JPH10154256A - Device and method for recognizing symbol and number of paper sheets - Google Patents

Abstract

PROBLEM TO BE SOLVED: To surely recognize even when the deviated amount is large, by using images of 1st and 2nd sensors and a parameter for identification to perform an identification operation. SOLUTION: A paper money 1 is carried on the path below a light emitting and receiving module, a line sensor 3 takes in an image for denomination identification, and a line sensor 4 takes in the image of an area that is preliminary defined for symbol and number identification. On the path for the paper money 1, a passage sensor 7 which detects passage and an authenticity sensor 8 which detects the authenticity of the paper money 1 are arranged, and a passage signal PS is inputted to an image processing deciding part 10 and an identification device controlling part 20. A detection signal from the sensor 8 is also inputted to the part 20 through an A/D converter 53. The part 20 is connected to the part 10 and a body controlling part 30. The part 10 compares the reference pattern of flash memory for a reference code pattern with the data of an identified paper money and outputs a decision result.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a serial number of a paper sheet such as a bill, a check, etc. (a number which is assigned as a different serial number at the time of printing). The present invention relates to a number recognition device and method.

[0002]

2. Description of the Related Art Normally, a different serial number is printed on each banknote, but the same serial number is printed on counterfeit banknotes.
It is often printed in large quantities and used in large areas. For this reason, it is desired to store the serial number of the counterfeit banknote and promptly report when the serial number is identified.

A device for recognizing denomination characters of banknotes is disclosed in, for example, Japanese Patent Application Laid-Open No. 1-281591. The recognition device disclosed in this publication is configured to determine a denomination, a front and back, and a transport direction of a banknote by recognizing a specific denomination character. Because it is not the recognition of the serial number, it cannot provide sufficient protection against counterfeit banknotes. However, the reference position data of the area including the denomination character, the density distribution data and the outline data of the reference area are set in advance, and the character recognition is performed in the area where the denomination character fits. In this recognition method, first, data is binarized by applying a filter to the data in this area, and a density distribution in a vertical direction and a horizontal direction is created for the binarized data. It is supposed to do. When recognizing denomination characters, it is only necessary to be able to identify the characters "1,000", "five", and "million", but if you try to recognize the serial number printed when a bill is issued , 26 letters of the alphabet and 10 numbers need to be identified, and the above-mentioned conventional method or general-purpose character recognition method cannot obtain a highly accurate recognition and high efficiency. In other words, there is a disadvantage that the print position is affected by fluctuations such as a print position shift, missing characters, and the like, resulting in incorrect recognition.

Japanese Patent Application Laid-Open No. 4-107797 discloses a device for storing banknote numbers and identifying counterfeit banknotes by detecting the banknote numbers.
Japanese Patent Application Laid-Open No. 7-1921 as a recognition device for identifying bill numbers
No. 61, there is a description for the purpose of recognizing a serial number printed on a bill or the like.

[0005]

The serial number of a paper sheet can be read to separate and recover old tickets. If the automatic depositing machine is used, the serial number can be read and stored. It is possible to identify who deposited which banknote, and to identify the depositor when a counterfeit banknote is discovered. Furthermore, when a banknote with a known serial number is received, by specifying the depositor, it is also useful to identify a criminal in the event of a cash-related incident such as a theft incident.

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a printing apparatus which has a large amount of misalignment even for paper sheets having a large amount of misalignment (for example, US dollar bills). An object of the present invention is to provide a serial number recognizing device capable of surely recognizing a serial number even if there is any. In addition, a method for easily learning serial number recognition is provided. By reliably recognizing the serial number, sufficient protection against counterfeit bills and the like can be realized. After determining the type of paper sheet and the transport direction, the position where the serial number image is located is searched to determine the search start character position, and then the serial number recognition is performed, thereby performing the serial number character recognition with high accuracy. obtain. further,
By learning a large number of paper sheets, it is possible to realize recognition of serial numbers which are not affected by fluctuations such as positional deviation of paper sheets and missing characters.

