CN101847287B - Number identifying device and method, paper processing apparatus and automatic transaction processing apparatus - Google Patents

Abstract

Provided is a serial number identification device and method, a paper processing device, and an automatic transaction processing device, capable of reliably identifying serial numbers marked on paper. The banknote identification unit (number identification device) (25) takes a color photograph of the banknote, uses the color photograph to determine the denomination and direction of the banknote, and identifies the serial number. The banknote identification unit (25) cuts out the area where the banknote number is displayed, erases the background color or pattern of the area, extracts the image of the number, and performs character recognition when identifying the number.

Description

Number recognition device and method, paper processing device, automatic transaction processing device

technical field

The present invention relates to a serial number recognition device for recognizing a serial number marked on paper sheets, a paper sheet processing device, an automatic transaction processing device, and a serial number recognition method.

Background technique

Conventionally, a serial number recognition device having a plurality of light sources and an image sensor and selectively driving a light source of a color to drop out a background color based on currency denomination information has been proposed (see Patent Document 1).

Patent document: Japanese Patent Laid-Open No. 2004-213560

In general, banknotes often have different background colors for each denomination. For this reason, in the invention described in Patent Document 1, in order to select the light source corresponding to the currency type, first, the banknotes are successively sent out from the continuous delivery part, and the currency type information and direction information are obtained by the currency type identification part, and collected in the banknotes. Temporary reservations. Then, the banknote is sent to the serial number reading part and the light source corresponding to the currency type is turned on, and the serial number (recording serial number) of the banknote is read to carry out the identification process of the serial number. Thus, in the conventional serial number recognition apparatus, there existed a problem that it took time until the serial number of a banknote was read.

In addition, in order to correctly recognize the serial number, it is necessary to provide light sources of a plurality of colors corresponding to the background color of the banknote, and there is a problem of cost.

In addition, when a banknote with a background color different from the conventional one is issued, there is a problem that a new color light source must be added in order to recognize the serial number of this denomination.

Contents of the invention

Therefore, an object of the present invention is to provide a serial number recognition device, a paper sheet processing device, an automatic transaction processing device, and a serial number recognition method capable of reliably recognizing a serial number of paper sheets.

In the present invention, the color image of the paper is taken by the imaging mechanism; the type and direction of the paper are judged by the discrimination mechanism based on the color image. And, through the selection mechanism, based on the information of the type and direction of the paper, the numbered area marked with the number is determined, and only the pixels of the color determined based on the type information of the paper are selected from the image of this area, and the color is selected by the character recognition means. , perform character recognition on the image composed of the selected pixels. Although the number or the background color or pattern around the number differs depending on the type of paper, since the color of the number can be determined based on the information of the type of paper, using this information enables selection of only pixels corresponding to the number, and reliable Extract the numbered string.

In addition, in the present invention, the selection unit acquires the luminances of R, G, and B for each pixel constituting the image in the numbered area, and selects the pixel based on the values of these luminances within the threshold range. When each luminance value of a pixel is within the threshold range, the pixel is a pixel constituting a character string of numbers. Therefore, even if the serial number or the background color of the paper changes due to variations in printing on the paper or deterioration associated with use of the paper, the serial number can be reliably extracted for character recognition.

In addition, in the present invention, the selection means acquires the luminances of R, G, and B for each pixel constituting the image in the numbered area, and calculates the luminances of R, G, and B to be acquired and the corresponding values according to the paper type. The value of the linear combination of the eigenvectors defined for each category of , if the value is within the threshold range, it is selected. When the value of the linear combination is within the threshold, that pixel is the one that makes up the numbered string. Therefore, even if the serial number or the background color of the paper changes due to variations in printing on the paper or deterioration associated with use of the paper, the serial number can be reliably extracted for character recognition.

In addition, in the present invention, if the value based on each luminance of R, G, and B is out of the range of the above-mentioned threshold value, the selection means replaces it with another luminance set in advance. When each luminance of a pixel or a linearly combined value is outside the threshold range, the pixel is not a pixel constituting the numbered character string but a background pixel. Therefore, only the character string of the number can be extracted by substituting all the pixels of the background of the number with other luminances set in advance.

In addition, in the present invention, the useless area removing means performs image processing with a median filter on the image constituted by the pixels selected by the selecting means to remove unnecessary pixels. Image processing with a median filter can remove noise components such as tiny dots. Therefore, when unnecessary pixels are included in the image selected by the selection means, the character string of the serial number can be reliably extracted by performing image processing by the useless area removal means.

Invention effect

According to the present invention, numbered character strings can be reliably extracted.

Description of drawings

Fig. 1 is a block diagram showing the configuration of the main parts of the ATM.

Fig. 2 is a schematic diagram showing a banknote transport path formed inside the ATM.

