according to the first embodiment, a character table that recognizes a character part that has been frequently excluded from the matching process and manages a character part that is specific to the age of the coin is provided. By using it, the featured part template image of the recognized age is collated with the corresponding part of the input image. Therefore, the collation rate can be improved by efficiently collating the fine features of the coins, and the fake coins can be efficiently detected by maintaining the feature table according to the tendency of the fake coins. .

  Next, the outline | summary of the process sequence of the money identification device 1 which concerns on Example 1 is demonstrated using FIGS. FIG. 2 is a flowchart illustrating an outline of a processing procedure of the money identifying apparatus 1 according to the first embodiment. Details of the processing procedures of the first authenticity identification process (step S102), the age identification process (step S104), and the second authenticity identification process (step S106) shown in FIG.

  First, when the denomination identification processing unit 10 identifies a denomination of a coin based on information detected by the optical sensor 2 and the magnetic sensor 3 (step S101), a first true / false identification process in which the identified denomination is received. The unit 12 collates the coin image (input image) captured by the CCD camera 4 with the registered image 21 in the storage unit 20, thereby identifying the authenticity of the denomination identification performed by the denomination identification processing unit 10. Processing is performed (step S102). If the result of the true / false identification is not “true” (No at Step S103), the true / false determining unit 15 determines “false money” (Step S110) and outputs the determination result.

  Here, an outline of the first authenticity identification process (step S102) will be described with reference to FIG. FIG. 3 is an explanatory diagram for explaining an overview of the first authenticity identification process shown in FIG. As shown in FIG. 3, the first authenticity identification processing unit 12 performs polar processing on the input image received from the CCD camera 4 after performing preprocessing such as edge enhancement. Then, the input image after the polar coordinate conversion is collated with the registered image 21 (template image in FIG. 3) registered after the same preprocessing and polar coordinate conversion. In the first embodiment, a collation method based on a positive / negative separation correlation, which will be described later, is used for the first true / false discrimination process, but another collation method (for example, a general normalized correlation method) may be used.

  Returning to the description of FIG. 2, the processing procedure after step S <b> 103 will be described. When the identification result of the first authenticity identification processing unit 12 is “true” (Yes at Step S103), the rotated input image 22 stored in the storage unit 20 by the first authenticity identification processing unit 12 Using the character collation table, the age identification processing unit 13 performs age identification processing by performing character recognition of the age portion of the coin (step S104). If the age identification process fails (No at Step S105), the authenticity determination unit 15 determines “reject target” (Step S109) and outputs a determination result. On the other hand, when the age identification process is successful (Yes at Step S105), the age that is normally identified is passed to the second authenticity identification processing unit 14.

  Here, an outline of the age identification process (step S104) will be described with reference to FIG. FIG. 4 is an explanatory diagram for explaining an overview of the age identification process shown in FIG. When the authenticity of the denomination is identified by the first authenticity identification processing unit 12, as shown in FIG. 4, the age identification processing unit 13 searches the character verification table 23 in the storage unit 20 to find the corresponding gold Get a character template image of the seed. Then, the age is identified by sequentially collating each character of the age portion cut out from the rotated input image 22 and position-corrected with the acquired character template image. The age normally identified in this way is passed to the second authenticity identification processing unit 14 (Yes in step S105). On the other hand, if any of the characters of the age part related to the rotated input image 22 does not match any of the character template images, the fact that it is “rejected” is passed to the authenticity determination unit 15 ( Step S105 negative).

  In the first embodiment, the character identification process of the age identification processing unit 13 uses a collation method based on the positive / negative separation correlation described later, but uses another collation method (for example, a general normalized correlation method). Also good. In addition, the generalized correlation method uses a collation method based on positive / negative separation correlation for collation processing such as “3” and “5”, which may cause a collation error, and is generally used for collation processing of other characters. A normalized correlation method may be used. Moreover, the character collation table 23 does not necessarily have to be collected for each denomination, and it is sufficient to include at least characters that may be recognized.

  Returning to the description of FIG. 2, the second authenticity identification process will be described. When the age identification is successful in the age identification process (Yes at Step S105), the age that is normally identified is passed to the second authenticity identification processing unit 14. Then, the second authenticity identification processing unit 14 that has received the age searches the feature table 24 of the storage unit 20, acquires the feature template image of the corresponding age, and corresponds to the corresponding portion of the rotated input image 22. (Step S106).

  Here, the outline of the second authenticity identification process (step S106) will be described with reference to FIG. FIG. 5 is an explanatory diagram for explaining the outline of the second authenticity identification process shown in FIG. As shown in FIG. 5, each record of the feature table 24 includes a denomination, an issue year, an X coordinate and a Y coordinate of the feature portion, a width and a height of the feature portion, and a feature portion template image. The table shown in FIG. 5 is an example of the feature table, and may include information such as font information in addition to the information described above.

  The second authenticity identification processing unit 14 searches the feature table 24 using the denomination identified by the first authenticity identification processing unit 12 and the age identified by the age identification processing unit 13. For example, in the figure, since the denomination is “500 yen” and the identified age is “Showa 58”, the denomination of the feature table 24 is “500 yen” and the issue year is “Showa”. The record “58 years” is extracted. Then, the position information (X coordinate, Y coordinate, width, and height) of the feature portion is acquired from the feature table 24, the corresponding portion of the rotated input image 22 is cut out, and the cut out portion and the feature table 24 are acquired. The obtained feature part template image is collated.

  In the first embodiment, a collation method based on a positive / negative separation correlation, which will be described later, is used for the collation processing of the second authenticity identification processing unit 14, but other collation methods (for example, a general normalized correlation method) are used. It is good also as using.

  Returning to the description of FIG. 2, the processing procedure after step S107 will be described. If the result of the second authenticity identification process is not “true” (No at step S107), the authenticity determination unit 15 determines “false coins” (step S110) and outputs the determination result. On the other hand, when the result of the second authenticity identification process is “true” (Yes at step S107), the true / false determining unit 15 determines “true coin” (step S108) and outputs the determination result, The process ends.

