JP2519791B2 - Pattern recognizing device for paper sheets - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION Object of the Invention (Industrial field of application) The present invention determines the type of a banknote, for example, by the front and back patterns of the banknote, and determines the authenticity and the damage of the banknote based on the determination result. The present invention relates to a paper sheet pattern identification device suitable for application in, for example, the determination of the type of paper money in the paper money identification device.

(Prior Art) Conventionally, this type of pattern identification device scans the front surface and the back surface of a conveyed banknote with a dedicated reflection optical system, respectively, to thereby generate an image signal corresponding to each pattern on the front surface and the back surface of the banknote. To each image signal, and 1 for each image signal
Set the identification area of the location, create a sample pattern for this identification area, find the similarity between this sample pattern and the standard pattern of the banknote stored in advance, The front and back patterns of the bill are identified based on the large value.

However, in such a conventional pattern identification device,
The similarity is calculated for each of the two identification areas, and the pattern is identified based on the calculated value.For example, if a bill is partially damaged or soiled, the sample pattern will change significantly. However, there is a problem in that an identification error is likely to occur or the identification tends to be impossible.

(Problems to be Solved by the Invention) As described above, the present invention has a problem that, for example, when a part of a paper sheet is damaged or soiled, an identification error occurs or the identification is apt to occur. It was made to solve the problem, for example, even if a part of the paper sheet is damaged or soiled, it is possible to always perform correct pattern identification, and an identification error does not occur or it becomes impossible to identify. An object of the present invention is to provide a pattern identification device of a class.

[Configuration of the Invention] (Means for Solving the Problem) A paper sheet pattern identification apparatus according to the present invention is a conveyance means for conveying paper sheets, a surface of the paper sheets conveyed by the conveyance means, and There are provided a plurality of optical means, which are arranged to face each other on the back surface, respectively, for obtaining image signals corresponding to the patterns of the front and back surfaces of the paper sheet, and a plurality of identification areas to be subjected to the identification processing for the front and back surfaces of the paper sheet. Sample pattern creating means for setting and creating a pattern for each of the set plurality of identification areas from the image signal of the paper sheet obtained by the optical means, and a plurality of types of paper sheets to be transported by the transporting means Standard pattern storage means for storing in advance standard patterns corresponding to the respective identification areas, and all types of standard patterns stored in the standard pattern storage means. Pattern similarity calculation processing means for performing similarity calculation between the pattern and the pattern for each identification area created by the sample pattern creation means, and the optical means based on the calculation result by the pattern similarity calculation processing means. As a paper sheet candidate subjected to the processing of step 1, first editing processing means for independently extracting a plurality of types of paper sheet candidates in descending order of similarity for each of the identification areas, and the first editing Second editing processing means for independently extracting a plurality of types of paper sheet candidates for each of the front surface and the back surface of the paper sheet based on the extraction result by the processing means, and the second editing processing means. And a third editing processing means for extracting one paper sheet candidate as an identification result from the plurality of the selected candidates.

(Operation) When pattern identification processing is performed by the image signal obtained from the optical system, a plurality of identification areas are set, and the pattern identification processing for each identification area is performed by the image signal in each identification area as a first editing processing means. Then, a plurality of types of paper sheet candidates are independently extracted for each identification area set on the front and back of the paper sheet. The second edit processing means independently extracts a plurality of types of paper sheet candidates for each of the front and back sides of the paper sheet based on the extraction result of the first edit processing means, and further the third edit processing means. By this, one candidate of the paper sheet to be the identification processing target is determined and extracted. In this way, the paper sheet candidates are independently extracted from the identification area stage, and the processing is performed stepwise to finally narrow down the candidates to one. Even if the corresponding part is damaged or soiled and it is impossible to extract the paper sheet candidates, reliable pattern identification can be performed based on the information in the remaining identification area that is not damaged or soiled.

(Example) Hereinafter, one example of the present invention will be described with reference to the drawings.