Fig. 3 is a block diagram showing a configuration of a main part of a banknote recognition unit.

Fig. 4 is an external view showing an example of a banknote serial number.

Fig. 5 is a side view showing a state when an image of a banknote is captured by an imaging unit.

FIG. 6 is a diagram showing an extraction result of a serial number.

FIG. 7 is a comparison explanatory diagram of two processing methods for selecting pixels.

FIG. 8 is a diagram showing the result of numbered useless region removal processing.

9 is an example of a table in which data of the first pixel selection process, the second pixel selection process, and the unnecessary area removal process are recorded.

Fig. 10 is a flowchart for explaining the operation of the banknote recognition unit.

FIG. 11 is a flowchart illustrating first pixel selection processing.

FIG. 12 is a flowchart for explaining second pixel selection processing.

Symbol Description:

1…ATM

2…Main control unit

4a...Banknote deposit and withdrawal port

20...Banknote conveying path (transfer section)

21…ROM

21…Paper money box

22…recycling box

23…Temporary reservation department

24... Deposit and withdrawal banknote retention department

25...Banknote identification department

30…control department

31…ROM

32…RAM

33...camera department

34...Currency identification department

35...cut-out part

36...Selection Department

37...Useless area removal part

38...Character recognition department

Detailed ways

In the following description, an automatic teller machine (hereinafter referred to as ATM) will be described as an example of the automatic transaction processing apparatus. In addition, a case where banknotes are used as an example of paper sheets will be described. Fig. 1 is a block diagram showing the structure of the main part of the ATM.

As shown in Figure 1, ATM1 has: main control part 2, display/operating unit 3, banknote processing unit 4, coin processing unit 5, card/transaction list processing unit 6, passbook processing unit 7, communication unit 8, and user Detection unit 10. This ATM1 is installed in a branch of a financial institution, a convenience store, etc., and processes the transaction (deposit transaction or withdrawal transaction) of the type requested by the user. The main control unit 2 controls the operation of each unit. Moreover, the main control part 2 switches the main body mode of ATM1 between a standby mode and an execution mode based on the signal input from the user detection means 10. The standby mode is a mode in which the supply of operating power to some functions of the ATM1 main body is stopped, and the power consumption of the ATM1 main body is suppressed. The execution mode is a mode for executing processing requested by the user.

The display/operation unit 3 has a display 3a (not shown) provided on the front of the main body, and a touch panel 3b (not shown) attached to the display 3a. The display/operation unit 3 displays an operation guidance screen for the user on the display 3a. In addition, the display/operation unit 3 accepts the user's input operation related to the password and the transaction content (processing type, deposit and withdrawal amount, etc.) by detecting the pressed position of the touch panel 3b. In addition, the display/operation unit 3 may include a biological information reading unit for reading user biological information (fingerprint, finger vein, palm print, etc.).

The banknote processing unit 4 has a banknote conveyance section (not shown) that conveys banknotes along a banknote conveyance path formed by a banknote deposit and withdrawal port 4a (not shown) provided on the front of the main body, and a banknote delivery port (not shown) provided inside the main body. Between banknote boxes (picture omitted) according to different currencies. In addition, the banknote processing unit 4 further includes a banknote identification unit (not shown) that performs denomination (type of banknote) and authenticity identification for each banknote conveyed along the banknote conveyance path. The coin handling unit 5 has a coin conveying section (not shown) that conveys coins along a coin conveying path formed in a coin deposit and withdrawal port 5a (not shown) provided on the front of the main body, and a Between coin boxes (not pictured) according to different currencies. In addition, the coin processing unit 5 also has a coin identification unit (not shown) for identifying the type and authenticity of each coin conveyed along the coin conveying path. The banknote handling unit 4 corresponds to the banknote handling device of the present invention.

The card/transaction list processing unit 6 takes in the card inserted from the card slot 6a (not shown) arranged on the front of the main body, reads the card information (financial institution number, business network number, account number, etc.) recorded on the card, Or rewrite the card information, etc. The card/transaction list processing unit 6 may be a structure for processing magnetic cards, or an IC card, or two types of cards (magnetic card and IC card). In addition, the card/transaction statement processing unit 6 also has a printing unit (not shown) that prints transaction details on the transaction statement. The card/transaction statement processing unit 6 sends out the transaction statement printed with the transaction content from the transaction statement delivery port 6b (not shown) provided on the front of the main body.

The passbook processing unit 7 takes in a passbook inserted through a passbook slot 7a (not shown) provided on the front of the main body, and has a printing unit (not shown) for printing transaction history on the passbook. In addition, the passbook processing unit 7 also has a page-turning mechanism for turning the pages of the incoming passbook, a barcode reader for reading the barcode printed on the passbook to indicate the page number, and a barcode reader for reading the barcode recorded on the magnetic stripe attached to the passbook. The magnetic head of the passbook information (financial institution number, business outlet number, account number, etc.). The communication unit 8 controls data communication with a high-level device (not shown) installed at the center.

