JPH04291482A - Pattern discriminating method - Google Patents

Abstract

PURPOSE:To shorten the discrimination time for pattern discrimination by comparing the binarized picture of a discrimination object pattern Gy with respective binarized pictures of B kinds of reference pattern Gx. CONSTITUTION:Reference patterns Gx corresponding to peculiar binarization levels Lx are binarized by these levels Lx so that the number of binary quantities W1 whose values are '1' is Ax, thereby obtaining B kinds of reference binarized pictures Px. One discrimination object pattern Gy is binarized by C kinds of binarization level Ly to obtain C pairs of measured binarized pictures Py and numbers Ay of binary quantities W2 which constitute these pictures Py and have binary quantity value '1', and the picture Py which pairs with Ay most approximating Ax is compared with B kinds of Px to discriminate the classification of Gy.

Description

[Detailed description of the invention]

0001

[Industrial Field of Application] The present invention relates to a pattern discrimination method used in the case where images of patterns on banknotes are binarized to identify the denomination and authenticity of the banknotes.
The present invention relates to a method capable of achieving high discrimination accuracy and shortening the discrimination time.

0002

[Prior Art] Conventionally, when distinguishing the denomination and authenticity of a banknote, a binary image obtained by binarizing the local brightness of a pattern on the banknote at a predetermined binarization level and a reference image are used. The above patterns may be determined by comparison, but
In this case, even if the pattern is the same, there is a difference in the local brightness of the pattern based on the newness and oldness of the banknote, as can be seen between official notes that are uncirculated banknotes and circulation notes that have been in circulation for a considerable period of time. Naturally, the binarized image becomes a different image, making it impossible to accurately identify the pattern. Therefore, it is necessary to correct the local brightness as described above. It is being carried out to determine the Note that changes in the local brightness of the pattern on a banknote depending on whether the banknote is old or new usually occur at the same rate in all parts of the pattern. is sometimes referred to as the overall density of the pattern.

That is, FIG. 3 is an explanatory diagram of the configuration of the main parts of a conventional banknote validating machine 1. In the figure, 2 conveys banknotes 3 placed on a flat conveyance surface 2a linearly in the direction of the arrow H. The conveyance paths 4 arranged in this manner are fixed to the housing 1a of the discriminator 1 so as to be arranged along a straight line orthogonal to the direction of the arrow H, and are arranged to face the banknotes 3 conveyed by the conveyance path 2. The plurality of brightnesses are arranged above the conveyance path conveyance surface 2a so that the local brightness of a pattern placed on the conveyance surface 2a is detected as the brightness of a portion of an arbitrary pattern. All of them are reflective optical sensors as detectors, and 5 is a plurality of reflective optical sensors that are virtually set to be orthogonal to the direction of the arrow H to the observation surface 7, which is a part of the conveyance surface 2a on which the banknotes 3 are placed. A linear first test line 6 indicates a portion of the observation surface 7 where the pattern whose brightness is detected by the sensor 4 is not placed, as the observation surface 7 is moved by the conveyance path 2. This is a second test line as a locus formed on the observation surface 7. Then, Ri is the intersection of the test lines 5 and 6, Qi is the two-dimensional position of the intersection Ri on the observation plane 7, and the discriminator 1 detects the part of the above pattern on the observation plane 7 at the position Qi, that is, The sensor 4 is configured to detect the brightness of the pattern portion of the banknote on the surface 3a of the banknote at the two-dimensional position Qi, and obtain brightness data Di as a measured value of this brightness.

