JPH10198837A - Paper sheet discriminating device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a discriminating device for paper money, paper sheets or the like that can fast perform discriminating processing with high accuracy and also contribute to the reduction of misreading ratio. SOLUTION: The outline dimensions of a paper money P are calculated (S10), and a read data pattern of the paper money P itself is sought (S20). Next, the read data pattern RP of the paper money P is aligned to a reference data pattern BP of an initial discriminating denomination based on a passage position and an oblique angle θ(S30), and the passage direction of the paper money P is decided with respect to a part of the pattern RP and the bill P of the initial discriminating denomination (S40). A segmenting position at the time of being oblique is calculated (S50) by referring to a segmenting position that is preliminarily stored, the oblique angle θ of an image of a segmenting position is corrected on memory according to a prescribed formula (S60), and normalization correlation matching is performed (S70).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a sheet discriminating apparatus for discriminating the type and authenticity of bills, checks, and other sheets, and more particularly, to a highly accurate sheet discriminating process at a high speed. Regarding what you did.

[0002]

2. Description of the Related Art Generally, a bill discriminating apparatus or a cash depositing / dispensing apparatus uses a bill discriminating apparatus for discriminating the denomination and authenticity of a bill inserted by a customer and storing the bill in a safe. In FIG.
The outline of a banknote discriminating device generally known conventionally is shown. The image sensor 2 is a linear image sensor that detects a one-dimensional image. A bill P to be discriminated is transferred so as to pass through the image sensor 2 in a direction transverse to the transfer direction F in order to increase the discrimination processing speed. You. Therefore, the image sensor 2 optically scans the bill P in the longitudinal direction (longitudinal direction). Image sensor 2
The timing sensor 1 composed of a light emitting element and a light receiving element provided on the upstream side detects the leading end of the bill P which is being conveyed, and outputs a bill detection signal by a light shielding effect of the leading end of the bill P being transferred. . In response to this bill detection signal,
The analog video signal for each scan output from the image sensor 2 is converted into digital gradation (multi-valued) data and then stored in the memory 3. Thus, the multi-value data of the entire surface of the bill P is stored in the memory 3 by moving the bill P with respect to the image sensor 2. Thereafter, the banknote determination unit 4 detects the edge of the banknote P based on the multi-value data on the entire surface of the banknote P stored in the memory 3 in this manner, thereby denominating the denomination of the banknote P and The authenticity is determined with respect to the external dimensions. Further, the banknote determination unit 4 stores each banknote P in the reference memory 5 based on a data pattern of the multi-value data corresponding to a print pattern such as a seal or a portrait printed on the banknote P (hereinafter, referred to as a read data pattern). The denomination and authenticity of the banknote P are determined for the print pattern by pattern matching with a reference data pattern prepared in advance for each banknote.

[0003] In consideration of the possibility that bills P are inserted in four directions in total, front and back, front and back, there is a possibility of passing through the image sensor 2 in the four directions. Reference data patterns are prepared in the reference memory 5 for each of these passing directions. Further, in this type of bill discriminating apparatus, since the bills P of a plurality of denominations pass through the timing sensor 1 and the image sensor 2 in the horizontal direction, respectively, as described above, the transfer to the timing sensor 1 and the image sensor 2 is performed. The road is a banknote having a maximum length L (that is, a 10,000-yen bill) so that banknotes P of different denominations having different longitudinal lengths (hereinafter, simply referred to as lengths) can be transported with an appropriate margin. It has a width somewhat larger than P. For this reason, each bill (particularly 5,000-yen bill and 1,000-yen bill shorter than the 10,000-yen bill) inserted into the apparatus passes through various positions in the width direction of the transfer path. Become. Therefore, the passing position of the bill P with respect to the image sensor 2 is also different.
In order to cope with this, the reference data pattern for each bill P is also stored for each of various different passing positions of the bill P with respect to the image sensor 2. That is, conventionally, the reference data pattern is stored for each passing direction of each bill P to the image sensor 2.