The user detection unit 10 detects the user of the ATM1 body. The user detection unit 10 can also be installed on the ceiling of the ATM1 body or built in the ATM1 body.

Next, the banknote conveyance in the banknote processing unit 4 will be briefly described. Fig. 2 is a schematic diagram showing a banknote transport path formed inside the ATM. In the lower part of ATM1, the banknote cassette (paper cassette) 21 (21a-21d) which accommodates banknotes for each denomination, and the collection box 22 which collects defective banknotes are accommodated. The banknote box 21 and the recycling box 22 are freely detachable with respect to the ATM1 body. In addition, on the upper part of ATM1, an escrow unit 23 for temporarily holding banknotes, a deposit and withdrawal banknote storage unit 24 provided at the banknote deposit and withdrawal port, and a banknote identification unit 25 for identifying currency types and authenticity of banknotes are provided. The banknote transport path 20 is connected with a banknote cassette 21 , a collection box 22 , an escrow unit 23 , a deposit and withdrawal banknote storage unit 24 , and a banknote identification unit 25 . The banknote handling unit 4 controls gates 26 ( 26 a to 26 f ) provided at branch points of the banknote transport path 20 to switch the transport destination (transport path) of the banknote transported on the banknote transport path 20 .

In addition, the banknote processing unit 4 detects the presence or absence of banknotes conveyed on the banknote conveyance path 20 and the inclination of the banknotes by a plurality of sensors arranged along the banknote conveyance path 20 . For example, this sensor is a transmissive optical sensor in which a light-emitting part and a light-receiving part are arranged facing each other across the banknote conveyance path 20 . The banknote transport path is driven by a motor (not shown) as a power source.

Next, a brief description will be given of the transport paths for depositing banknotes and withdrawing banknotes. Deposited banknotes are sent out one by one from the deposit and withdrawal banknote retention unit 24, and are transported to the banknote identification unit 25 along the banknote transport path 20 to identify the currency type and authenticity. The banknotes whose denomination has been recognized by the banknote identification unit 25 and are genuine are transported to the escrow unit 23 along the banknote transport path 20 . On the other hand, the banknotes whose denomination and authenticity fail to be identified by the banknote identification part 25 are sent back to the deposit and withdrawal banknote retention part 24 along the banknote conveying path 20 . The banknotes retained in the escrow unit 23 are sent out one by one from the escrow unit 23 after the transaction is completed, and sent to the banknote identification unit 25 along the transport path 20 for identification of currency type and authenticity again. Then, along the banknote conveyance path 20, it is conveyed to the banknote cassette 21 suitable for a denomination.

In addition, the withdrawn banknotes are fed out one after another from the banknote cassette 21 . The withdrawn banknotes fed out one after another from the banknote cassette 21 are sent to the banknote identification unit 25 along the banknote conveyance path 20 , and the currency type and authenticity are identified by the banknote identification unit 25 . The banknotes whose denominations have been identified by the banknote identification unit 25 and are genuine banknotes are delivered to the deposit and withdrawal banknote storage unit 24 along the banknote delivery path 20 , and other banknotes are delivered to the temporary storage unit 23 . Then, the banknote transport path 20 is reversed, and the banknotes in the escrow unit 23 (the banknotes whose denomination and authenticity failed through the banknote identification unit 25 ) are sent to the collection box 22 .

Next, the concrete structure of the banknote recognition part 25 is demonstrated. Fig. 3 is a block diagram showing a configuration of a main part of a banknote recognition unit. The banknote recognition unit 25 corresponds to the serial number recognition device of the present invention.

The banknote recognition unit 25 includes a control unit 30 , a ROM 31 , a RAM 32 , an imaging unit 33 , a denomination determination unit 34 , a cutting unit 35 , a selection unit 36 , an unnecessary area removal unit 37 , and a character recognition unit 38 .

The control part 30 controls the banknote recognition part 25 as a whole. ROM 31 is a non-volatile memory, and is equipped with a control program used by the control unit 30, a template image for distinguishing currency types (images of banknotes of a plurality of currency types processed on the ATM), and a serial number area for each currency type. position and size etc. The RAM 32 is a volatile memory, and stores variables used by the control unit 30 , color images captured by the imaging unit 33 , and the like. The imaging unit 33 captures a color image of paper sheets. The denomination discriminating unit 34 discriminates the denomination and orientation (direction) of the banknotes. The cutting unit 35 specifies the numbered area based on the information on the type and orientation of the banknote, and cuts out the color image of the numbered area from the color image of the entire paper sheet. The selection part 36 extracts the pixel of a character by selecting the pixel of the specific color determined based on the color information of each pixel and the type information of a banknote from the color image of a serial number area. The useless area removal unit 37 removes useless areas from the selected image. The character recognition unit 38 recognizes the numbered characters based on the image from which the characters are extracted.