FIG. 1 shows a bill discriminating apparatus to which the paper sheet pattern discriminating apparatus according to the present invention is applied. That is, P
Is a banknote (paper sheet) and is conveyed by the conveyor belt 1 in the Y direction at a constant speed. The banknote The P side along the transport path of the first optical system 2 ₁ is installed to obtain an image signal of a color corresponding to the surface pattern of the bill P conveyed. The first optical system 2 ₁ is a reflection type optical system, and includes a light source 3 ₁ that illuminates the detection area A on the banknote P and an X direction that receives reflected light from the detection area A on the banknote P and converts it into an electrical signal. The color line sensor 4 ₁ having a scanning function, and a lens (not shown) that forms an image on the color line sensor 4 ₁ with reflected light from the detection area A on the banknote P. Color line sensor 4 ₁ is constituted by a CCD color line sensor having 512 elements, for example.

Color image signals of the surface of the bill P obtained from the color line sensor 4 ₁ is amplified by the amplifier 5 _1, A / D converter 6
Sent to ₁ where it is converted to a digital signal. This digitized image signal is classified by color (eg IR, R, G, B)
Sequentially stored in the image storage unit ₇₁ to the.

On the other hand, in the conveyance path of the banknote P, the banknote P being conveyed
A detector 8 including a light emitting element and a light receiving element for detecting the tip of the detector is provided, and the output of the detector 8 is supplied to the timing pulse generator 9. The timing pulse generator 9 generates a variety of timing pulses based on the output of the detector 8, which color line sensor 4 _1, A / D converter 6
Supplied as drive pulses to ₁ and the image storage unit _71.

Further, with respect to the back side of the banknote P as well as the front side, the second optical system 2 ₂ , the amplifier 5 ₂ , the A / D conversion unit 6 ₂ and the image storage unit 7 are provided.
₂ are provided respectively.

In this manner, when the digital image signals of the front surface and the back surface of the banknote P, that is, the patterns of the front surface and the back surface of the banknote P are stored in the image storage units 7 ₁ and 7 ₂ for each color,
(U, etc.) 10 determines whether each banknote P is genuine or false by processing each color-based pattern stored in the image storage units 7 ₁ and 7 ₂ based on a program.

That is, first, the sample pattern creation processing routine 21
In, as shown in FIG. 2, the pattern RP of one color of the Color of the pattern stored in the image storage unit ₇₁ (for example R), with respect to the conveying direction of the banknote P Identification areas A ₁ and A ₂ are set by the front end reference and the rear end reference, respectively, and a sample pattern SP is created for each of the identification areas A ₁ and A ₂ . As shown in FIG. 2, this sample pattern SP is an identification area from X5 to X11 in the X direction (scanning direction of the color line sensor) and Y1 to Y10 in the Y direction (banknote transport direction). A ₁ , and X
X5 to X11 for direction and Y21 to Y3 for Y direction
The pattern data corresponding to the identification areas A ₂ up to 0 are read from the image storage unit 7 ₁ , and 7 points are 4 samples in the X direction and 10 points are 10 samples in the Y direction for each identification area A ₁ and A _2. Create a sample pattern with 40 samples in two dimensions and a level for each sample.

Further, the sample pattern SP is created on the back surface side of the banknote P by the same processing as that on the front surface side. Therefore, two for the front side of the banknote P and two for the back side.
A total of four sample patterns will be created.

Next, in the pattern similarity calculation processing routine 22,
Each of the four sample patterns created by the sample pattern creation processing routine 21 is created and stored in the standard pattern table 11 in advance.
One (two on the front side and two on the back side) standard patterns of all the ticket types corresponding to the identification areas are subjected to similarity calculation processing based on the following similarity calculation formulas, thereby And the degree of similarity (M) to That is, the pattern identification process is performed for each identification area.
Here, FIG. 3 (a) shows a sample pattern example of one identification area of the bill P, and FIG. 3 (b) shows a standard pattern example for one bill type. The similarity calculation processing is performed. In the similarity calculation formula below, Si (i = 1, 2, ..., Sample number)
Is one sample value of the sample pattern, for example 8
Bit positive / sample. Also, Di (i = 1,2, ...,
Sample number) is one sample value of the standard pattern,
For example, 7 bits + 1 bit of code.