  Next, the first authenticity identification process shown in FIG. 2 will be described in more detail with reference to FIGS. FIG. 6 is a flowchart showing a processing procedure of the first authenticity identification processing shown in FIG. As shown in FIG. 6, the first authenticity identification processing unit 12 performs an image cutout process of cutting out a coin portion from the received input image (step S201). Specifically, only the image in the square area circumscribing the input image including coins is cut out.

  FIG. 7 is an explanatory diagram for explaining the outline of the image cut-out process. As shown in the figure, in the image cutting process, the input image 30 acquired from the CCD camera 4 is scanned in the horizontal direction, and the density values of all the pixels are accumulated to generate a horizontal projection 30a. Further, the input image 30 is scanned in the vertical direction, and the vertical projection 30b is generated in the same procedure. Then, the horizontal projection 30a and the vertical projection 30b are scanned, and the rising coordinates and falling coordinates of the accumulated density values are calculated. Subsequently, as indicated by the four broken lines in the figure, the region surrounded by the calculated coordinates is cut out as a cutout image 31, and this cutout image 31 is output to the edge extraction process (step S202). .

  Returning to the description of FIG. 6, the edge extraction process (step S202) will be described. The edge extraction process (step S202) acquires the cutout image 31 from the image cutout process (step S201), and avoids the influence based on individual differences such as the brightness and the hue of the cutout image 31. The density change (edge strength) is calculated. Also, in order to suppress variation in the calculated edge strength, the edge strength is normalized. Specifically, the edge intensity is calculated by performing edge extraction processing using the Sobel operator on the cut image 31, and the calculation result is normalized. In the first embodiment, the Sobel operator is used, but edge extraction can also be performed using a Roberts operator or the like.

FIG. 8 is an explanatory diagram for explaining the Sobel operator. As shown in the figure, in the edge extraction process (step S202), edge strength is calculated using two Sobel operators, ie, a horizontal edge calculation 31a and a vertical edge calculation 31b. Specifically, the Sobel operators (31a and 31b) are scanned for all the pixels of the cutout image 31, and the horizontal edge calculation result Gx and the vertical edge calculation result Gy are acquired. Then, after calculating the edge strength (G) in each pixel, the edge strength is normalized (E).

  As shown in Equation (1), the edge strength (G) in each pixel is expressed as the sum of the absolute value of the horizontal edge calculation result Gx and the absolute value of the vertical edge calculation result Gy. Further, as shown in Expression (2), the normalized edge strength (E) in each pixel is the product of a constant c and an edge strength (G) for which a predetermined value is set for each type of coin, Divided by the sum of the edge strengths (G).

  In this way, by normalizing the edge strength, it is possible to suppress the occurrence of edge strength variation between a new coin that is easy to produce an edge and a current currency that is difficult to produce an edge. Regardless of old or new, various coins can be verified with high accuracy.

  FIG. 9 is an explanatory diagram for explaining an overview of processing (image conversion processing) performed by the edge extraction processing (step S202). As shown in the figure, the cut-out image 31 is converted into an edge extracted image 32 by an edge strength calculation process using a Sobel operator. Then, the edge extracted image 32 is image-converted into an edge normalized image 33 by edge intensity normalization processing using Expression (1) and Expression (2). In the edge extraction process (step S202), the edge normalized image 33 is output to the polar coordinate conversion process (step S203) of the matching process (steps S203 to S205).

  Each pixel value of the edge extraction image 32 shown in the figure takes a value of 0 to 255, for example, 0 corresponds to black, and 255 takes a gray scale value corresponding to white. In the edge extracted image 32 of FIG. 3, the white portion is the extracted edge portion, and the black portion is the background portion. Further, each pixel value of the edge normalized image 33 takes a value of 0 to 255, for example, 0 corresponds to black, and 255 corresponds to a gray scale value corresponding to white. Note that, in the edge normalized image 33 shown in the figure, the white portion corresponds to the edge portion and the black portion corresponds to the background in the same manner as the edge extracted image 32.

  Returning to the description of FIG. 6, the matching process (steps S203 to S205) will be described. In this matching process, the edge normalization image 33 is acquired from the edge extraction process (step S202), and the edge normalized and polar coordinate converted template image is acquired from the registered image 21 of the storage unit 20. Then, the edge normalized image 33 is subjected to polar coordinate conversion, and a deviation angle between the polar coordinate-converted image and the template image is detected by parallel movement of the template image, and the front / back determination is performed to determine the edge normalized image 33. The deviation angle corrected template image is output to the collation process (step S206) based on the positive / negative separation correlation. In the first embodiment, the case where the shift angle is detected by translating the template image will be described. However, the shift angle is detected by translating an image obtained by converting the edge normalized image 33 into polar coordinates. Also good.

を用いる。 In the polar coordinate conversion process (step S203), the edge normalized image 33 is subjected to polar coordinate conversion. Specifically, the center point of the edge normalized image 33 is calculated, and this center point is set as the origin of polar coordinates. Then, each pixel is specified by the rotation angle θ and the distance ρ from the center point, and the polar coordinate conversion is performed by moving each pixel to the ρ-θ space. For such conversion,

Is used.

  FIG. 10 is an explanatory diagram for explaining the processing outline of such polar coordinate conversion. When the center point of the xy space (edge normalized image 33) is the origin and the coordinates of each pixel are represented by (x, y), the x and y and the above-described ρ and θ are expressed by the following equation (3). And the relationship shown in Formula (4). Therefore, by converting each pixel (x, y) in the edge normalized image 33 into (ρ, θ) that satisfies the relationship of the expressions (3) and (4), the polar coordinate conversion process (step S203): A polar coordinate transformed edge normalized image 34 is generated.

  In the figure, the distance ρ from the center point takes a value of 10 to 100, and the rotation angle θ takes a value of 0 to 255, but the range of these values is arbitrary. Can be set.

  Returning to FIG. 6, the rotation angle detection process (step S204) will be described. In the rotation angle detection process (step S204), a deviation angle between the polar coordinate-converted edge normalized image 34 and the template image 40 that has been previously subjected to polar coordinate conversion by the same polar coordinate conversion process is detected, and the deviation angle between both images is corrected. Perform the process. FIG. 11 is an explanatory diagram for explaining the processing outline of the rotation angle detection processing (step S204).