Fig. 4 is an external view showing an example of a banknote serial number. Typically, banknotes have serial numbers printed on them in at least one place. On the banknote 40 shown in FIG. 4, the serial number 41 is printed at one place. In addition, as shown in FIG. 4 , the area including the number and the background surrounding the number is called a number area 42 . In addition, as shown in FIG. 4, the serial number which combined letters and numbers was illustrated, but it is not limited to letters and numbers, and symbols and characters which can identify a banknote, etc. can also be used.

Fig. 5 is a side view showing a state when an image of a banknote is captured by an imaging unit. The imaging unit 33 includes a lamp 330A and a lamp 330B that emit white light, and a line sensor (line sensor) 331A and a line sensor 331B that capture a color image. In FIG. 5, the banknote 40 is conveyed in the direction of the arrow 45 on the conveyance path 20 (not shown). The lamp 331 illuminates the entire upper surface of the banknote 40, and the light reflected from the banknote 40 is received by the line sensor 331A. The color image data output by the line sensor 331A is equivalent to one line. As the banknote 40 is transported, the line sensor 331A sequentially captures images of the banknote 40. By making such settings, a 2D image can be obtained as a result. In addition, the lower surface of the banknote 40 is photographed by the combination of the lamp 330B and the line sensor 331B, and a two-dimensional image is obtained in the same manner as above.

Here, the printing position and printing surface of the serial number on the banknote of each currency are determined. Japanese banknotes, for example, have serial numbers on one side and no serial number on the other. Therefore, the imaging unit 33 shown in FIG. 5 is configured to capture images of the upper and lower sides of the banknote 40 so that the banknote can be conveyed no matter which side faces upward. However, when dealing with foreign banknotes with serial numbers printed on both sides, it does not matter even if an image is captured on one side only. In addition, even if it is a processing device that handles paper with a serial number printed on only one side, such as Japanese banknotes, if the orientation of the paper can be fixed, it does not matter if the light and line sensor are installed on only one side. Also, it does not matter to use a 2D image sensor instead of a line sensor. In this case, the sheet being conveyed may be photographed in one to several times to obtain an image of the entire sheet.

Next, in the present invention, in order to correctly perform character recognition on the number, the number is extracted by the method described below.

1st pixel selection process

In the banknote recognition unit 25 , after the imaging unit 33 captures a color image of the banknote 40 , the denomination discriminating unit 34 reads out the template image from the ROM 31 and compares it with the captured image to determine the denomination of the banknote. Next, the cutting unit 35 reads the position information of the serial number from the ROM 31 based on the denomination information after the discrimination, and cuts out the serial number area corresponding to the position information from the color image of the banknote. At this time, as the first pixel selection process, the selection unit 36 can extract pixels of characters by selecting pixels of colors within a range determined for each denomination from the color image in the numbered area.

For example, for each pixel, it is checked whether or not the three luminances of R (red), G (green), and B (blue) are within a range preset for each denomination. The brightness of the selected pixel, if the red lower limit RL ≤ R brightness ≤ red upper limit RH, and the green lower limit GL ≤ G brightness ≤ green upper limit GH, and the blue lower limit BL ≤ B brightness ≤ upper limit value BH of blue, the original pixel value is substituted into the selected pixel value. On the other hand, if any of the luminances of the selected pixels is out of the range (outside the range), 255 (maximum value) is substituted into each selected pixel value, and white is set (replaced) as the background.

Also, when the brightness of the selected pixel is out of the range, the number may be converted to a color that is significantly different from the background color. For example, it may be converted to a color corresponding to the paper color (color of the medium) of the banknote, and when the paper color of the banknote is white, white is set as the background as described above.

A specific example of this processing is shown in FIG. 6 . Fig. 6 is a graph showing the extraction result of the serial number. FIG. 6(A) is an image of the numbered area 42 cut out by the cutting unit 35 . It was originally a color image, but based on the situation on paper, it is expressed in gray image here. The serial number "EL956293YC" is printed in red ink. Various lines are printed in water tone on the background around the numbers.

FIG. 6(B) is an image obtained by converting the color image shown in FIG. 6(A) into a gray image and then performing binarization. In this way, if the entire numbered area is binarized without selecting pixels, a part of the fourth character 9 from the right is cut off, and around the four characters "EL95" from the left, Lines and dots in the background other than numbered characters remain. Even if character recognition is performed on such an image, the accuracy is poor, and correct character recognition is impossible.