[0004] The discriminating machine 1 is constructed as described above, and in this discriminating machine 1, the conveying path 2 is
The pattern on the surface 3a of the banknote as the discrimination target pattern Gy placed on the conveying surface 2a of the banknote 3 is determined according to the denomination of the banknote 3, the front and back sides of the banknote 3, and the orientation of the pattern with respect to the direction of the arrow H. That is, as shown in FIG. 4, the discriminator 1 first sets the position Qi of the intersection Ri in the first step T1 as described above. Then, in the second step T2, the denomination to be determined, the degree of newness and oldness of the banknote, and the banknote surface 3 are determined.
Obtain the measured value Xi as the brightness data Di at all positions Qi on the surface 3a for a large number of N banknotes 3, each with a different state of the front and back sides of the banknote, whether a is the front or back side of the banknote. Then, the total value of Xi in the N banknotes 3 is divided by N to determine the reference value Sx of the overall density on the banknote surface 3a, and in a third step, the total value of Xi in the N banknotes 3 is determined. From 3, banknote side 3a
The above-mentioned types of patterns are classified by denomination, front and back, and different orientations of the pattern B types (here, B represents a plural number).
For each type of pattern F that requires discrimination, select a reference pattern Gx as the pattern F having the density reference value Sx or a total value of Xi close to this Sx,
After that, if the brightness data Xi is equal to or higher than the appropriately set binary level La, the value of the binary amount W3 in this case is set to "1".
The binarization by the above La is one reference pattern Gx
The value of the binary quantity W3 obtained for Xi in the image memory M3 corresponding to this one pattern Gx is
All Xi in the one pattern Gx are stored in the storage area corresponding to the above-mentioned position Qi where i was obtained.
B type reference pattern Gx
For each case, a reference image Pxa is determined as a collection of binary quantities W3 exhibiting a value of "1" in the memory M3, and as a result, each of the B types of Pxa can be easily distinguished from one another. The binarization level La is determined to be the binarization level Lb so that an image with characteristics is obtained, and B type image P when Xi is binarized using this Lb.
xa is determined as the final B-type reference image Pxb.

Next, in a fourth step T4, brightness data D at all positions Qi in the pattern on the surface 3a of the banknote 3 placed on the conveyance surface 2a as the discrimination target pattern Gy is determined.
The brightness data Yi as i is obtained by the sensor 4, and finally the total value Sy=ΣYi of all Yi is calculated,
After that, in the fifth step T5, the correction coefficient K is calculated as K=Sx using Sx determined in step T2 and Sy calculated in step T4.
/Sy is calculated, and in the sixth step T6, step T
Using Yi obtained in step 4 and K calculated in step T5, corrected data Zi is calculated as Zi=K·Yi. Then, in the seventh step T7, the above data Zi is transferred to step T3.
If it is equal to or higher than the binarization level Lb determined in , the value of the binary quantity W4 in this case is set to "1". Binarization is performed using Lb of Zi, and the binary quantity W4 obtained by this binarization is By storing the value in the image memory M4 corresponding to the above-mentioned image memory M3, a measurement image Pz as a collection of binary quantities W4 exhibiting a value of "1" in the memory M4 is obtained, and then in an eighth step T8. , the image Pz is compared with the B type reference image Pxb obtained in step T3, and the B type Pxb is determined.
Px that matches or most closely matches Pz from among
Image Pxb0 as image Pxb0 is selected, and it is determined that the above-described discrimination target pattern Gy belongs to the type of pattern F corresponding to image Pxb0 among the B types of patterns F that require discrimination.

[0006]

[Problem to be Solved by the Invention] In the discriminating machine 1, the discrimination target pattern Gy is discriminated as described above, so in this case, the overall density Sy, which varies depending on whether the banknote of the pattern Gy is old or new, is corrected to the reference value Sx. However, in this case, since the number of optical sensors 4 is several, the brightness data Y
Usually, the number of i is small, and as a result, in the pattern discrimination method described above, the accuracy of pattern discrimination is low because both the measurement image Pz as a binary image and the B type reference image Pxb are not precise. In addition, after obtaining all the data Yi in the discrimination target pattern Gy in step T4, the measurement image Pz is
The problem with this latter problem is that in order to solve the former problem, all the optical sensors 4 are set to a minute brightness corresponding to this sensor 4. When the detectors are replaced with a single one-dimensional image sensor having a configuration in which a large number of detectors are closely arranged in a straight line, it is equivalent to providing a large number of optical sensors 4,
This is particularly noticeable.