により算出する。 As shown in the figure, in the ρ-θ space, the polar coordinate-converted edge normalized image 34 is moved in parallel with the θ coordinate axis. Then, the shift angle (φ) in the θ coordinate direction and the matching degree M (φ) between the template image 40 and the edge normalized image 33 at each shifting angle (φ) are calculated, and the matching degree M (φ) is the maximum. The rotation angle φ _max is obtained. The coincidence degree M (φ) is

Calculated by

  As shown in Expression (5), the degree of coincidence M (φ) at each shift angle (φ) is the density value t of each pixel of the template image 40 when the polar coordinate-converted edge normalized image 34 is shifted by φ. The product of (k, θ) and the density value s (k, θ−φ) of each pixel of the edge normalized image 33 is summed over each pixel. Here, k is a selection value obtained by selecting a feature-prone distance from the distance ρ from the center point in the edge normalized image 33. For example, k is selected by extracting 16 values of ρ that are prone to feature out of ρ (0 to 100) of the polar coordinate-converted edge normalized image 34 shown in FIG.

The value of M (φ) is a mountain graph having a maximum value at a certain rotation angle as shown in FIG. In the rotation angle detection process (step S204), the value of φ _max at which M (φ) is maximum (the peak portion of the mountain shape) is acquired. As described above, the rotation angle detection process (step S204) corrects the shift angle by the parallel movement of the ρ-θ image subjected to the polar coordinate conversion, and thus is calculated in comparison with the method of correcting the shift angle by the rotation of the xy image. The amount can be reduced.

により求められる。なお、式（６）中のＮは、判定対象となる画素数を示す。 Returning to the description of FIG. 6, the front / back determination processing (step S205) will be described. First, in the front / back determination process (step S205), the maximum value of the degree of matching M (φ) described above between the polar template-converted front and back template images and the polar coordinate-converted edge normalized image 34. M (φ _max ) is calculated, and a normalized correlation coefficient R is obtained from this M (φ _max ). Specifically, this normalized correlation coefficient R is

Is required. Note that N in Expression (6) indicates the number of pixels to be determined.

In the front / back determination process (step S205), a template image having a larger normalized correlation coefficient R is selected and output to the determination process (step S206) based on the positive / negative separation correlation together with the polar coordinate-converted edge normalized image 34. For example, the normalized correlation coefficient R between the back template image and the polar coordinate transformed edge normalized image 34 is larger than the normalized correlation coefficient R between the front template image and the polar coordinate transformed edge normalized image 34. In this case, the back template image and the polar coordinate-converted edge normalized image 34 are output to the matching process (step S206) based on the positive / negative separation correlation. Here, the polar converted edge normalized image 34 to be output to the matching processing by the positive-negative separation correlation (step S206), the angle phi _max only translate to polar coordinate converted edges normalization the displacement angle was corrected with a template image This is a converted image 34.

  Returning to the description of FIG. 6, the registered image 21 used in the matching process (steps S203 to S205) will be described. The registered image 21 stores a plurality of template images corresponding to various coins registered in advance, and provides these template images to the matching process (steps S203 to S205). For such a template image, an average image obtained by combining a plurality of images of the same type of coins is used in order to suppress variations due to individual differences in coins. By using such an average image, the correlation value between the uneven pattern part unique to each coin, such as the issue year, and the corresponding part of the template image becomes the correlation value for the average image (average value), so the effect at the time of matching Is less likely to occur. That is, although it is a true coin, it can be prevented from being determined as a false coin due to the influence of the issue year portion.

  In order to collate such a template image with an input image that has been subjected to polar coordinate conversion processing and edge normalization processing, after being subjected to polar coordinate conversion processing and edge normalization processing in the same manner as the input image, the registered image 21 is stored in the storage unit 20. Registered as The registered image 21 includes template images of the front and back surfaces of each denomination.

  Returning to the description of FIG. 6, the collation process (step S <b> 206) based on the positive / negative separation correlation will be described. The collation processing by the positive / negative separation correlation (step S206) acquires the polar coordinate-converted edge normalized image 34 (hereinafter referred to as “input image 34”) and the template image 40 shown in FIG. By collating images, it is collated to determine whether the coin to be collated is a true coin, and the determination result is output.

  Here, the processing procedure of the collation processing (step S206) based on the positive / negative separation correlation will be described with reference to FIG. FIG. 12 is a flowchart showing the processing procedure of the collation processing based on the positive / negative separation correlation shown in FIG. As shown in FIG. 12, in the collation process based on the positive / negative separation correlation, the input image 34 and the template image 40 output from the matching process are received, and the normalized correlation value calculation process of these images (34 and 40) (step S301). To do.

により各画素の正規化相関値ｒ（ｋ，θ）を算出する。なお、式（７）に示す各画素における正規化相関値ｒ（ｋ，θ）は、たとえば、−１．０〜＋１．０の値をとる。また、式（７）中のｎは、画素数を示す。 Specifically, in the normalized correlation value calculation process (step S301), the correlation value for each corresponding pixel of the input image 34 and the template image 40 is calculated, and the normalized correlation value image is obtained by normalizing the correlation value. Generate. Specifically, for each pixel having a coordinate value of (k, θ), the density value s (k, θ) of the input image 34 and the density value t (k, θ−φ _max ) of the template image 40 are used.

To calculate the normalized correlation value r (k, θ) of each pixel. Note that the normalized correlation value r (k, θ) in each pixel shown in Expression (7) takes a value of −1.0 to +1.0, for example. Moreover, n in Formula (7) shows the number of pixels.

Then, in the normalized correlation value calculation process (step S301), a positive normalized correlation value image (r + image) and a negative normalized correlation value image are determined depending on whether the pixel value of the normalized correlation value image is 0 or more. (R-image). Further, the template image 40 is separated into a positive template image (t + image) and a negative template image (t−image) depending on whether each pixel value is equal to or greater than a predetermined threshold value (T _t ).

  The pixel value of the r + image takes a value of 0.0 to 1.0, for example, and the pixel value of the r− image takes an absolute value of each pixel value, for example, 0.0 to 1. Takes a value of zero. Further, the pixel values of the t + image and the t− image take a binary value of 0 or 1, for example. In other words, the t + image and the t− image serve as image masks used for image conversion of each normalized correlation value image.