FIG. 6(C) is an image after performing pixel selection processing on FIG. 6(A). Since the character color is red, and since the character portion G luminance and B luminance are very small, the upper limit value of green GH and the upper limit value of blue BH are set to small values. On the other hand, since the background is based on water color, its B luminance is higher than the blue upper limit value BH, so it is converted to white. As a result of converting only the background to white, only character strings can be extracted and correct character recognition can be performed.

In addition, the pixels extracted as shown in FIG. 6(C) may be binarized as shown in FIG. 6(D). As a result, the contrast between the color of the character string and the background color becomes clear, so that the accuracy of character recognition improves.

As described above, when performing numbered character recognition for currencies with different character colors and background colors, the numbered characters can be correctly extracted by selecting pixels according to their brightness.

2nd pixel selection process

For the color image of the numbered area cut out by the cutting unit 35, as the selection process of the second pixel, the selection unit 36 performs an operation on a value obtained by linearly combining (linearly combining) a preset eigenvector and brightness, and by selecting the Character pixels can be extracted from pixels whose values are within a predetermined range for each currency.

For example, the predetermined eigenvector for judgment is set to (WR, WG, WB), and the luminance of the pixel is set to (RS, GS, BS), as a characteristic of the value obtained by linearly combining the judgment vector and the luminance of the pixel. The quantity is set to F, then F=WR*RS+WG*GS+WB*BS. It is confirmed whether or not this feature quantity F is within a predetermined range for each denomination. If the feature quantity F satisfies: threshold WL≤F≤threshold WH, the original pixel value is substituted into the selected pixel value. On the other hand, if any of the luminances of the selected pixels is out of the range, 255 (maximum value) is substituted into each selected pixel value, and white is set as the background.

Also, similar to the first pixel selection process, when the luminance of the selected pixel is out of the range, the number may be converted to a color that is significantly different from the background color.

A specific example of this processing is shown in FIG. 7 . FIG. 7 is a comparison explanatory diagram of two processing methods for selecting pixels. FIG. 7(A) is a first pixel selection processing method for selecting pixels whose luminance is within a specific range, and FIG. 7(B) is a second pixel selection processing method.

In FIG. 7(A), in the space where R and G are the axes, the pixels of the character portion are drawn by ○, the pixels of the pattern of the background color A are drawn by △, and the pixels of the pattern of the background color B are drawn by △. Note that the data shown in FIG. 7 is actually three-dimensional data of R, G, and B, but it is handled as two-dimensional data of R, G for simple explanation here. As shown in the figure, since the character color is similar to the background color A and the background color B, using thresholds RH, RL, GH, and GL, not only the numbered characters but also a part of the background are selected. That is, within the determination range (inside the rectangle) surrounded by the upper limit value RH of red, the lower limit value RL of red, the upper limit value GH of green, and the lower limit value GL of green, it is determined to be a character color pixel. , but a part of the pixel group of the background color A is also included in this judgment range (inside the rectangle), and these pixels are the background and are also judged as the pixels constituting the character. In addition, a part of the pixel group of the background color B is also included in this determination range (inside the rectangle), and these pixels are background and are also determined as pixels constituting a character. Therefore, although the characters can be recognized, the recognition accuracy is reduced.

On the other hand, in the present processing method shown in FIG. 7(B), instead of projecting to the RG space, it is projected to the axis linearly combined with the pixel brightness and the eigenvector (WR0, WG0), and the eigenvector ( In the space represented by the axes after the linear combination of WR1, WG1), the feature quantity F0 is, F0=WR0*RS+WG0*GS, and the feature quantity F1 is, F1=WR1*RS+WG1*GS. At this time, within the discrimination range (in the rectangle) surrounded by threshold WH0, threshold WL0, threshold WH1, and threshold WL1, pixels of background color A and background color B are not included, and only character pixels are included. Therefore, with this processing method, only the pixels constituting the characters of the serial number can be extracted.

In order to extract only the characters of the serial number by this method, a plurality of data of banknotes obtained by sampling in advance for each denomination can be obtained by carrying out an experiment, etc., and an intrinsic vector and a threshold value can be appropriately set. In this way, the extraction accuracy of characters can be improved.

Also, depending on the type of paper, when the background is drawn in a plurality of colors, it is also possible to separate the character colors for the plurality of colors.

Useless area removal processing

By executing the first pixel selection process or the second pixel selection process as described above, the characters of the numbers can be extracted. However, when the character color of the number is the same color as the background color of the number area, the pattern of the background is also extracted. Therefore, in this case, after the first pixel selection process or the second pixel selection process, the useless area removal process described below may be performed.