[0004]

However, in the above-mentioned conventional bill discriminating apparatus, in order to discriminate each inserted bill P, a reference data pattern for each bill P is defined by:
Since the data pattern is stored for each passing direction with respect to the image sensor 2, the number of reference data patterns to be matched with the data pattern becomes extremely large, and as a result, the bill discrimination processing time becomes very long. In addition, a large reference memory 5 is required. If the number of reference data patterns is reduced in order to reduce these problems, the degree of positional matching between the two patterns is reduced, and accurate matching cannot be performed. As a result, high precision There is a problem that the bill cannot be discriminated. further,
In the conventional banknote discriminating apparatus, when performing discrimination processing on banknotes of each country, it is necessary to perform such feature extraction for each country, and there is a problem that it takes a lot of time to develop the apparatus. The present invention has been made in view of the above, and an object of the present invention is to provide a paper sheet discriminating apparatus capable of performing high-accuracy sheet discriminating processing at a high speed and contributing to a reduction in an erroneous reading rate. Is to do.

[0005]

SUMMARY OF THE INVENTION A sheet discriminating apparatus according to the present invention optically scans one side of an inserted sheet and multi-valued data corresponding to the density of the sheet side for each pixel. Reading means for reading, contour detecting means for detecting an edge of the sheet based on the multi-valued data to detect a contour of the sheet and a skew angle at the time of reading, and reading the multi-valued data value of the sheet. Calculating means for calculating for each of a plurality of adjacent pixels; creating means for creating a read data pattern of the sheet using the arithmetic result value as a data element; and reference data corresponding to a characteristic range of various authentic sheets. A reference pattern storing means for storing a pattern in advance with a data element having twice the resolution of the read data pattern, and a read pattern corresponding to the reference data pattern based on the detected outer shape and skew angle. Positioning means for pre-positioning, and a cutting position storing means for storing in advance a cutting position for extracting a part of the read data pattern corresponding to a characteristic range of various authentic paper sheets; A cut-out position calculating means for calculating a cut-out position corresponding to the skew angle of the paper sheet by referring to the position, and cutting out only a range corresponding to the calculated cut-out position from the read data pattern aligned. It has range cutout means and discriminating means for discriminating the type of the inserted paper sheet by performing correlation value matching between the read data pattern of the cutout area and the reference data pattern. As a result, a read data pattern corresponding to only a characteristic range of a part of the paper sheet can be cut out, and the processing time related to correlation value matching can be reduced. In addition, high-precision correlation value matching can be performed using a single reference data pattern, and as a result, high-precision sheet discrimination processing can be performed at high speed without requiring a huge amount of data processing. it can.
Furthermore, by cutting out only a characteristic portion such as a character portion and performing correlation value matching, the erroneous reading rate can be reduced. Furthermore, by cutting out only the characteristic part corresponding to each banknote in each country and performing correlation value matching, the misreading rate can be reduced. In this case, it is not necessary to perform feature extraction for each country, so that the time required for device development can be reduced. According to the present invention, the determining means includes:
Correlation value matching may be performed between one or more ranges of the read data pattern cut out by the range cutout means and the reference data pattern. As a result, it is possible to perform high-precision sheet discrimination processing at high speed without requiring a huge amount of data processing, and further, cut out only a characteristic portion corresponding to each banknote in each country and perform correlation value matching. Thereby, the erroneous reading rate can be reduced.

[0006]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below in detail with reference to the accompanying drawings. Hereinafter, an example in which the sheet discriminating apparatus of the present invention is configured and implemented as a bill discriminating apparatus will be described. However, it is needless to say that the sheet discriminating apparatus can be configured as a discriminating apparatus for other sheets such as checks and securities. . FIG.
FIG. 2 is a block diagram showing an example of a hardware configuration of a bill discriminating apparatus according to an embodiment of the present invention, and FIG. 2 is a diagram illustrating the operation of the apparatus shown in FIG. Here, for convenience, the description will be limited to the case of processing Japanese banknotes (10,000 yen, 5,000 yen, and 1,000 yen bills). First, in this embodiment, as shown in FIG.
(Here, it is indicated by a broken line for simplicity.)
In order to speed up the bill discrimination process, the bill P to be discriminated is moved laterally with respect to the transfer direction F by the transfer roller 13 in the image sensor 10.
, While being optically scanned by the image sensor 10 in the length L direction. Here, the image sensor 10 is a linear image sensor that detects a one-dimensional image by receiving reflected light (reflection image) of light emitted from an LED array (not shown). In this embodiment, the image sensor 10 detects a reflection image of 1 mm per pixel for performing a line scan (indicated by a dashed line S) of the banknote P. A transfer roller 13 for transferring the banknotes provided below the image sensor 10 transfers the banknotes P in the transfer direction F with respect to the image sensor 10 at a predetermined scan pitch proportional to the speed of the transfer. The image sensor 10 can scan the entire surface of the bill P.