  Here, the meaning of each image will be described. The r + image indicates a pixel having a correlation (similarity) between images to be collated, and if there is a strong correlation, the pixel takes a large value. Further, the r-image indicates that there is no correlation (not similar) between the images to be collated, and if there is a strong negative correlation, the pixel takes a large value. The t + image indicates the edge portion of the template image, and the edge portion has a value of 1 and the background portion has a value of 0. Further, the t-image indicates a background portion (a portion that is not an edge) of the template image, and the background portion has a value of 1 and the edge portion has a value of 0.

  Returning to the description of FIG. 12, the positive / negative separated correlation image generation processing (step S302) will be described. In this positive / negative separated correlation image generation process (step S302), a positive / negative separated correlation image is generated by a combination of r + image, r− image, t + image and t− image generated by the normalized correlation value calculation process (step S301). . Specifically, an A + area image is obtained from the r + image and the t + image, an A- area image is obtained from the r− image and the t + image, a B + area image is obtained from the r + image and the t− image, and the r− image and the t− image. A B-region image is generated from each image.

  Here, the meaning of each area image will be described. FIG. 13 is an explanatory diagram for explaining these four areas. As shown in the figure, the A + region image is a region image obtained by superimposing the r + image and the t + image, and is correlated with the edge portion, that is, the edge is appearing where the edge should appear. Show. The A-region image is a region image in which the r-image and the t + image are overlapped, and indicates that there is no correlation with the edge portion, that is, no edge appears where the edge should appear. The B + region image is a region image obtained by superimposing the r + image and the t− image, and indicates that there is a correlation with the background portion, that is, no edge appears where the edge should not appear. The B-region image is a region image in which the r-image and the t-image are overlapped, and indicates that there is no correlation with the background portion, that is, no edge appears where the edge should not appear.

  Returning to the description of FIG. 12, the expansion process (step S303) will be described. In the expansion process (step S303), using a predetermined image mask, the pixels of the A-region image are moved to the A + region image, and the pixels of the B-region image are moved to the B + region image. The reason why such expansion processing is performed is that an isolated point having a negative correlation value in the form of noise appears in the normalized correlation value. In other words, by performing such expansion processing, it is possible to suppress the influence of such isolated points from affecting the collation value determination result.

式（８）により、照合値（Ｚ）を求める。ここで、係数ａ_ij、ｂ_ij、ｃ_ijおよびｄ_ijは学習サンプルを用いて線形判別分析により最適解を求める。なお、各領域画像のブロック値であるＡ⁺ _ij、Ａ^- _ij、Ｂ⁺ _ijおよびＢ^- _ijは、各ブロック内の画素値の総和を示す。 The collation value calculation process (step S304) divides each of the A + region image, the A− region image, the B + region image, and the B− region image into a total of 64 blocks, for example, 16 in the horizontal direction and 4 in the vertical direction.

The collation value (Z) is obtained from equation (8). Here, the coefficients a _ij , b _ij , c _ij, and d _ij obtain an optimal solution by linear discriminant analysis using a learning sample. The block values A ⁺ _ij , A ^- _ij , B ⁺ _ij and B ^- _ij of each area image indicate the sum of the pixel values in each block.

  Subsequently, if the collation value (Z) calculated by the collation value calculation process (step S304) is equal to or greater than the threshold (Yes in step S305), “match” is set as the return value (step S306), and the process returns. On the other hand, if the collation value (Z) is smaller than the threshold value, “no match” is set as the return value (step S307), and the process returns.

  Returning to the description of FIG. 6, the processing after step S207 will be described. When the collation process based on the positive / negative separation correlation (step S206) outputs the collation result, it is determined whether or not the collation result is “match” (step S207). If the collation result is “match” (Yes at step S207), the input image 34 translated in a matching process so as to match the template image 40 is stored in the storage unit 20 as the rotated input image 22 ( In step S208), “true” is set as the return value (step S209), and the process returns. On the other hand, if the collation result is not “match” (No at Step S207), “false” is set as the return value (Step S210), and the process returns.

  As described above, in the first authenticity identification process, the coin part is cut out from the input image received from the CCD camera 4 to perform edge extraction, and after performing the matching process, the collation process by the positive / negative separation correlation is performed, The denomination of the denomination once identified by the denomination identification processing unit 11 is identified. As a result, it is possible to improve the identification accuracy of denominations such as 100-yen coins and 500-yen coins, and fake coins obtained by transforming foreign currencies with similar weight and metal composition.

  Hereinafter, the collation process based on the positive / negative separation correlation shown in FIG. 12 will be described more specifically. First, the normalized correlation value image positive / negative separation process performed by the positive / negative separated correlation image generation process (step S302) will be described with reference to FIGS. FIG. 14 is a flowchart showing a processing procedure of the normalized correlation value image positive / negative separation process included in the positive / negative separated correlation image generation process shown in FIG. 12, and FIG. 17 shows images corresponding to the respective regions shown in FIG. It is explanatory drawing for demonstrating a production | generation procedure.

  As shown in FIG. 17, when the normalized correlation value calculation process (step S301) generates the normalized correlation value image 111 from the input image 33 and the template image 40, the normalized correlation value image positive / negative separation process (step S302) Normalized correlation value positive / negative separation processing is performed by using the generated normalized correlation value image 111 as an input. Specifically, the normalized correlation value image 111 is separated into an r + image 111a that is a positive correlation value image and an r− image 111b that is a negative correlation value image.

  As shown in FIG. 14, in the normalized correlation value positive / negative separation process (step S302), first, the process moves to the start point pixel of the normalized correlation value image 111 (step S501). Such a start pixel is, for example, a pixel with k = 0 and θ = 0. Then, the normalized correlation value r (k, θ) of the pixel is calculated using the equation (4) (step S502). If the calculated r (k, θ) is 0 or more (Yes in step S503), Such a pixel value is set as a pixel value at the same coordinates of the r + image 111a (step S504). On the other hand, if the calculated r (k, θ) is smaller than 0 (No in step S503), the absolute value of the pixel value of the pixel is set as the pixel value of the same coordinate of the r-image 111b (step S505).