Next, in the edit discrimination processing routine 23, the edit processing is performed by using the similarity (M) of each identification area obtained by the pattern similarity calculation processing routine 22 and the edit information table 12, whereby the bill P Species discrimination (pattern identification) is performed.

As shown in FIG. 4, the edit discrimination processing routine 23 firstly, by the first edit processing routine 31, the top two ticket types having the similarities to the respective ticket types (first candidate, second (Candidate) is extracted based on the similarity slice level value for each identification area and the similarity difference level value between the first candidate and the second candidate, which are preset and stored in the editing information table 12, and the similarity of the candidate is extracted. Based on the conversion conditions shown in Fig. 5, W0 to W5 (W5>W4>W3>W2>W1> W0 = 0)
Convert to the weight value of.

Next, by the second edit processing routine 32, the weights previously set and stored separately in the edit information table 12 in advance in the edit information table 12 based on the candidates and the converted weight values extracted in the first edit processing routine 31. Up to three candidates are extracted for both the front surface and the back surface by comparison processing with weight level information including slice level values and weight difference level values. That is, the weight value of each identification area converted by the first editing process is added for each ticket type (including the transport direction), and the maximum value is 3 based on the value and the information of the edit information table 12 based on the following conditions. Extract even candidates. Note that the following conditions indicate a case where the first, second, and third candidates are extracted in descending order of the added weight value for the third editing process, and the output information of the candidates extracted by this condition is shown. An example is shown in FIG.

(1) Condition for extracting first candidate Weight value of first candidate ≧ weight slice level value of edit information table (2) Condition for extracting second candidate Satisfying the condition of (1) above Weight difference value <Weight difference level value in edit information table (3) Condition for extracting third candidate Satisfies the above condition (1) Weight difference value for first and third candidates <Weight difference level value in edit information table Last Then, the third edit processing routine 33 edits the front surface and the back surface, which are preset and stored in the edit information table 12, from the ticket types and the weight values of the respective candidates of the front surface and the back surface extracted in the second edit processing routine 32. Mode, and second
Candidate 1 based on weight level information similar to edit processing
The candidate ticket type is used as the discriminating ticket type. That is, the weight values of each of the front and back candidates obtained in the second editing process are added for each ticket type and direction, and the weight values of the top two candidates are designated in the editing mode shown in FIG. In the edit mode, the weight value and the difference are compared with the weight slice level value and the weight difference level value in the edit information table 12 to narrow down to one candidate and the discrimination processing is performed. The edit mode is a mode for designating the combination of the discrimination processing by dividing the bill type and the direction and the processing by the combination of the front surface and the back surface, as shown in FIG.
There is a street and is designated by the edit information table 12.

On the other hand, the sample data creation processing routine 24
In, create sample data for authenticity fitness determination by accumulating a pattern data corresponding to the determination area B, such as from the image storage unit _71, 7 ₂ shown in FIG. 8 read, to separate colors. As shown in FIG. 8, the discrimination area B is an area from Xa to Xb in the X direction and from Ya to Yb in the Y direction.

Next, in the color calculation processing routine 25, the sample data for each color created in the true / false / damage sample data creation processing routine 24 is processed based on a color calculation expression suitable for each discrimination.

Next, in the comparison / determination processing routine 26, among the minimum reference value and the maximum reference value preset and stored in the authenticity / damage reference table 13, the reference value corresponding to the note type determination result of the edit determination processing routine 23. Is used to perform a comparison process with the value obtained by the color calculation process routine 25, thereby discriminating between true and false and a loss.

Finally, in the comprehensive discrimination processing routine 27, comprehensive discrimination is performed based on the discrimination result of the edit discrimination processing routine 23 and the discrimination result of the comparison discrimination processing routine 26, and whether the bill P is right or wrong is output as the comprehensive discrimination result.