[0007]

SUMMARY OF THE INVENTION According to the present invention, a serial number of a paper sheet in a paper sheet discriminator for processing at least the kind and direction of the paper sheet by processing optical image image data of the paper sheet. An object of the present invention is to provide a serial number recognition device for identifying a serial number, a first sensor for reading an image of the sheet, a second sensor for reading a serial number in the image of the sheet, Non-volatile memory for storing parameters for use, and overall control in cooperation with ROM and RAM, and identification operation for identifying the paper sheet with the image data of the first sensor and the second sensor and the parameter This is achieved by providing

Further, the present invention can be more effectively achieved by providing a general-purpose personal computer which is connected to the discriminating operation section via an interface and has an external interface and learns the discriminating operation. That is, in the sheet serial number recognition device, after the type and direction of the sheet conveyed based on the image data by the first sensor are determined, a denomination and a denomination set in advance from the image data are determined. The coordinates are converted in the x-axis direction and the y-axis direction in the image area by the second sensor based on the information of the print character search area having the serial number different depending on the direction, and the similarity is calculated each time. , And a similarity greater than a predetermined value is obtained, and the similarity with all the characters to be identified is calculated, and the first recognition candidate character and the second similarity are calculated. When there is a predetermined difference in similarity between the candidate character having the degree and the degree of similarity, the first recognition candidate character is determined as the identification result. Further, the number characters of the paper sheet are learned, and the image cut out for calculating the similarity is binarized from 1 and −1 by a predetermined threshold to obtain a first weight map, and the image to be learned is obtained. The sign of the binarized pixel value is compared with the sign of each pixel of the created weight map, and only the portion having a different sign is added and subtracted, and the addition and subtraction by the comparison are repeated until a predetermined evaluation value is obtained. The similarity is calculated as a second candidate with the character most similar to the corresponding character from the data. If the similarity difference between the learning character and the second candidate is not more than a predetermined value, the weight of the second candidate is calculated. Learning is performed by manipulating the weight of the coordinates so that the difference between pixels having a different sign from the map is increased.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 shows an example of a serial number recognizing device according to the present invention. A bill 1 includes a light emitting means 2 comprising a light emitting diode array or the like, a line sensor 3 as a light receiving means for receiving reflected light, and 4 is conveyed along the lower surface passage of the light emitting / receiving module integrated with the light emitting / receiving module. Video signals VSA1 and VSA2 from line sensors 3 and 4
Are converted into digital video signals VSB1 and VSB2 by A / D converters 51 and 52, respectively, and input to the image processing determination unit 10. The line sensor 3 captures an image for denomination identification, and the line sensor 4 converts an image of a predetermined area for serial number identification to a higher resolution (for example, 0.25 mm).
× 0.25 mm / dot).
In addition, the relationship between the banknote 1 and the serial number is as shown in FIG. 2, the serial numbers 1A and 1B are printed at a plurality of places on the banknote 1, and the line sensor 4 indicates the serial number (1A or 1B). It is designed to capture in high resolution. Image processing determination unit 10
3 is shown in FIG. 3. In the image processing determination unit 10, the video signals VSB1 and VSB2 are stored in FIFO memories 111 and 112, respectively.
DSP (Digital Signal Procedures)
(sor) 100 and sequentially transferred to a selected area of the main memory 12 through the correction unit 101 and written. On the main memory 12, image data for one bill is developed. Then, the denomination, direction, and authenticity of the banknote 1 are determined based on the image data of the entire banknote and data from another authenticity sensor 8. The image data of the serial number part is
Only the corresponding part of 1A or 1B is the main memory (RAM)
Expanded to 12. The DSP 100 sends a main memory (double buffer) 12 from the ROM 110 to increase the processing speed of the control program stored in the ROM 110.
In cooperation with the ones copied on the FIFO (First-in
First-out) The skew correction / block compression / encoding after performing skew correction on digital image data of the video signals VSB1 and VSB2 input via the memories 111 and 112 and the correction unit 101. A block compression encoder 102, a comparison / determination controller 103 that outputs a determination result DR, and a binarizer 1 that binarizes image data BD on which skew correction / blocking has been performed.
04 and a serial number recognition control unit 105. DS
P100 is connected to personal computer 40 via dual port RAM17.