The unnecessary region removal unit 37 shown in FIG. 3 includes a median filter, and performs unnecessary region removal (transformation) processing on each pixel selected by the selection unit 36 through the median filter. As we all know, the median filter is a method of replacing the value of the pixel of interest with the value of the pixel of interest by sorting (sorting) the values (such as luminance) assigned to n×n pixels composed of the pixel of interest and its surrounding pixels. , by which unwanted noise components such as background patterns can be removed (removed).

FIG. 8 shows a specific example of this processing. FIG. 8 is a diagram of numbered useless area removal processing results. FIG. 8(A) is a part of pixels in the numbered area cut out by the cutout unit 35 . It was originally a color image, but based on the situation on paper, it is expressed in gray image here. The number "AL8" is printed in red ink. In addition, the background around the number is the same color, which is a bright red tone.

FIG. 8(B) is an image of a result of pixel selection processing performed on FIG. 8(A). Since the character color and the background color are in the same color system as described above, only the character part cannot be extracted even if the pixel selection process is performed, and the background pattern is also extracted. In particular, patterns other than characters remained around the character "A".

FIG. 8(C) is an image after unnecessary region removal processing is performed on FIG. 8(B). After image processing with a 5×5 median filter, the image shown in FIG. 8(C) and the small speckle-like pattern seen in FIG. 8(B) disappeared.

FIG. 8(D) is an image obtained by binarizing the image shown in FIG. 8(C). After the image is processed by the median filter, the image is binarized to make the contrast between the numbered string color and the background color clearer.

In this way, even if the numbers and character colors are similar to the background color, only characters can be extracted by performing removal processing of useless areas after selection processing.

Next, in ATM1, in order to perform the 1st pixel selection process, the 2nd pixel selection process, and useless area removal process, the table is memorize|stored for every denomination of banknotes in ROM31. 9 is an example of a table in which data of the first pixel selection process, the second pixel selection process, and the unnecessary area removal process are recorded.

FIG. 9(A) is an example of a table for the first pixel selection process. In FIG. 9(A), the direction is the direction of the banknote conveyed on the conveyance path. If the side facing the conveying direction is referred to as "up", the banknotes can be directed in four directions: left side up, right side up, top side up, and bottom side up. Call them Direction 0, Direction 1, Direction 2, and Direction 3, respectively. In addition, the table records: the area start X coordinate, the area start Y coordinate, the area size X coordinate, the area size Y coordinate used for cutting out each currency and each direction, and the selection used for each currency. R (red), G (green), B (blue) brightness upper limit and lower limit (R upper limit, R lower limit, G upper limit, G lower limit, B upper limit, B lower limit).

In addition, FIG. 9(B) is an example of a data table for the second pixel selection process. Only the parts different from those in Fig. 9(A) will be described. The number of determinations is the number of determination expressions used for determination with respect to one pixel. In the example shown in FIG. 9(B), two are set. Coefficient group 0 is an eigenvector (WR0, WG0, WB0) used in the first decision expression. The characteristic value 0 is obtained by linear combination of these eigenvectors and the luminance of the pixel. WH0 and WL0 are threshold values (upper limit value and lower limit value) for judging feature amount 0. FIG. Similarly, (WR1, WG1, WB1) are eigenvectors used to obtain feature value 1, and WH1, WL1 are thresholds (upper limit and lower limit) for determining feature value 1.

FIG. 9(C) is an example of a table for useless area removal processing. Only the parts different from those in Fig. 9(A) will be described. Median filter usage flag, set whether to execute median filter processing. In addition, in the processing of the median filter, different parameters may be set for each coin type. For example, it is possible to set the size of the filter (3×3, 5×5, etc.), and select the Nth value from the front instead of the middle value instead of the middle value to replace the parameters corresponding to the characteristics of the banknote.

Next, the operation of the recognition section of the ATM 1 will be described. Fig. 10 is a flowchart for explaining the operation of the banknote recognition unit. FIG. 11 is a flowchart for explaining first pixel selection processing. FIG. 12 is a flowchart for explaining the second pixel selection process.

The banknotes are sent out one by one from the banknote deposit and withdrawal port 4a of ATM1, and after being sent to the banknote recognition unit 25, the control unit 30 detects the arrival of paper by the paper detection sensor (figure omitted), and the camera unit 33 (lamp light) detects the arrival of the paper. 330A, lamp 330B, line sensor 331A, and line sensor 331B) take a color image of the banknote 40 (s1).

Next, the control unit 30 outputs the color image captured by the imaging unit 33 to the denomination discriminating unit 34 to discriminate the type and orientation (direction) of the sheet and acquire these information (s2). The control unit 30 refers to the table stored in the ROM 31, reads out the information on the starting position and size of the numbering area 42 corresponding to the acquired currency and direction information, and cuts out the numbering area from the color image by the cutting unit 35 (s3) .