[0007] An object of the present invention is to measure the pattern Gy whose overall density has been equivalently corrected at the time when all the brightness data Yi in the pattern Gy to be determined has been measured.
Even if the formation of the binarized measurement image Py based on the above-mentioned one-dimensional image sensor is increased and a precise binarized image is obtained by increasing the number of data Yi using the above-mentioned one-dimensional image sensor, etc., the time is short. The purpose is to enable pattern discrimination based on time.

[0008]

[Means for Solving the Problems] In order to achieve the above object, according to the present invention,

1) Two-dimensional positions Qi of each of a plurality of fixed points on a planar observation surface and the respective positions Q
Each brightness data Xi as a measurement value of the local brightness of the pattern on the observation surface measured at
If the binarization level Ly for binarizing with the binarization level Ly of C types representing multiple types for each, and the said Xi is equal to or higher than the unique binarization level Lx, the binary quantity in this case is The first binary quantity exhibiting the value of "1" obtained by performing binarization of the Xi by the Lx with the value of the first binary quantity as "1" for all the Xi in the pattern. A first step of setting the value Ax of the number α, and setting the Xi at each of all the positions Qi of the reference pattern for each of the B types of reference patterns as the B types of patterns representing a plurality of types. and the value of the number α in each of the reference patterns is the Ax
a second step of determining the Lx of type B so that it is equal to
The first binary amount obtained by binarizing with the Lx set to this reference pattern is stored in a storage area corresponding to the position Qi where the Xi was obtained in the first image memory corresponding to this reference pattern. By performing the operation on all the Xi in the one reference pattern, the "
a third step of performing an operation for each of the B types of the reference patterns to obtain a reference image Px as a collection of the first binary quantities on the first image memory exhibiting a value of 1; and the observation. After measuring the local brightness at the position Qi in the discrimination target pattern placed on the surface and obtaining the brightness data Yi as the measured value, if this Yi is greater than or equal to the Ly, the binary quantity in this case Immediately after the Yi is determined, binarization of the Yi by the Ly is performed so that the value of the second binary quantity is "1", and the second binary quantity is converted to the second binary quantity corresponding to the discrimination target pattern. The second binary quantity exhibiting the value of "1" is stored in the storage area corresponding to the position Qi where the Yi is obtained in the two-image memory for all the Qi in the discrimination target pattern. The set of the measurement image Py as a collection on the second image memory of , and the number Ay of the second binary quantity exhibiting the value of "1" constituting this Py is defined as the L of the C types.
A fourth step of determining each of the y, and selecting the number Ay0 closest to the Ax from the C types of Ay, and then measuring the Py that forms the same group as this Ay0. The image Py0 is compared with the Px of the B type, and the discrimination target pattern is determined as the Py
a fifth step of determining that the pattern belongs to the reference pattern that produced the reference image Px that is most similar to 0;

2) In the method described in item 1) above, the fourth step is performed by measuring the local brightness at a position Qi in the pattern to be discriminated placed on the observation surface and obtaining brightness data Yi as this measurement value. Then, if this Yi is equal to or higher than the binarization level Ly, the value of the second binary quantity as the binary quantity in this case is set to "1", and the binarization of the Yi by the Ly is performed as the Yi Immediately after the
A binary quantity is stored in a storage area corresponding to the position Qi where the Yi is found in the second image memory corresponding to the discrimination target pattern, and the "1" appears as the value of the second binary quantity. The work of integrating this "1" each time as a number 1 is performed for all the Qs in the discrimination target pattern.
By performing this for i, the value of the second second quantity is “1”.
" measurement image P as a collection on the second image memory
y and the value “1” of the second binary quantity as a result of the integration.
The task of simultaneously finding the number Ay of C types of the above Ly
The pattern discrimination method is configured so that each of the steps is performed in parallel, and furthermore,

[0011]
3) In the method described in either 1) or 2) above, the position Qi of the fixed point is determined by a number of brightness detectors arranged in a straight line, each of which detects the local brightness of a pattern placed on the observation surface. each position of the brightness detector in a one-dimensional image sensor arranged in and relative movement in a direction perpendicular to the linear arrangement of the brightness detector between the observation surface and the one-dimensional image sensor; The pattern discrimination method is configured to be set based on the following.