The image processing determination unit 10 stores a reference code pattern (data for a denomination, a direction, authenticity determination and denomination, and a search area corresponding to a direction) of various bills. Code pattern flash memory (ROM) 13
The comparison code controller 103 compares the reference code pattern RC with the data CS of the banknote to be identified from the main memory 12 and outputs a judgment result DR.
Communication is performed with the discriminator control unit 20 that controls the discriminator via the control unit 14. The flash memory 13 is an electrically rewritable read-only memory, and the main memory 1
Reference numeral 2 denotes a RAM that functions as a double buffer and has a memory for image data, a memory for a work area, and the like. In order to store a weight map for serial number identification, a flash memory 16 for serial number recognition / weight map storage is prepared.

Further, the image processing determination unit 10 receives a mechanical clock signal ES of the rotary encoder 6 for controlling lighting / flashing of the light-emitting means 2 and determining a scanning interval of the line sensor 3 during bill conveyance, and receives an A / A signal. D converters 51 and 5
2 and a read control unit 15 that performs read control of the FIFO memories 111 and 112 and generates read control timing RT of the line sensors 3 and 4.
A pass sensor 7 for detecting the passage of the bill 1 and a true / false sensor 8 for detecting the authenticity of the bill 1 are provided on the transport path of the bill 1. The pass signal PS from the pass sensor 7 is input to the reading control unit 15 in the image processing determination unit 10 and also to the discriminator control unit 20. The detection signal from the true / false sensor 8 also passes through the A / D converter 53, and the discriminator control unit 20
And the discriminator control unit 20 is connected to the image processing determination unit 10 and the main unit control unit 3 such as a banknote deposit processing device.
Connected to 0. In FIG. 3, for convenience of description, components used for authenticity identification are omitted.

FIG. 4 shows the image processing determining unit 1 shown in FIGS.
It is a flowchart showing an operation example of the DSP 100 in 0, waits for reception of a start command from the main body control unit 30 (step S1), and when the start command is received, sets the initial values for correction of the line sensors 3 and 4 to Read (step S2). First, in order to correct the video signal VSA1 from the line sensor 3, black level data, which is so-called dark output data, such as reading the output of the line sensor 3 with the light emitting unit 2 turned off, is collected, and then the light emitting unit 2 is turned on. (Step S3). The read control unit 15 returns a ready response to the main body control unit 30 and then turns on the light emitting unit 2, writes an identification preparation completion response in the dual port RAM 14, generates an interrupt, and notifies the host CPU of the response. Execute (step S4). Then, it waits for the bill 1 to be conveyed. The passage of the bill 1 is determined based on the passage sensor 7 (step S5). If there is no passage of the bill 1, it is determined whether or not there is an end command (step S6). (Step S7), and the process returns to step S5 if there is no end command. When the passage of the bill 1 is detected in step S5, image data is acquired from the line sensors 3 and 4 via the A / D converters 51 and 52, respectively (step S5).
100), denomination identification is executed (step S10). It is determined whether or not the banknote should be rejected according to the result of the denomination identification (step S11). If the banknote is not a rejected banknote, a serial number recognition process is executed (step S200), and the denomination identification and the serial number recognition result are compared and determined. The control unit 103 sends the identification result DR to the main body control unit 30 (step S13). This transmission is performed by writing the result in the dual port RAM 14, generating a response interrupt and notifying the host CPU. Then, it waits for the next bill to be conveyed. If it is determined in step S11 that the banknote is a reject banknote, the comparison / determination control unit 103 sends a reject code as the identification result DR (step S12).