The control unit 30 focuses on the color component of the character part from the image of the numbered area cut out by the selection unit 36, and selects and extracts the character part (s4). Specifically, when the upper and lower limits of R, G, and B are recorded in the table referred to in step s3 as shown in FIG. 9(A), the control unit 30 performs the first pixel selection as shown in FIG. deal with. On the other hand, when the upper limit value and the lower limit value of the eigenvector and the feature value are recorded in the table referred to in step s3 as shown in FIG. 9(B), the control unit 30 performs the second pixel selection as shown in FIG. 12 deal with.

When performing the first pixel selection process, the following steps are performed. As shown in FIG. 11 , in the selection process, first, the R luminance, G luminance, and B luminance of one pixel are extracted from the color image of the numbered area that has been cut out (s11). Next, it is checked whether it is within the range of the threshold values (RH, RL, GH, GL, BH, BL) read from the table (s12).

If RL≤Rluminance≤RH, GL≤Gluminance≤GH, and BL≤Bluminance≤BH (s12: Yes), the original pixel value is substituted into the selected pixel value (s13). On the other hand, if the above condition is not satisfied (s12: No), 255 (maximum value) is substituted into each of the selected pixels to set white as the background.

Then, it is checked whether all the pixels have been processed (s15), and if not, it goes to the processing of the next pixel (s11), and if it is completed, the selection process is terminated.

In addition, when performing the second pixel selection process, the following procedure is performed. As shown in FIG. 12, first, the R luminance, G luminance, and B luminance of one pixel are extracted from the color image of the cut-out numbered area (s21). Next, linearly combine the extracted luminance with the eigenvector read from the table to obtain the feature value F=WR*RS+WG*GS+WB*BS (s22), and perform the judgment formula WL on this feature value F Determination of ≤F≤WH (s23). Check whether the judgment of all the judgment expressions is completed (s24), if not completed (s24: No), then find the next feature quantity (s21), if all judgments are completed (s24: Yes), then check whether all the judgment expressions are true (s25). If all the judgment expressions are true (s25: YES), the original pixel value is substituted into the selected pixel value (s26). On the other hand, if even one of the judgment expressions is false (s25: No), 255 (maximum value) is substituted into each of the selected pixel values, and white is set as the background (s27).

Then, it is checked whether all the pixels have been processed (s28), and if not (s28: No), go to the processing of the next pixel (s21), and if it is completed (s28: Yes), the selection process ends.

Also, although the selected pixel values are also processed as a color image, it does not matter if it is converted into a gray image or a binary image. In other words, it may be determined according to what kind of image expression processing is to be used in the post-processing.

As shown in the flow chart shown in Figure 10, after the processing of step s4 ends, then, the control unit 30 will refer to the table of each currency set on the ROM21 to determine whether it is necessary to perform useless area removal processing (s5 ).

When the median filter use flag of the referenced table is "Use" (s5: Yes), the control unit 30 performs unnecessary area removal processing (s6), and when it is "Not used" (s5: No), it does not perform unnecessary area removal processing (s5: No). The processing is removed and the next step s7 is executed.

The control unit 30 performs character recognition by the character recognition unit 38 on the numbered character string obtained through the above-mentioned respective image processes (s7).

In addition, there are various well-known methods for character recognition, and any method may be used, and detailed description thereof will be omitted here.

To sum up, in the present invention, when recognizing the serial number, the area where the banknote serial number is displayed is cut out, and the background color or pattern of this area is eliminated, and only the image of the serial number in this area is extracted. Therefore, character recognition can be performed reliably. In addition, genuine and counterfeit bills can be easily discriminated based on the serial number.

In addition, in the above description, the case of using banknotes as an example of paper is described, but the present invention is not limited to this, and other papers marked with numbers such as lottery tickets (くじ), tickets, securities, etc. are used. ,sure.

Claims (12)

1. numbering recognition device, identification are marked on the numbering on the paper class that transmits along transfer path, it is characterized in that,

Possess:

Image mechanism is taken the coloured image of above-mentioned paper class;

Mechanism for identifying, the coloured image of taking based on above-mentioned image mechanism is differentiated the kind and the direction of above-mentioned paper class;

Selection mechanism, the kind and the information of direction of the paper class of differentiating based on above-mentioned mechanism for identifying confirm to be marked with the numbering area of above-mentioned numbering, the color pixel that determines based on the kind of information of above-mentioned paper class of selection only from this regional image; And

Character recognition mechanism, the image that the pixel of being selected by above-mentioned selection mechanism is constituted carries out character recognition.

2. numbering recognition device as claimed in claim 1 is characterized in that:

For each pixel, if based on to the numerical value of R that this pixel obtained, G, each brightness of B in threshold range, this pixel is selected in then above-mentioned selection mechanism from the image of above-mentioned numbering area.