Further, in this embodiment, the transfer roller 13 provided below the image sensor 10 in the transfer path 12 is black so that the banknote P can be detected by a change in the gray level of the reflection image. and,
The image sensor 10 performs a scanning operation before the bill P is transferred (for example, from the time when the bill P is detected to be inserted into a predetermined bill insertion slot of the apparatus). In this way, as will be described in detail later, after the bill P is inserted, first, the image sensor 1
When the output level of 0 rises from the black level by the transport roller 13 to the white level when the banknote P is not present, it is detected that the leading end of the banknote P in the transport direction F has passed through the image sensor 10. From the point of detection of the leading edge, an operation of storing output data for each horizontal line scan (hereinafter simply referred to as a horizontal scan S) of the image sensor 10 in the memory (that is, an operation of being taken in) as described later. Is started. After that, after a number of horizontal scans S are performed at a predetermined scan pitch, the output level of the image sensor 10 falls to a black level, and when the black level continues for a predetermined time, after the transfer direction F of the bill P, It is detected that the end has passed the image sensor 10, and the storage operation of the output data of the image sensor 10 is terminated at the time of the detection of the rear end. With such a configuration, in this embodiment, there is no need to provide a timing sensor (reference numeral 1 in FIG. 9) which was conventionally required, and the configuration of the device is correspondingly simplified.

[0008] Japanese banknotes P consisting of 10,000-yen bills, 5,000-yen bills and 1,000-yen bills have the same width but different lengths L. Therefore, the length L of each bill P is determined by bill discrimination. It can be an important feature point. Focusing on this point, in this embodiment, the denomination and authenticity of the banknote P are determined,
For the purpose of detecting an unacceptable portion of the bill P that is not acceptable, the length L of each bill P is detected with a fine resolution of 1 mm unit. However, in the bill discriminating apparatus, although there is a demand for speeding up the discriminating process of the bill P as much as possible, each bill P
However, if the print pattern printed on the printer is detected with a fine resolution, the amount of data processing becomes enormous and the processing time becomes longer. Therefore, in this embodiment, the external dimensions including the length L, which is an important characteristic point of the banknote P, are detected with a fine resolution by the configuration illustrated in FIG. 1, however, as data indicating the detected external dimensions. Tries to capture the least. On the other hand, a print pattern that can be automatically detected relatively easily is detected with a coarser resolution than the detection of the external dimensions.

In FIG. 1, a microprocessor 16
Controls the entire operation of the bill discriminating apparatus.
The clock generator 14 generates a basic operation clock signal of the device. For example, when a bill input sensor detects that a certain bill P has been inserted into the apparatus, the start pulse generator 18 causes the image sensor 1 to operate in accordance with the timing specified by the operation clock signal.
A start pulse for instructing the start of each horizontal scan S by 0 is generated (see FIG. 2B). In response to the start pulse, the image sensor 10 starts a horizontal scan S at a predetermined pitch corresponding to the speed at which the banknote P is transported. The scan counter 19 outputs the pixel position (bit position) for each horizontal scan by counting the number of pixels from the start to the end of each horizontal scan S, as shown in (d) of FIG. The analog video signal output from the image sensor 10 is, for each pixel,
The data is converted by the A / D converter 20 into digital gradation data, that is, multi-value data. Each multi-value data is data obtained by quantizing the gray level of one pixel output from the image sensor 10 by multi-value. A waveform diagram represented by the value of the multi-value data per each horizontal scan S is as shown in FIG.