FIG. 5 is a flowchart showing an operation example of the details of the image capture of one banknote in step S100. First, when the passage sensor 7 detects passage of the bill 1, a passage signal PS is input to the reading control unit 15, and the video signal VSA2 from the line sensor 3 is converted to an A / D converter 5
2 to convert analog to digital value and FIFO
Write to the memory 111. Thereafter, a predetermined correction is performed by the correction unit 101 in the DSP 100, and the result is written into one double buffer of the main memory 12. The line sensor 3 uses the black level data fetched and processed at the start of the identification and the white level data and the black level data written in the flash memory 13 by executing the setting program in advance, and uses the correction unit 101. The image data is collected while performing the correction in (Step S10)
1). Then, when the data for one bill is completed, the double buffer of the main memory 12 is switched (step S102). This switching permission is executed by enabling the passage sensor interrupt, thereby entering a data collection standby state for the next bill (step S103).
In order to increase the processing speed, the configuration is such that recognition processing can be performed while image data is being captured. Next, the skew feeding / correction is performed by the skew feeding correction / blocking compression encoding unit 102 (step S104). Here, the image data of the coarse image of the entire bill and the image area of the high-resolution image area of the serial number portion are corrected. This completes the data acquisition subroutine and returns to the called place.

Next, the serial number recognition processing in step S200 will be described with reference to the flowchart of FIG. Here, this serial number identification is possible even if there is a printing shift of the character by searching for the character position.In recognition, the automatic weight learning function of the character enables recognition of similar characters and recognition of a missing character. ing. In addition, since the skew / adjustment has already been corrected, it is possible to identify the skew that has occurred in the denomination discriminating unit without being affected by the conveyance state such as the skew / width adjustment.

First, the maximum value and the minimum value of the pixel value of a predetermined area are obtained, the intermediate value is binarized as a threshold value (step S201), and the first character search start position is determined based on denomination and direction information. A preset search position (x, y) is set (step S202). The binarization of the serial number is performed as shown in FIG. 7. Since the character interval of the serial number is constant, a character search is performed by adding coordinates in the x direction by the interval. Then, as in the algorithm shown in FIG. 8, the product of the weight W determined by learning and the value img (binary: −1, 1 (0 is not used)) of each coordinate is used as the value of each coordinate. The ratio between the sum and the sum of the absolute values of all the weights is determined according to Equation 1 and used as the similarity to the character of each number. Then, the similarity is calculated for all the candidate characters, the first candidate character having the maximum similarity χ1 is specified, and the second candidate character having the next semi-maximum similarity 準 2 is specified. (Step S203).

FIG. 8 is a diagram for explaining the relationship between the weights Wj, [i] between the image data in the area of the character to be recognized and each character as a recognition candidate. That is, the weight W used when obtaining the similarity χ [j] in Expression 1 is present between the character serving as the identification result and j as each identification result.
This indicates that there are all the pixel numbers i for the th character. In the present embodiment, since there are 144 i, there are 144 weights for one character.
This shows that the pixel value img [i] of the character to be identified calculates the similarity χj using the weight Wj, [i] written in the flash memory 16.

[0018]

(Equation 1) Further, if the value of the similarity χ1 with the first candidate is equal to or greater than a predetermined value and the difference from the similarity χ2 with the second candidate is equal to or greater than the predetermined value, the corresponding number is recognized. That is, it is determined whether or not the first candidate # 1 is larger than a predetermined value θ (step S204). If the first candidate # 1 is larger than the predetermined value θ, the difference between the first candidate # 1 and the second candidate # 2 is larger than a predetermined value ε. It is determined whether or not the character is the predetermined character (step S205). If the value is larger than the predetermined value ε, the recognition result (the character of the first candidate # 1) is determined (step S206), and it is checked whether the character is the last character (step S215). If it is not the last character, the next character position is set (step S216), and the process returns to step S203. Step S2 above
If it is smaller than the predetermined value ε in 05, there is a high possibility of erroneous recognition, and the process jumps to the determination of non-recognition (step 214). Also, in step S204, the first
When the candidate # 1 is equal to or smaller than the predetermined value θ, x and y are changed to search for the character position. That is, it is advanced by "+1" in the x-axis direction (step S210), and it is checked whether the x coordinate has reached the right end of the area (step S211). If it is not the right end, the process returns to step S203. Advance by “+1” to return the x-axis value to the initial value (step S2)
12). Here, if y has not reached the maximum value (step S213), the process returns to step S203, and the calculation and evaluation of the similarity are repeated. If y is the maximum value (step S213), it is determined that recognition is not possible (step S21).
4). Thus, it is determined whether or not the search range is a predetermined search range. If y is within the range of the maximum value, the above-described step S20 is performed.
The process returns to step S3. If the character is out of the range, it is determined that recognition is not possible (step S214), and it is determined whether or not the last character is recognized (step S215). Skip to step S216.