  If the positive / negative separation process has not yet been completed for all the pixels of the normalized correlation value image 111 (No at Step S506), the process moves to the next target pixel (Step S507), and the processes after Step S502 are performed. repeat. On the other hand, when the positive / negative separation process has been completed for all the pixels (Yes in step S506), the process ends. By such normalized correlation value positive / negative separation processing, the r + image 111a and the r− image 111b are generated as images having pixels having pixel values of 0.0 to 1.0. In the first embodiment, the pixel value of the pixel of the r-image 111b is described as a pixel value of 0.0 to 1.0, but the pixel value is −1.0 to 0.0. It is good also as taking the value of.

  Next, the template image positive / negative separation process performed by the positive / negative separated correlation image generation process (step S302) shown in FIG. 12 will be described with reference to FIGS. FIG. 15 is a flowchart illustrating a processing procedure of template image positive / negative separation processing included in the positive / negative separation correlation image generation processing illustrated in FIG. 12. As shown in FIG. 17, in the template image positive / negative separation process, the template image 40 is separated into a t + image 40a that is a positive template image and a t− image 40b that is a negative template image.

As shown in FIG. 15, in the template image positive / negative separation process, first, the template image 40 is moved to the start point pixel (step S601). Such a start pixel is, for example, a pixel with k = 0 and θ = 0. If the density value of the pixel is equal to or greater than a predetermined threshold value (T _t ) (Yes at step S602), the pixel value at the same coordinate of the t + image 40a is set to 1 (step S603). On the other hand, if the density value is smaller than the predetermined threshold value (T _t ) (No at Step S602), the pixel value at the same coordinate of the t-image 40b is set to 1 (Step S604).

  If the positive / negative separation process has not yet been completed for all the pixels of the template image 40 (No at Step S605), the process moves to the next pixel of interest (Step S606), and the processes after Step S602 are repeated. On the other hand, when the positive / negative separation process has been completed for all the pixels (Yes in step S605), the process ends. By this template image positive / negative separation processing, the t + image 40a is generated as a binary image with an edge portion of 1 and a background portion of 0, and the t-image 40b is a binary image with an edge portion of 0 and a background portion of 1. Generated.

  Next, the positive / negative separated correlation image generation processing performed by the positive / negative separated correlation image generation processing (step S302) shown in FIG. 12 will be described with reference to FIGS. FIG. 16 is a flowchart of a process procedure of the region image generation process included in the positive / negative separated correlation image generation process shown in FIG.

  As shown in FIG. 17, in the positive / negative separated correlation image generation processing, the r + image 111a, the r− image 111b, the t + image 40a generated in the normalized correlation value image positive / negative separation processing and the template image positive / negative separation processing described above, and Using the t-image 40b as an input image, an A + region image 121, an A- region image 122, a B + region image 123, and a B- region image 124 are generated.

  For example, when the r + image 111a and the t + image 40a are used as input images, first, as shown in FIG. 16, the image is moved to the start point pixel of each image (step S701). If the pixel value of the t + image 40a at this pixel is 1 (Yes at step S702), the pixel value of the A + region image 121 is set as the pixel value of the r + image 111a (step S703). On the other hand, when the pixel value of the t + image 40a in such a pixel is not 1 (that is, 0) (No in step S702), the pixel value of the A + region image 121 is set to 0 (step S704).

  If the region image generation processing has not been completed for all the pixels (No at Step S705), the process moves to the next pixel of interest (Step S706), and the processing from Step S702 is repeated. On the other hand, when the region image generation processing has been completed for all the pixels (Yes in step S705), the A + region image 121 is generated, and thus the processing ends.

  Similarly, an A-region image 122 from the r-image 111b and the t + image 40a, a B + region image 123 from the r + image 111a and the t-image 40b, and a B-region image 124 from the r-image 111b and the t-image 40b. , Generate each.

  Next, the expansion process (step S303) shown in FIG. 12 will be described with reference to FIGS. 18 is an explanatory diagram for explaining an image mask used in the expansion process shown in FIG. 12, FIG. 19 is a flowchart showing a processing procedure of the expansion process shown in FIG. 12, and FIG. It is explanatory drawing for demonstrating the image produced | generated by the expansion process shown in FIG.

  In such dilation processing, noise-like isolated points (pixels) included in the negative region images (A− region image 122 and B− region image 124) are converted into positive region images (A + region image 121 and B + region image 123). ) To move to. By performing such processing, the accuracy of the collation value can be increased.

  As shown in FIG. 18, in the expansion process, two image masks of a positive area image mask 130a and a negative area image mask 130b are used. Each image mask has P5 and M5 and eight regions surrounding these regions. For example, when the expansion process from the A− region image 122 to the A + region image 121 is performed, M5 of the negative region image mask 130b is matched with the target pixel of the A− region image 122, and P5 of the positive region image mask 130a is focused. Fit to the pixel corresponding to the pixel. Then, the expansion process is performed by sequentially comparing the pixel value of M5 and the pixel values of P1 to P9.

  Next, the procedure of the expansion process will be described with reference to FIG. 19, taking as an example the case where the expansion process from the A-region image 122 to the A + region image 121 is performed. First, it moves to the start point pixel of each image (121 and 122) (step S801). Such a start pixel is, for example, a pixel with k = 0 and θ = 0. Then, 1 is set to n in order to sequentially switch the nine regions (P1 to P9) of the primary region mask 130a (step S802). That is, when step S802 is completed, the target area of the normal area image mask 130a is P1.

  Then, the value of Pn is compared with the value of M5, and if the value of P1 is larger than the value of M5 (Yes at Step S803), the value of P5 is replaced with M5 and the value of M5 is set to 0 (Step S803). S805). That is, the pixel M5 is moved to the pixel P5. On the other hand, when the value of Pn is less than or equal to the value of M5 (No at Step S803), 1 is added to the value of n (Step S804), and when the value of n is 9 or less (No at Step S806). Step S803 is performed again.

  As described above, if any one of the values P1 to P9 is larger than the value M5, the pixel M5 is moved to P5. On the other hand, if any of the values of P1 to P9 is equal to or less than the value of M5 (Yes at step S806), the pixel is not moved.

  If the process has not been completed for all the pixels of the A-region image 122 (No at Step S807), the process moves to the next target pixel (Step S808), and the processes after Step S802 are performed. On the other hand, when the process is completed for all the pixels of the A-region image 122 (Yes in step S807), the expansion process is terminated.