In FIG. 1, in order to detect the outer dimensions of a bill P to be discriminated, an edge of the bill P in the horizontal scan S direction is detected at a resolution of each pixel (1 mm in this example), and the detected edge is detected. The components for capturing the data indicating the position are as follows. The slice level register 21 stores a predetermined slice level (SL in FIG. 2C) for converting the multi-valued data into binary data.
) Is stored. As the slice level SL, a level is set at which it is possible to distinguish between “black” of the transport roller 13 and “white” of a non-printed portion (particularly, an edge portion) of the banknote P. The binarization unit 22 compares the level of each of the multi-valued data with the slice level, and thereby converts the multi-valued data into binary data indicating, for example, “black” and “white”. The binary data for each horizontal scan S output from the binarization unit 22 is provided to a rise / fall detection unit 24.

The rising / falling detecting section 24 determines, for each horizontal scan S, a point in time when the binary data changes from “black” data to “white” data by the image sensor 10.
Is detected as the rising edge of the output level (corresponding to the left edge of the bill P), and the point in time when the binary data changes from “white” data to “black” data is defined as the falling edge of the output level of the image sensor 10 (bill). (Corresponds to the right edge of P)
Is detected. In this embodiment, after detecting that the banknote P has been inserted, the microprocessor 16, as described above,
First, when it is detected that the output level of the image sensor 10 has risen from the black level to the white level, it is detected that the leading end of the bill P in the transfer direction F has passed through the image sensor 10. After that, a number of horizontal scans of the banknote P at a predetermined pitch are completed, and the image sensor 1
0 falls to the black level, and when the black level state continues for a predetermined time, the transport direction F of the bill P
It is configured to detect that the rear end has passed the image sensor 10.

When a rise is detected by the rise / fall detection unit 24 for each horizontal scan S, the bit position output by the scan counter 19 at that time is latched by a rise latch 26. Also,
Thereafter, when the falling edge is detected by the rising / falling detecting unit 24, the bit position output by the scan counter 19 at that time is latched by the falling latch 28. Thus, the rising and falling bit positions latched by rising latch 26 and falling latch 28 are provided to microprocessor 16. The microprocessor 16 stores the bit position in the external shape data memory 30. Thus, the outer shape data memory 30
The rising and falling edges of each horizontal scan S, that is, the bit positions of the edges of the bill P are stored as the outline data of the bill P. afterwards,
The microprocessor 16 detects the outer dimensions of the banknote P based on the outer shape data stored in the outer shape data memory 30 in this manner, and stores the reference outer dimensions of each bill stored in the reference outer shape memory 38 in advance. The banknote P is discriminated with respect to its outer dimensions by comparison and collation.

Further, in FIG.
The components for detecting and capturing the print pattern with a resolution lower than that at the time of detecting the external dimensions are as follows. When the rise is detected by the rise / fall detection unit 24 as described above, the multi-valued data for each horizontal scan S digitized by the A / D converter 20 is given to the first addition circuit 34. Can be In the first adder circuit 34, as shown in FIG. 3A, according to a timing signal generated by the frequency divider 35 and obtained by dividing the clock signal by 1 / N (for example, 1/4). Are added to the values D0, D1,... Of the multivalued data for every four pixels in the horizontal direction to generate horizontal addition values H0, H1,. Further, in the second adding circuit 36, as shown in FIG. 3B, the horizontal direction added values H0, H
Are added for every four pixels in the vertical direction to create vertical addition values V0, V1,. The vertical addition values V0, V1,... For each horizontal scan S are stored in the FIFO memory 37.
Is stored. The FIFO memory 37 is for asynchronously controlling the timing so that the vertical addition values V0, V1,. Microprocessor 1
In step 6, the vertical addition values V0, V1,... Stored in the FIFO memory 37 are stored in the pattern data memory 39 as read pattern data. Thus,
The pattern data memory 39 stores 4 × 4 pixels of the image sensor 10, that is, read pattern data V0, V1,... Every 4 mm × 4 mm in this case. Hereinafter, for the sake of explanation, each of the read pattern data V0, V1,.
Is referred to as a read data element V. As explained later,
In the host computer 32, based on the read data elements V stored in the pattern data memory 39 in this manner, the entire read data pattern R of the bill P itself is read.
P is obtained, and a cutout position for extracting a part of the read data pattern PR in advance corresponding to a part of the read data pattern RP and a characteristic range of a genuine bill of each denomination is stored in the cutout position memory 41 in advance. To be stored. Here, the cutout memory 43 for storing a part of the read data pattern PR after the skew correction, and the reference data pattern BP stored in advance in the reference pattern memory 40 are
Matching is performed using a correlation value calculation formula described later, whereby the bill P is determined with respect to its print pattern.