Next, a serial number learning process using an external personal computer will be described with reference to the flowcharts of FIGS.

First, in the process shown in FIG. 9, the weight of the serial number recognition is automatically learned, but the following (1 or 1) binarized serial number image output from the serial number identifying device is obtained as follows. Learning, and a weight map is obtained by learning. If the character with the serial number j is to be learned, i indicates the position of each image pixel. In this example, it is from “1” to “144” (see FIG. 7), and img [i] is the binarized pixel i.
Indicates the second value. When learning of multiple images is repeatedly executed by the following algorithm, the weight of a pixel in which a large amount of variation such as missing or no missing of a character image occurs decreases,
The weight of a pixel that does not fluctuate has a large value (the weight of a pixel that always has a character has a larger number in the plus direction, and the weight of a pixel that does not always have a character has a larger number in the minus direction). As a result, even a missing character can be recognized. In other words, the weight of the portion of the pixel that is a missing character is small, so that it is not affected.
In recognition of a similar character (second candidate), the weight of a pixel (feature pixel) having a sign different from that of a pixel value (binary) is increased, and as a result, a large similarity difference is obtained. It becomes possible.

First, the learning mode is externally set.
When the personal computer 40 is connected as shown in FIGS. 1 and 3, the learning mode may be input according to an instruction from the personal computer 40. First, the process waits for a start command as a trigger for determining whether to start the process (step S40).
1) When the start command is received, the initial data is read and stored by the line sensors 3 and 4 (step S4).
02), the light emitting means 2 is turned on (step S403).
When the preparation is completed, a response is returned (step S404), and the banknote 1 is ready to be conveyed. Waiting for a bill (step S405), if a bill arrives, fetches image data for one bill into the main memory 12 (step S100), and identifies the denomination direction prior to learning the serial number (step S410). ). Although the denomination is not particularly described, the denomination can be sufficiently identified by the known technique. As a result of the denomination identification, if the denomination identification result is unknown, a reject code is sent to perform rejection (steps S411 and S412).
If it is not a reject, the serial number part image data identification result is sent (step S413), and the denomination identification result /
The serial number part image data is transmitted (step S41).
4) Turn off the light emitting means 2 (step S415). Thereafter, the process waits for a start command from the personal computer 40 (step S401).

Thereafter, learning of the serial number is started on the personal computer 40 side based on the image data of the collected bills. This will be described with reference to the flowchart of FIG.

The image img [i] is binarized (step S
301), the above-mentioned step S40 is manually performed by the personal computer 40.
The corresponding character data of the learning character number j and its character position selected at 0 and input is read (step S3).
02), it is determined whether it is the first learning (step S30).
3) For the first learning, the weight Wj, [i] (i
Is the pixel position from 1 to 144, and j indicates the character number. Therefore, the weight of the character number j is cleared by 144 for the character number j specified in step S302) (step S304). If the learning is not the learning, the weight Wj, [i] is read from the flash memory 16 (step S305), and as shown in FIG. If the pixel is a non-character pixel, the weight value is set to "-1" (step S306). FIG.
(A) shows the first learning of the character "5". In the first learning, the pixel [i] is based on the image data in the left diagram, and the img [i] with a character is "1" and the character is absent. Img [i]
Becomes "-1". FIG. 11B shows the second learning of the character "5". When the image data of the partially missing portion in the left diagram is learned, W33,
[I] = W33, [i] + img [i] = W33,
[I] +1, and W33, [i] at coordinates i without characters
= W33, [i] + img [i] = W33, [i] -1
Becomes

Then, the similarity χj of the learning target character is calculated from the above equation 1 (step S307), and a predetermined value is set as a threshold θ, and it is determined whether or not the similarity χj is smaller than the threshold θ (step S310). ),
If it is smaller than the threshold θ, the process proceeds to step S311.