In this way, when performing the pattern identification processing by the image signal obtained from the optical system, the identification area is set with both the leading end reference and the trailing end reference with respect to the conveyance direction of the banknote,
A two-dimensional sample pattern is created for each of the set identification areas, the standard pattern is used for each of the created sample patterns, the similarity is calculated, and candidates are extracted. Finally, the front and back patterns (types) of banknotes are identified by performing the three-stage editing process using the editing information in the editing information table based on the candidate bill type, the direction, and the degree of similarity thereof. .

As a result, even if a part corresponding to an identification area of a banknote is damaged or is soiled and it is impossible to extract the ticket type candidates, it is possible to use the information in the remaining identification area that is not damaged or soiled. The pattern can be identified by the editing process. Therefore, correct pattern identification can always be performed, an identification error does not occur and the identification cannot be made, and the identification performance is significantly improved as compared with the conventional art.

In addition, in the said Example, although the case where the banknote discrimination device discriminate | determines the front-back pattern of a banknote was demonstrated, this invention is not limited to this, For example, securities other than banknotes, such as a check, and other paper sheets. It can also be applied to identify front and back patterns of.

[Effects of the Invention] As described in detail above, according to the present invention, for example, even if a part of a paper sheet is damaged or soiled,
It is possible to always provide correct pattern identification, and to provide a paper sheet pattern identification device that does not cause an identification error or become unidentifiable.

[Brief description of drawings]

The figure is for explaining one embodiment of the present invention. FIG. 1 is an overall configuration diagram, and FIG. 2 is a pattern from a surface pattern of a banknote stored in an image storage unit to a sample pattern. FIG. 3 (a) is a processing diagram, FIG. 3 (a) is a diagram showing a sample pattern example of one identification area of a bill, and FIG.
FIG. 4 shows an example of a standard pattern for one bill type, FIG. 4 is a flow chart for explaining an edit discrimination process, and FIG. 5 is a diagram for explaining a conversion condition when a similarity is converted into a weight value in the first edit process. FIG. 6 is a diagram showing an example of output information of the candidates extracted in the second editing process, FIG. 7 is a diagram explaining the processing contents in the editing mode in the third editing process, and FIG. It is a figure explaining the discrimination area when performing. P ... bill (paper sheet), A _1, A ₂ ... identification area, 1 ... conveyor belt, 2 _1, 2 ₂ ... optical system, 6 _1, 6 ₂ ... A / D conversion section, 7 _1, 7 ₂ ...
Image storage unit, 10 ... Logical operation unit, 11 ... Standard pattern table, 12 ... Editing information table (storage unit), 21 ... Sample pattern creation processing routine, 22 ... Pattern similarity calculation processing routine (identification processing unit), 23 ... Editing discrimination processing routine (editing processing means).

Claims (1)

(57) [Claims] 1. A conveying means for conveying a paper sheet, and a conveying means arranged to face a front surface and a back surface of the paper sheet conveyed by the conveying means, and correspond to a pattern on the front surface and the rear surface of the paper sheet. The optical means for obtaining each of the image signals and a plurality of identification areas to be subjected to the identification processing for each of the front surface and the back surface of the paper sheet are set, and the setting is performed from the image signal of the paper sheet obtained by the optical means. Sample pattern creating means for creating a pattern for each of a plurality of identification areas, and a plurality of types of paper sheets to be transported by the transport means, a standard for storing in advance standard patterns corresponding to each of the identification areas Pattern storing means, all kinds of standard patterns stored in the standard pattern storing means, and patterns for each identification area created by the sample pattern creating means. Based on the pattern similarity calculation processing means for performing the similarity calculation with the turn, based on the calculation result by the pattern similarity calculation processing means, as a candidate of the paper sheet subjected to the processing by the optical means, each of the identification areas First editing processing means for independently extracting a plurality of types of paper sheet candidates in descending order of similarity, and the front and back surfaces of the paper sheets based on the extraction result by the first editing processing means. Second edit processing means for independently extracting a plurality of types of paper sheet candidates for each, and paper sheet candidates as identification results from the plurality of candidates extracted by the second edit processing means. A paper sheet pattern identification apparatus comprising: a third edit processing unit for extracting one.