In the above embodiment, in order to detect the print pattern of the banknote P, as described above,
The reason for taking in the data of the entire surface without thinning out is that, in the case of thinning out and taking in, because the print pattern is a line drawing, moire fringes occur due to the relationship between the sampling pitch and the pitch of the printed line. This is because the detection is unstable. Further, since the captured data is continuous data without omission, a pattern at an arbitrary position can be selectively detected. The components surrounded by the dashed line in FIG. 1 are not limited to hardware circuits, but may be realized by software processing. Further, using the apparatus according to this embodiment, the reference outer shape and the reference data pattern RP for all the denomination bills to be discriminated are used.
If it is stored, not only Japanese banknotes but also foreign banknotes can be easily handled. Further, the image sensor 10 is not limited to the type in which the image sensor 10 receives the reflected light as in the above embodiment, but may be the type in which the image sensor 10 receives the transmitted light.

The reference data pattern BP stored in the reference pattern memory 40 for each denomination banknote P will be described with reference to FIG. In this embodiment, only one reference data pattern BP having twice the resolution of the read data pattern RP is stored for each passing direction of the banknotes P of each denomination with respect to the image sensor 10. Absent. That is, as described below, in this embodiment, the reference data pattern BP has a reference data element having the same size of 4 mm × 4 mm as the read data element V at a pitch of 2 mm ( The read data element V has a pitch of 4 mm). More specifically, in order to create the reference data pattern BP of each denomination banknote, the authenticity of each denomination is determined by performing substantially the same processing as described above using the above-described apparatus of this embodiment. 4A is optically read to obtain an original data pattern OP composed of 2 mm × 2 mm data elements A relating to the entire surface of the bill P as shown in FIG. Next, the values of four adjacent data elements in the original data pattern OP in the vertical and horizontal directions are sequentially added while being shifted by 2 mm, thereby obtaining a value of 4 mm × 4 mm as shown in FIG.
A reference data pattern BP including a 4 mm reference data element B is created. In this example, for example, the reference data element B11 is obtained by adding the data elements A00, A01, A10 and A11 of the original data pattern OP, and the reference data element B12 is obtained by adding A01, A02, A11 and A12. The reference data element B21 includes A10, A1
1, A20 and A21 are added, and the reference data element B22 is obtained by adding A11, A12, A21 and A22.

In this manner, as shown in FIG. 5, each of the reference pattern memories 40 has a size of 4 mm × 4 mm.
Are stored in advance at a pitch of 2 mm. In addition, in FIG.
a reference data element B11 having a size of mx 4 mm, a reference data element B1 shifted rightward by 2 mm from the element B11;
2, and only the reference data element B21 shifted downward by 2 mm from the element B11 is illustrated, but other reference data elements are also sequentially stored at a pitch of 2 mm as indicated by broken lines.