In step S311, weight coordinates (sig) having a sign different from the current learning weight of the learning target character are used.
coordinates i such that n (img [i]) ≠ sign (Wj, [i])
Is determined as dif Wj, [i], and it is determined whether or not it exists, that is, whether or not Wj, [i] ≧ 0 (step S312). Update the weight according to 3.

[0026]

## EQU2 ## if (２j <θ) χif (dif Wj, [i]
≦ 0): dif W [i], j + 1; If the similarity χj is smaller than the threshold θ (Yes in step S310), step S31
If the weight Wj, [i] of the corresponding coordinate i is negative or 0 in step 2, "1" is added to make a new weight Wj, [i] (step S314).

[0027]

## EQU3 ## if (＜ j <θ) （if (dif Wj, [i]
> 0): dif W [i], j−1; conversely, if the similarity χj is smaller than the threshold θ (Yes in step S310), and if the weight Wj, [i] of the corresponding coordinate i is positive in step S312, For example, "1" is subtracted to obtain a new weight Wj, [i] (step 31).
3).

[0028]

If (χj ≧ θ): end Then, the similarity χj is calculated (step S307). If the similarity χj becomes larger than the threshold θ (step S310), the loop is ended (equation 4). ).

Although the weights are partially overlapped, the weight updating by the above equations (2) and (3) is as shown in FIG. 12, and the similarity χj of the n-th learning target character is χj <θ, and the lower row of the lower row showing the updated state. The shaded area in the figure indicates the updated weight. That is, only the weight of the hatched portion is operated so as to approach 0 in the portion where the hatched portion in FIG. 12 (B) is different from the sign of the weight before update in FIG. 12 (A). Such an update is performed only for a pixel i having a different sign (step S
311) It is determined whether the weight Wj, [i] is positive or not (step S312). If the weight is Wj, [i], the weight is determined to be Wj, [i].
-1 (old weight value -1) (step S31)
3) If negative or 0, the weight is updated to Wj, [i] +1 (old weight value + 1) (step S314).

On the other hand, if the similarity χj is equal to or larger than the threshold θ in step S310, the similarity of the learning target character is set to χj in Equation 1, and the character number having the largest similarity among the characters other than the learning target is set to を j. j2, this similarity is χ
j2 and the similarity 文字 j of the character most similar to the character number j
2 (step S320), and the similarities χj and χj
It is determined whether the difference between the two is equal to or smaller than a predetermined value α (step S321). If the difference is equal to or smaller than the predetermined value α, the weight is updated according to the following equation (5). The weight update is executed by performing FIG. 13 as shown in FIG. Character with character number j and character number j2
In the image coordinate i of the character of, if weight coordinates having different signs are difi, the condition (χj−χj
2) If ≦ α, the following equation 5 is executed.

[0031]

[Mathematical formula-see original document] if (W [dif i], j≥0): W [di
f i], j + 1; if (W [diff i], j <0): W [dif
i], j-1; That is, if the difference is equal to or smaller than the predetermined value α in step S321, the process is executed only for i having a different sign (step S322), and the sign of Wj, [i] is determined (step S322). Step S323), if Wj, [i] is positive, Wj, [i] +1
(Step S324), and if Wj, [i] is negative, update to Wj, [i] -1 (step S32).
5). Thereafter, it is determined whether or not there is an end command (step S330). If there is no end command, the process returns to step S302. If there is an end command, the learned weight is written into the flash memory 16 to end the learning (step S331).

The weights Wj to the flash memory 16 are
Writing of [i] is performed on the discriminator side by a write command (step S500) from the personal computer 40 separately as shown in FIG. 15 (steps S501 and S5).
02). The weight map of “5” after learning is shown in FIG.
(A), and the weight map of the character “6” after learning is as shown in FIG. This is shown in detail in FIG. 17, which is used to identify similar characters such as characters "5" and "6" and characters "C" and "G" in order to increase the degree of identification. The weight is increased to intentionally emphasize the feature points. Since the arrow in FIG. 17 is a point to be distinguished from "6", the weight is increased.
FIG. 18 is for explaining the character "C", and the weight of the arrow is made large to distinguish it from the character "G". Further, the weight is reduced because the hatched portion in the peripheral portion may or may not be chipped depending on the image.