3. numbering recognition device as claimed in claim 2 is characterized in that:

For each pixel; The inherent vector linear combination that calculating will be set respectively to R that this pixel obtained, G, each brightness of B with by each kind of paper class and the value that obtains; If should be worth in threshold range, this pixel is selected in then above-mentioned selection mechanism from the image of above-mentioned numbering area.

4. numbering recognition device as claimed in claim 2 is characterized in that:

If beyond above-mentioned threshold range, then above-mentioned selection mechanism is replaced as predefined other brightness with this brightness based on the above-mentioned numerical value of R, G, each brightness of B.

5. numbering recognition device as claimed in claim 3 is characterized in that:

If beyond above-mentioned threshold range, then above-mentioned selection mechanism is replaced as predefined other brightness with this brightness based on the above-mentioned numerical value of R, G, each brightness of B.

6. numbering recognition device as claimed in claim 1 is characterized in that:

Possess the useless region removal mechanism, this useless region removal mechanism is carried out Flame Image Process for the image of being selected by above-mentioned selection mechanism that pixel constituted through median filter, removes unwanted pixel,

Above-mentioned character recognition mechanism, the image that the useless region removal mechanism has been removed after the unwanted pixel carries out character recognition.

7. numbering recognition device as claimed in claim 2 is characterized in that:

Possess the useless region removal mechanism, this useless region removal mechanism is carried out Flame Image Process for the image of being selected by above-mentioned selection mechanism that pixel constituted through median filter, removes unwanted pixel,

Above-mentioned character recognition mechanism, the image that the useless region removal mechanism has been removed after the unwanted pixel carries out character recognition.

8. numbering recognition device as claimed in claim 3 is characterized in that:

Possess the useless region removal mechanism, this useless region removal mechanism is carried out Flame Image Process for the image of being selected by above-mentioned selection mechanism that pixel constituted through median filter, removes unwanted pixel,

Above-mentioned character recognition mechanism, the image that the useless region removal mechanism has been removed after the unwanted pixel carries out character recognition.

9. numbering recognition device as claimed in claim 4 is characterized in that:

Possess the useless region removal mechanism, this useless region removal mechanism is carried out Flame Image Process for the image of being selected by above-mentioned selection mechanism that pixel constituted through median filter, removes unwanted pixel,

Above-mentioned character recognition mechanism, the image that the useless region removal mechanism has been removed after the unwanted pixel carries out character recognition.

10. bank note treatment device in the going out bank note depositing port and be arranged between the paper class receiver of body interior of this bank note treatment device of the body front that is arranged at this bank note treatment device, transmits the paper class that deposits taking-up in, it is characterized in that,

Possess:

Image mechanism, the coloured image of the numbered paper class of shot mark;

Mechanism for identifying, the coloured image of taking based on above-mentioned image mechanism is differentiated the kind and the direction of paper class;

Selection mechanism, the kind and the information of direction of the paper class of differentiating based on above-mentioned mechanism for identifying confirm to be marked with the numbering area of above-mentioned numbering, the color pixel that determines based on the kind of information of above-mentioned paper class of selection only from this regional image; And

Character recognition mechanism, the image that the pixel of being selected by above-mentioned selection mechanism is constituted carries out character recognition.

11. an automated transaction is handled device; Be arranged at this automated transaction handle device the body front go out bank note depositing port and be arranged at this automated transaction to handle between the paper class receiver of body interior of device; Paper class to depositing taking-up in transmits; And to deposit and withdraw the transaction handle, it is characterized in that

Possess:

Image mechanism, the coloured image of the numbered paper class of shot mark;

Mechanism for identifying, the coloured image of taking based on above-mentioned image mechanism is differentiated the kind and the direction of paper class;

Selection mechanism, the kind and the information of direction of the paper class of differentiating based on above-mentioned mechanism for identifying confirm to be marked with the numbering area of above-mentioned numbering, the color pixel that determines based on the kind of information of above-mentioned paper class of selection only from this regional image; And

Character recognition mechanism, the image that the pixel of being selected by above-mentioned selection mechanism is constituted carries out character recognition.

12. numbering recognition methods is characterized in that,

Possess:

Take step, the image mechanism shot mark coloured image of paper class of numbering;

Discriminating step, the coloured image that mechanism for identifying is taken based on above-mentioned image mechanism is differentiated the kind and the direction of above-mentioned paper class;

Select step; Selection mechanism is based on the kind of the paper class of above-mentioned mechanism for identifying differentiation and the information of direction; Confirm to be marked with the numbering area of the numbering of paper class, from this regional image, only select the color pixel that determines based on the kind of information of above-mentioned paper class; And

Character recognition step, character recognition mechanism carries out character recognition to the numbering image that the pixel of being selected by above-mentioned selection mechanism constitutes.

Images