Next, the bill discriminating process performed by the microprocessor 16 will be described with reference to FIG. First, in step S10, the microprocessor 16 detects a passing position and a skew (skew) angle θ of the bill P with respect to the image sensor 10 based on the outline data of the banknote P stored in the outline data memory 30. Further, based on this detection, the outer dimensions, particularly the length L of the bill P are calculated. The length L is obtained, for example, by using the edge data of the non-defective portion of the bill P among the left and right edge data of the bill P, and calculating the average value of the difference values of the left and right edge positions of the bill P. , The average value is corrected by the skew angle θ. Then, the reference external dimensions in the reference external memory 34 that match the external dimensions determined in this way are detected, and the denomination of the adapted reference external dimensions is determined as the initial discrimination denomination. Next, in step S20, of the read data elements V stored in the pattern data memory 39, only the data relating to the banknote P (that is, the data excluding the data relating to the transport roller 13 as the background) is cut out, thereby A read data pattern of the bill P itself is obtained.

Further, in step S30, the bill P
The read data pattern RP of the banknote P stored in the pattern data memory 39 is stored in the reference pattern memory 40 based on the passing position and the skew angle θ with respect to the image sensor 10. (Normalization) with respect to the reference data pattern BP. In this case, since the pitch of the reference data element of the reference data pattern BP is 2 mm, the positioning of the read data pattern RP can be performed with high accuracy equal to this pitch. In this embodiment, since such a configuration is adopted, it is not necessary to prepare a reference data pattern BP for each passing position of each banknote P with respect to the image sensor as in the related art. Step S4
In the case of 0, a part of the read data pattern RP and the four kinds of references stored for each of the passing directions (one of four kinds of passing directions of front and back, front and back) for the banknote P of the initial discrimination denomination. A corresponding overall pattern matching of the data pattern BP is performed.
Is determined, that is, the passing direction to the image sensor 10 is determined. That is, in this case, the passing direction is determined from the direction of the reference data pattern BP having the highest correlation value among the four types of reference data patterns BP.

Here, the skew correction processing and the cutout processing will be described with reference to FIGS. Step S50
Then, the cutout position at the time of skew is calculated with reference to the cutout position stored in the cutout position memory 41 in advance. The set value (Sx, Sy) of the cutout range previously stored in the cutout position memory 41 on the xy plane, and the set value (Lx, Ly) of the distance from the center of the bill to the center of the cutout range on the xy plane. And the center of gravity G (Gx, Gy) on the xy plane of the input image calculated from the outline of the banknote P stored in the outline data memory 30 and the center of gravity of the cutout range of the input image. Distance (Cx,
Cy) The equation to be obtained is

Cx = Lx · sin θ + Ly · sin θ Cy = −Lx · sin θ + Ly · cos θ In addition, as shown in FIG. 8, θ used in the above (Equation 1) is the angle of the banknote P when skewed.

Next, in step S60, the skew angle θ of the image at the cut-out position is corrected on the memory in accordance with the above (Equation 1). That is, the data read from the pattern data memory 39 is set such that the distance (fx, fy) of the upper left corner of the cut-out range from the cut-out range center matches the center of the cut-out memory 43 after the skew correction. The skew correction processing is performed by substituting the calculation result into (Equation 1) and sequentially storing the calculation results in the memory 43. Note that the addresses from the cutout start positions Ax and Ay on the pattern data memory 39 are:

## EQU2 ## Address = Rx.Ay + Ax

In step S70, normalized correlation matching is performed. That is, matching is performed a plurality of times with dictionary data having a margin for positional deviation to obtain the maximum correlation. That is, partial matching of the read data pattern RP stored in the cut-out memory 43 is performed using the reference data pattern (that is, the direction correspondence reference data pattern) BP. That is, the correlation value matching is performed for each read data element V of the read data pattern RP with the reference data element B corresponding to the position in the aligned reference data pattern. By the way, even if the read data pattern RP and the reference data pattern BP are aligned in step S30, the alignment is based on the outer shape of the banknote P. This usually results in some alignment error. However, in pattern matching using a correlation value calculation formula described later, the correlation value is significantly reduced due to such an alignment error.