  As shown in FIG. 20, the A + region image 121, the A− region image 122, the B + region image 123, and the B− region image 124 are respectively expanded by the expansion process. The image is converted into an image 132, an expanded B + region image 133, and an expanded B− region image 134. Since the isolated point on the A− region image 122 moves to the A + region image 121, the edge portion of the expanded A + region image 131 has an area larger than that of the A + region image 121. On the other hand, the area of the edge portion of the expanded A-region image 132 is smaller than that of the A-region image 122.

  Next, the collation value calculation process (step S304) shown in FIG. 12 will be described with reference to FIG. FIG. 21 is an explanatory diagram for explaining an example in which the expanded A + region image 131 is divided into blocks with respect to the block division of the expanded region image (131 to 134). In the collation value calculation process (step S304), first, as shown in the figure, the expanded A + region image 131 is divided into a total of 64 blocks of 16 in the horizontal direction and 4 in the vertical direction. Similarly, the block division is performed on the expanded A-region image 132, the expanded B + region image 133, and the expanded B-region image 134.

In the matching value calculation process (step S304), the matching value (Z) is calculated using equation (8). Here, for each coefficient a _ij , b _ij , c _ij, and d _ij in equation (8), an optimal solution is obtained by linear discriminant analysis or the like using a learning sample. Specifically, because of the difference in the design of the concavo-convex pattern of coins, there are coins that are likely to have an edge and coins that are difficult to come out, so these coefficients take different values for each type of coin. By optimizing these coefficients with learning samples, accurate image matching can be performed.

In the matching value calculation process (step S304), the matching value (Z) is calculated using the coefficients a _ij , b _ij , c _ij and d _ij for which the optimum values are set, and each image block. Then, as shown in FIG. 12, when the collation value is greater than or equal to the threshold value (Yes at Step S305), “match” is set as the return value (Step S306), and when it is smaller than the threshold value (Step S305). No), “unmatched” is set as the return value (step S307). In the first embodiment, the case where each image is divided into 64 blocks has been described. However, the number of blocks may be an arbitrary number.

If the coefficients c _ij and d _ij are set to 0 in equation (8), the collation value (Z) can be calculated from only the A + area image block and the A− area image block. If the coefficients a _ij and b _ij are set to 0, the collation value (Z) can be calculated from only the B + region image block and the B− region image block.

  As described above, in the collation value calculation process (step S304), the image collation is efficiently performed by adjusting the number of image blocks and the values of the coefficients of the equation (8) according to the type of coin and the capability of the hardware. Can be done.

  In the collation value calculation process (step S304) according to the first embodiment, each region image is divided into blocks, and the collation value (Z) is calculated by Expression (8). For example, other methods such as a multilayer neural network, a support vector machine, and a secondary discriminant function may be used.

  Next, the age identification process (step S104) shown in FIG. 2 will be described with reference to FIGS. FIG. 22 is a flowchart showing a processing procedure of the age identification process shown in FIG. 2, and FIG. 23 is an explanatory diagram for explaining the character matching table 23 used in the age identification process shown in FIG.

  As shown in FIG. 22, the age identification processing unit 13 first reads out the rotated input image 22 from the storage unit 20 and performs a process of cutting out the age part corresponding to the corresponding denomination (step S1001). Then, after correcting the position of the extracted age part (step S1002), one character included in the extracted age part is extracted (step S1003).

  Subsequently, the character collation table 23 in the storage unit 20 is searched to obtain a character template image corresponding to the corresponding denomination (step S1004), and the collated processing using the positive / negative separation correlation between the cut out character image and this character template image. (Step S206).

  Here, the character collation table 23 will be described. As shown in FIG. 23, the character collation table 23 has a record including a denomination and a character template image. For example, each record relating to a 500 yen coin includes a character template image of characters to be engraved in a 500 yen coin such as “Akira”, “Wa”, “Hira”, “Sei”, “I”, “2”.

  The character matching table 23 shown in FIG. 23 is an example of the character matching table, and includes information indicating the position of each character included in the input image 34 in addition to the denomination and the character template image. It is also possible to cut out each character representing the age from the input image 34 using such information. Moreover, the character collation table 23 does not necessarily have to be collected for each denomination, and it is sufficient to include at least characters that may be recognized.

  Returning to FIG. 22, the processing after step S1005 will be described. When the collation result of the collation process based on the positive / negative separation correlation is received, it is determined whether or not the collation result is “match” (step S1005). When the collation result is “match” (Yes at step S1005), the matched character is temporarily stored (step S1006), and it is determined whether or not the character to be collated is the last character (step S1006). Step S1007).

  When the character to be verified is the last character (Yes at Step S1007), it is determined whether or not the age of the recognized character is valid (Step S1009), and when it is valid (Step S1009). (Yes), the recognized “age” is set as the return value (step S1010), and the process returns. On the other hand, if the character to be collated is not the last character (No at Step S1007), the processing after Step S1003 is repeated in order to collate the next character included in the extracted age part.

  If the collation result of the collation process based on the positive / negative separation correlation is not “match” (No at Step S1005), it is determined whether or not it is the last character template image for the denomination to be collated (Step S1008). If it is the last character template image (Yes in step S1008), the character template image does not exist and character recognition is not possible, that is, the corresponding character in the input image 34 is invalid. “Reject target” is set (step S1011) and the process returns. On the other hand, if another character template image exists for the denomination to be collated (No at step S1008), the current character template image is changed to another character template image, and the processing from step S206 is repeated.

  In this way, the age identification processing unit 13 passes the age identified and identified to the age included in the input image 34 to the second authenticity identification processing unit 14. In addition, in Example 1, although the case where each character which comprises the age inscribed in the coin was collated using the character collation table 23 was demonstrated, for example, the serial number of a banknote, a serial number, and an issue year are shown. Information representing the relationship may be registered in the character collation table 23 and the issue year may be identified based on the serial number included in the input image of the banknote.

  Next, the second authenticity identification process (step S106) shown in FIG. 2 will be described with reference to FIGS. FIG. 24 is a flowchart showing a processing procedure of the second authenticity identification process shown in FIG. 2, and FIG. 25 is a diagram for explaining a feature table used in the second authenticity identification process shown in FIG. It is explanatory drawing of.