[0033]

According to the serial number recognizing device of the present invention, high-speed serial number recognition can be performed without affecting the transport state of the paper sheet. In addition, the weight learning function makes it possible to perform recognition with little effect on variations due to missing characters or the like.

[Brief description of the drawings]

FIG. 1 is a block diagram showing the overall configuration of the present invention.

FIG. 2 is a diagram showing the relationship between bills and serial numbers.

FIG. 3 is a block diagram illustrating details of an image processing determination unit according to the present invention.

FIG. 4 is a flowchart showing an overall operation example of the present invention.

FIG. 5 is a flowchart illustrating a detailed operation example of data acquisition in the identification processing.

FIG. 6 is a flowchart showing an operation example of serial number recognition according to the present invention.

FIG. 7 is a diagram illustrating an example of coordinates of a binarized character;

FIG. 8 is a diagram showing a relationship between coordinate values of pixels for serial number identification according to the present invention, a character to be identified, and a weight Wij.

FIG. 9 is a flowchart illustrating an example of a learning operation of serial number identification.

FIG. 10 is a flowchart illustrating an example of a learning operation in the personal computer.

FIG. 11 is a diagram for explaining weight update without an update condition in the first learning.

FIG. 12 is a diagram for explaining weight update when the similarity is equal to or less than a predetermined value.

FIG. 13 is a diagram for explaining weight updating when the maximum similarity and the second similarity are equal to or smaller than predetermined values.

FIG. 14 is a diagram for explaining weight updating when the maximum similarity and the second similarity are equal to or smaller than predetermined values.

FIG. 15 is a flowchart showing an example of a write operation to a flash memory by a personal computer.

FIG. 16 is a diagram showing an example of a weight map after learning of characters “5” and “6”.

FIG. 17 is a diagram illustrating an effect of a weight map by learning using a character “5” as an example.

FIG. 18 is a diagram illustrating an effect of a weight map by learning using a character “6” as an example.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Banknote 2 Light emitting means 3, 4 Line sensor 7 Passage sensor 8 True / false sensor 10 Image processing determination part 20 Discriminator control part 30 Main body control part

Claims (3)

[Claims] A first processing unit that processes optical image data of a paper sheet to identify at least a type and a direction of the paper sheet;
Cooperating with a sensor, a second sensor having a higher resolution than the first sensor for reading a serial number in the image of the sheet, a non-volatile memory for storing identification parameters, and a ROM and a RAM. And an identification operation unit for identifying the sheet by the image data and the parameters of the first sensor and the second sensor, and identifying the serial number of the sheet. An apparatus for recognizing a serial number of a paper sheet. 2. The serial number recognizing device for paper sheets according to claim 1, wherein the type and direction of the paper sheets to be conveyed based on the image data by the first sensor are determined, and thereafter, the serial number is determined from the image data. Based on information on a print character search area having a serial number that differs depending on a preset denomination and direction, the coordinates are converted in the x-axis direction and the y-axis direction in the image area by the second sensor, and the similarity is calculated each time. , The print character search area of the serial number is specified, the similarity greater than a predetermined value is obtained, and the similarity with all the characters to be identified is calculated. When there is a predetermined difference in similarity between the character and the character of the candidate having the second similarity, the character of the first recognition candidate is determined as the identification result. Number recognition method for paper sheets. 3. A learning method of learning the number characters of the paper sheet and binarizing an image cut out for calculating a similarity from 1 and −1 by a predetermined threshold to obtain a first weight map, Compare the sign of the binarized pixel value for each image to be learned with the sign of each pixel of the created weight map, add / subtract only portions having different signs, and repeat the addition / subtraction by the comparison until a predetermined evaluation value is obtained. A similarity is calculated as a second candidate using a character most similar to the corresponding character from the learned character data, and if the similarity difference between the learning character and the second candidate is smaller than a predetermined value, the second similarity is calculated. A learning method for recognizing serial numbers of paper sheets, wherein learning is performed by manipulating weights of the coordinates so that a difference between a pixel having a sign different from that of a candidate weight map is increased.