For this reason, in this embodiment, each read data element V of the read data pattern RP of the banknote P to be discriminated is replaced with a reference data element which corresponds to the position of the reference data pattern BP aligned as described above. One-to-one with B
In addition to the above case, the above-described read data pattern RP is displaced vertically (left and right) and diagonally with respect to the reference data pattern BP (positioning error is generated), and The reference data element B shifted by one pitch (2 mm) in the vertical, horizontal, and diagonal directions from the position (basic position) thus aligned is also targeted for matching. That is, in this embodiment, each read data element V of the read data pattern RP of the banknote P to be discriminated includes a corresponding reference data element B and one pitch in the up, down, left, right, and diagonal directions. It is matched with a total of nine reference data elements B which are reference data elements B shifted only by 9 positions (a total of nine patterns, and thus, one to nine).

The partial matching between the read data pattern RP and the nine positions of the reference data pattern BP for each denomination banknote P is performed using the following correlation value calculation formula.

(Equation 3) Here, n is the total number of read data elements to be matched in the read data pattern RP, i is the number of the data element to be matched, Bi is the value of each reference data element of the reference data pattern BP, and Vi is The value of each read data element of the read data pattern RP, Ba
Is an average value of all reference data elements to be matched in the reference data pattern BP, and Va is an average value of all read data elements to be matched in the read data pattern RP.

When the read data pattern RP and the reference data pattern BP relate to the same print pattern, the correlation value becomes the maximum value (peak) when the positions of the two patterns RP and BP completely match. In contrast, when the two patterns RP and BP relate to different print patterns, they exhibit flat characteristics without an extreme peak. The reason why the resolution of the reference data pattern BP at the basic position is twice that of the read data pattern RP, that is, the pitch of the reference data element is 要素 of that of the read data element is as follows. That is, if the pitch of each reference data element is increased in order to reduce the data amount of the reference data pattern BP to be matched with the read data pattern RP, the pitch can be obtained by shifting one pitch from the basic position in each direction as described above. The amount of change in the correlation value obtained is too large. For example, if the pitch of each reference data element of the reference data pattern BP is 4 mm, which is the same as the read data element, data from the position of −2 mm to the position of +2 mm with respect to the basic position jumps, so that the correlation peak value decreases considerably. Would.
However, if the pitch of each reference data element is １ ／ of the read data element, that is, 2 mm as in the above embodiment, a correlation peak value in a range of ± 1 mm from the basic position can be obtained.

The maximum value of the correlation values obtained by matching at the nine positions in the reference data pattern BP is adopted as the matching result. If the matching result is equal to or more than a predetermined value, the banknote P is finally determined as a genuine banknote, and the denomination and the insertion direction are further output to a computer (not shown) of a control system. In the above embodiment, the size of each read data element V of the read data pattern RP is 4 × 4 pixels of the image sensor 10 (that is, 4 mm × 4 mm), and the size of the reference data element of the reference data pattern BP is the same. However, the size of each data element of both patterns RP and BP is
It may be somewhat larger or smaller than those described above. In short, it is sufficient that the resolution of the reference data pattern BP is twice the resolution of the read data pattern RP, that is, the pitch is 1/2.

In the above embodiment, both patterns R
Each data element of P and BP is an addition value of multi-value data of a plurality of pixels, but is not limited thereto, and may be an average value of multi-value data of these pixels. Furthermore, by performing correlation value matching between one or more ranges of read data patterns cut out by the range cutout means and the reference data pattern, a huge amount of data processing is not required, and highly accurate bill discrimination. The processing can be performed at high speed, and only the characteristic portion corresponding to each banknote in each country is cut out and subjected to correlation value matching, whereby the misreading rate can be reduced.

[0027]