  As shown in FIG. 24, the second authenticity identification processing unit 14 first sets the age passed from the age identification processing unit 13 (step S1101), and searches the feature table 24 of the storage unit 20 ( Step S1102). Then, a feature range and a feature portion template image representing the position and range of the image area of the feature portion are acquired from the corresponding record in the feature table 24 (step S1103).

  Here, the feature table 24 will be described. As shown in FIG. 25, each record of the feature table 24 has a denomination, an issue year, a feature portion X coordinate and a Y coordinate, a feature portion width and height, and a feature portion template image. For example, in the case of a 500-yen coin, the “0” part of “500”, which represents the amount of money on the back of the coin, has a different design depending on whether it was issued before 1999 or after 2000. This “0” portion is registered as a feature portion template image. Further, the X coordinate, Y coordinate, width and height representing the position of the “0” portion are also registered.

  FIG. 25 shows an example of the feature table 24, and the case of a 500 yen coin currently issued has been described. However, the provision of fine security features to deal with counterfeit money is performed not only in Japanese money but also in foreign currencies, and in the future, it may have different security features depending on the year of issue. Therefore, as shown in the first embodiment, if the feature part corresponding to the issue year is managed in a table, the issue year is identified and the detailed matching process is performed using the feature part corresponding to the issue year. It is possible to determine whether or not it is a true coin with high accuracy.

  In FIG. 25, the case where one feature portion is registered in the feature table 24 for each issue year has been described, but a plurality of feature portions may be registered for each issue year. Furthermore, the feature part template image registered in the feature table 24 does not need to be a feature part that can be recognized by the naked eye, and may include fine features that can be recognized by a highly accurate camera.

  Returning to FIG. 24, the processing procedure after step S1104 will be described. If the feature range and feature portion template image representing the image area of the feature portion are acquired from the corresponding record of the feature table 24 (step S1103), the acquired feature range (X coordinate, Y coordinate, width and height) is obtained. The feature part to be collated is cut out from the input image 34 using it (step S1104). Subsequently, the feature part cut out from the input image and the feature part template image acquired from the feature table 24 are collated by the collation process based on the positive / negative separation correlation described above (step S206).

  When the collation result of the collation process using the positive / negative separation correlation is received, it is determined whether or not the collation result is “match” (step S1105). If the collation result is “match” (Yes at step S1105), “true” is set as the return value (step S1106), and the process returns. On the other hand, if the collation result is not “match” (No at step S1105), “false” is set as the return value (step S1107) and the process returns.

  In this manner, the second authenticity identification processing unit 14 collates the feature part related to the issue year based on the age recognized by the age identification processing unit 13. And the authenticity determination part 15 which received the collation result of the 2nd authenticity identification process part 14 will output a determination result.

  As described above, in the money identifying apparatus, the money identifying method, and the money identifying program according to the first embodiment, the denomination identifying processing unit obtains the denomination of the coin and receives the obtained denomination. The identification processing unit is configured to perform true / false identification processing of coins. Then, the age identification processing unit identifies the issue year of the coin, and the second authenticity identification processing unit searches the feature table using the previously identified denomination and issue year, and the feature part template image and the feature The range is acquired, and the feature portion template image is matched with the corresponding portion of the input image. Furthermore, the authenticity determination unit is configured to determine the authenticity of the coin based on the processing results of the first authenticity identification process, the age identification process, and the second authenticity identification process. Therefore, highly accurate currency identification can be performed. In addition, by recognizing the character included in the input image and collating the feature portion related to this character in detail, the image collation accuracy can be improved, and only the feature portion of the input image is subjected to detailed collation processing. By doing so, it is possible to efficiently collate fine features. Further, by maintaining the feature table according to the tendency of the false coins, the false coins can be efficiently identified.

  Example 1 mentioned above demonstrated the case where the denomination identification process of a coin was performed using sensors, such as an optical sensor and a magnetic sensor. In the second embodiment described below, a case where the denomination identifying process is performed using an image captured by a CCD camera or the like will be described.

  FIG. 26 is a functional block diagram illustrating the configuration of the money identifying apparatus according to the second embodiment. In the following description, only differences from the currency identifying apparatus according to the first embodiment will be described, and description of common points will be omitted.

  As shown in FIG. 26, the denomination identification processing unit 11 of the currency identification device 1 according to the second embodiment acquires a captured image (input image) of a coin from the CCD camera 4, and based on this input image, denomination of the denomination Identify. The denomination identification processing unit 11 outputs the identified denomination and the input image acquired from the CCD camera to the first authenticity identification processing unit 12.

  The denomination identification processing unit 11 uses a general normalized correlation method when identifying the denomination from the input image. And in the 1st authenticity identification process part 12, the authenticity identification of the money type which the money type identification process part 11 identified is performed by performing the collation process by above-mentioned positive / negative separation correlation. By doing in this way, it becomes possible to perform the denomination and authenticity determination of a coin, without using an optical sensor (refer 2 of FIG. 1) and a magnetic sensor (refer 3 of FIG. 1). In addition, it is good also as a denomination identification process part 11 identifying a denomination by the collation process by positive / negative separation correlation.

  In FIG. 26, the denomination identification processing unit 11 acquires an image (input image) obtained by imaging a coin from the CCD camera 4, and outputs the denomination identification result and the input image to the first authenticity identification processing unit 12. However, the denomination identification processing unit 11 itself may be omitted, and the first true / false identification processing unit 12 may perform denomination identification and first authenticity identification. Hereinafter, a configuration in which the denomination identification processing unit 11 is omitted will be described.

  FIG. 27 is a functional block diagram illustrating a modified example of the configuration of the money identifying apparatus according to the second embodiment. As shown in the figure, the first authenticity identification processing unit 12 uses an input image received from the CCD camera 4 and a registered image 21 stored in advance in the storage unit 20, and performs a collation process using positive / negative separation correlation. Denomination identification and first authenticity identification are performed. As described above, the collation process based on the positive / negative separation correlation has a high collation accuracy, so that the denomination identification accuracy is sufficiently high. Therefore, if denomination identification is performed by positive / negative separation correlation, this denomination identification can also serve as the first true / false identification.