As described above, according to the present invention, one side of the inserted paper sheet is optically scanned, and multi-value data corresponding to the density of the paper sheet is read for each pixel. When the sheet is read by the means, the edge of the sheet is detected based on the multi-value data, and the outer shape of the sheet and the skew angle at the time of reading are detected by the outer shape detecting means. On the other hand, the arithmetic means calculates the multi-value data value of the sheet for each of a plurality of adjacent pixels, and then creates a read data pattern of the sheet using the calculation result value as a data element. Here, a reference data pattern corresponding to a characteristic range of various genuine paper sheets is stored in the reference pattern storage means in advance with data elements having twice the resolution of the read data pattern, and the detected outer shape and obliqueness are stored. Based on the row angle, the read data pattern is aligned by the alignment means in correspondence with the reference data pattern. Further, a cutout position for extracting a part of the read data pattern corresponding to a characteristic range of various genuine paper sheets is stored in a cutout position storage unit in advance,
With reference to the cut-out position, a cut-out position corresponding to the skew angle of the sheet is calculated by a cut-out position calculating unit.
Next, of the read data patterns aligned, only the range corresponding to the calculated cut-out position is cut out by the range cut-out means, and between the read data pattern of the cut-out range and the reference data pattern,
The type of the paper sheet input by performing correlation value matching is determined by the determination unit. As a result, a read data pattern corresponding to only a characteristic range of a part of the paper sheet can be cut out, and the processing time related to correlation value matching can be reduced. In addition, high-precision correlation value matching can be performed using a single reference data pattern, and as a result, high-precision sheet discrimination processing can be performed at high speed without requiring a huge amount of data processing. it can.
Furthermore, by cutting out only a characteristic portion such as a character portion and performing correlation value matching, the erroneous reading rate can be reduced. Furthermore, by cutting out only the characteristic part corresponding to each banknote in each country and performing correlation value matching, the misreading rate can be reduced. In this case, it is not necessary to perform feature extraction for each country, so that the time required for device development can be reduced. Further, the discriminating means determines between the reference data pattern and one or more ranges of the read data pattern clipped by the range clipping means.
If correlation value matching is performed, high-precision paper discrimination processing can be performed at high speed without requiring a huge amount of data processing, and only the characteristic part corresponding to each banknote of each country is cut out By performing correlation value matching,
The misreading rate can be reduced.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an example of a hardware configuration of a bill discriminating apparatus according to an embodiment of the present invention.

FIG. 2 is a timing chart for explaining the operation of the embodiment in FIG. 1;

FIG. 3 is a view for explaining an operation of taking in a banknote read data pattern in the embodiment.

FIG. 4 is a view for explaining an operation of creating a reference data pattern for each genuine bill of each denomination in the embodiment.

FIG. 5 is a view for explaining a storage state of the reference data pattern.

FIG. 6 is an exemplary flowchart showing bill discriminating processing performed by the host computer according to the embodiment;

FIG. 7 is a view for explaining a cutout position corresponding to a characteristic range of a genuine bill stored in advance in the cutout position memory of FIG. 1;

FIG. 8 is a diagram illustrating skew correction processing and cutout processing.

FIG. 9 is a diagram illustrating a conventional bill discriminating apparatus.

[Explanation of symbols]

 Reference Signs List 10 image scanner 16 microprocessor 40 reference pattern memory 41 cutout position memory 43 cutout memory

Claims (2)

[Claims] 1. A reading means for optically scanning one surface of an inserted paper sheet and reading multi-valued data corresponding to the density of the paper sheet for each pixel, and reading the paper based on the multi-valued data. Outer shape detecting means for detecting the edge of the leaf to detect the outer shape of the sheet and the skew angle at the time of reading; calculating means for calculating a multi-value data value of the sheet for each of a plurality of adjacent pixels; Creating means for creating a read data pattern of the paper sheet using the operation result value as a data element; and a data element having a resolution twice as large as the read data pattern, using a reference data pattern corresponding to a characteristic range of various authentic paper sheets. A reference pattern storage means for storing in advance, and a positioning means for positioning the read data pattern in correspondence with the reference data pattern based on the detected outer shape and skew angle. A cutout position storing means for storing in advance a cutout position for extracting a part of the read data pattern in correspondence with a characteristic range in a genuine sheet of a seed; A cut-out position calculating unit that calculates a cut-out position corresponding to a line angle; a range cut-out unit that cuts out only a range corresponding to the calculated cut-out position from the aligned read data pattern; and the cut-out range And a discriminating unit for discriminating the type of the inserted sheet by performing correlation value matching between the read data pattern of the sheet and the reference data pattern. 2. The method according to claim 1, wherein the determining unit performs correlation value matching between the read data pattern of one or more ranges cut out by the range cutting unit and the reference data pattern. The paper sheet discriminating apparatus described in the above.