  Next, the processing procedure of the first authenticity identification processing shown in FIG. 27 will be described with reference to FIG. FIG. 28 is a flowchart showing a processing procedure of first authenticity identification processing according to the modification shown in FIG. Note that the processing procedure shown in FIG. 28 is obtained by adding processing for denomination identification to the processing procedure shown in FIG. 6, so only the differences from FIG. 6 will be described. The description will be omitted.

  First, in the processing after the rotation angle detection process (step S204), the template image 40 corresponding to a predetermined currency is selected from the plurality of template images 40, and the rotation angle detection process is performed. For example, the template image 40 corresponding to 1 yen coin is selected from the template images 40 of coins such as 1 yen coin, 5 yen coin, and 10 yen coin. In the subsequent processing, if the image matching using the template image 40 fails, the template image 40 such as the template image 40 corresponding to the 5-yen coin and the template image 40 corresponding to the 10-yen coin are displayed. By changing the template image 40, the template image 40 matching the coin to be verified is found. In this way, in the first authenticity identification process shown in FIG. 28, the denomination is identified.

  When the collation process based on the positive / negative separation correlation (step S206) outputs the collation result, it is determined whether or not the collation result is “match” (step S207). If the collation result is “match” (Yes at step S207), the input image 34 translated in a matching process so as to match the template image 40 is stored in the storage unit 20 as the rotated input image 22 ( In step S208), “true” is set as the return value (step S209), and the process returns.

  On the other hand, when the collation result is not “match” (No at Step S207), it is determined whether there is another money candidate (Step S211). Specifically, it is determined whether or not there is a template image 40 that is not yet used for the image matching process. If all denominations have been collated (No at Step S211), “false” is set as the return value (Step S210), and the process returns. If there is a template image 40 that has not yet been used for the image matching process (Yes at step S211), the template image 40 is changed to that of another currency, and the processing after step S204 is repeated, thereby denomination identification. Will be performed.

  As described above, in the second embodiment, the denomination identification processing unit performs denomination identification using an image captured by a CCD camera or the like, and further, the denomination identification processing unit itself is omitted, and the first true Since the false identification processing unit is configured to perform denomination identification and first authenticity identification, it is possible to efficiently perform image matching processing without using various sensors such as an optical sensor and a magnetic sensor.

  In addition, in each Example mentioned above, although the case where money identification was performed by performing image collation about the input image of a coin was explained, the present invention is not limited to this, for example, identification of a bill, The present invention can also be applied to identification of security signs attached to products distributed in the market. Moreover, by applying the present invention to a coin sorter and a money changer, it is possible to wrap every denomination and issue year after collecting true coins, or to collect old coins and fake coins.

  The coin classification device and the coin classification method according to the present invention are suitable for efficiently classifying coins in consideration of the age when the coins were issued.

DESCRIPTION OF SYMBOLS 1 Money identification device 2 Optical sensor 3 Magnetic sensor 4 CCD camera 10 Control part 11 Denomination identification process part 12 1st authenticity identification process part 13 Era identification process part 14 2nd authenticity identification process part 15 Authenticity determination part DESCRIPTION OF SYMBOLS 20 Memory | storage part 21 Registration image 22 Rotated input image 23 Character collation table 24 Feature table 30 Input image 30a Horizontal direction projection 30b Vertical direction projection 31 Cutout image 31a Sobel operator (for horizontal direction edge calculation)
31b Sobel operator (for vertical edge calculation)
32 Edge extracted image 33 Edge normalized image 34 Polar coordinate transformed edge normalized image (input image)
40 template image 40a t + image 40b t-image 111 normalized correlation value image 111a r + image 111b r-image 121 A + region image 122 A-region image 123 B + region image 124 B-region image 130a positive region image mask 130b negative region image Mask 131 Inflated A + area image 132 Inflated A- area image 133 Inflated B + area image 134 Inflated B- area image

Claims (6)

A coin classification device that performs the coin classification process by collating an input image of coins with a pre-registered character template image,
Polar coordinate conversion image acquisition means for converting the input image into polar coordinates;
Rotation angle acquisition means for acquiring a rotation angle by collating with a reference image registered in advance;
Image rotation means for matching the input image of the coin with the rotation angle of the character template based on the acquired rotation angle;
The partial image including the character part indicating the issue year that forms part of the input image of the coin and the character template image are collated in a state where the rotation angles of both images are aligned, and based on the collation value, An issue year identification means for identifying the issue year of a coin;
A coin classification device comprising: classification processing means for performing classification processing on the coins based on an identification result of the issuing year identification means.   2. The coin classification apparatus according to claim 1, wherein the classification processing unit classifies the coins of a predetermined issue year identified by the issue year identification unit with coins of another issue year.   3. The coin classification device according to claim 1, wherein the classification processing unit performs a classification process on a coin as a subject to be rejected when the issue year of the coin is not identified by the issue year identification unit. The classification processing means includes
Corresponding to the issue year when the issue year of the coin is identified by the feature template storage means for storing a plurality of feature part templates respectively indicating the characteristic portions of the coins of the plurality of issue years, and the issue year identification means Search means for searching for a feature template of coins from the feature template storage means;
Based on a collation value between a partial image showing a characteristic part of the coin that forms a part of the input image of the coin and a characteristic template searched by the search means, the coin is a real coin or a fake coin The coin classification device according to claim 1, 2 or 3, further comprising: authenticity identification processing means for performing identification processing.   The coin classification device according to any one of claims 1 to 4, wherein the classification processing unit classifies issued coins after a predetermined issuing year and old coins issued until then. A coin classification method of a coin classification device that performs a classification process of the coins by collating an input image of coins with a character template image registered in advance,
Converting the input image into polar coordinates;
A rotation angle acquisition step of acquiring a rotation angle by collating with a reference image registered in advance;
An image rotation step of matching the rotation angle between the input image of the coin and the character template based on the acquired rotation angle;
The partial image including the character part indicating the issue year that forms a part of the input image of the coin is collated with the character template image in a state in which the rotation angles match, and based on the collation value, the issue year of the coin Issue year identification process for identifying
And a classification processing step of classifying the coins based on the identification result of the issue year identification step.

Images