0012

[Operation] With the above configuration, the pattern discrimination method can be used in either the discrimination method that does not perform the formation of the measurement image and the integration of the number of the value "1" of the second binary quantity in parallel, or the discrimination method that does. Also in the case of , the formation of all C types of measurement images Py is completed at the time when all the brightness data Yi in the discrimination target pattern Gy have been measured, and in this case, the second binary quantity is selected from among these Py. Py selected through the number Ay of the value "1" matches the measurement image Pz in the conventional pattern discrimination method described above based on the determination target pattern Gy to which the correction of the count K has been added for the overall density. Or, since it is clear that the image is similar to this Pz, pattern Gy can be determined in a short time, and in the case of a method using a one-dimensional image sensor, the number of fixed points at position Qi is Since the number of patterns increases, precise pattern discrimination can be performed in a short time.

[0013]

[Embodiment] FIG. 1 is a flow chart explaining the configuration of a pattern discrimination method 8 according to the present invention, and FIG. The difference between FIG. 2 and FIG. 3 is that a large number of minute brightness detectors each detect the local brightness of an arbitrary pattern placed on the observation surface 7 as part of the conveyance surface 2a of the conveyance path 2. One-dimensional image sensors 10 are arranged closely in a straight line in a direction perpendicular to the direction of the arrow H, and are fixed to the housing 9a of the discriminator 9 in correspondence with the optical sensor 4 in FIG. In addition, the second test line 11 corresponding to the second test line 6 in FIG. surface 7 so as to match the locus formed on this surface 7 as the observation surface 7 moves by
In this case, Qi is set to test lines 5 and 1.
This is the two-dimensional position of the intersection point Ri with 1 on the observation plane 7.

Next, a pattern discrimination method 8 as a method for discriminating banknotes 3 performed in the discriminator 9 will be explained with reference to FIGS. 1 and 2. That is, in determination method 8, as shown in FIG. 1, in the first step V1, the position Qi of the intersection Ri is set as described above, and the brightness detection of the image sensor 10 is performed at the position Qi. Each brightness data Xi, which is a measured value of the local brightness of the pattern on the observation surface 7 measured using a device, is binarized using C types of binarization levels Ly representing a plurality of types for each Xi. If the binarization level Ly of these C types is set for the purpose of The value Ax of the number α of W1 exhibiting a value of “1” obtained by performing binarization of Xi with Lx with a value of “1” for all Xi in the pattern on the observation surface 7 is set. Then, here, the binarization level Ly always changes Xi to a plurality of levels even if the overall density of the discrimination target pattern Gy as a pattern on the surface 3a of the banknote 3 changes based on the newness or oldness of the banknote 3. This is a binarization level that can be binarized with .

[0015] Then, in the determination method 8, the second step V2
Then, as in the conventional discrimination method described above, for each pattern F of convenience B type that requires discrimination, a reference pattern Gx as a pattern F exhibiting an arbitrary overall density is determined. For each pattern Gx, the brightness data Xi at each position Qi of Gx is measured by the sensor 10, and the value of the number α explained in the above procedure V1 in each pattern Gx is determined to be Ax set in V1 as well. The above-mentioned binarization level Lx of convenience B type is set so that it is equal to the third
In step V3, the aforementioned first binary value obtained by binarizing Xi in one Gx of the B type Gx determined in step V2 at the binarization level Lx set for this Gx in step V2. The value of the amount W1 is stored in the first image memory M1 corresponding to this Gx.
By performing the operation of storing the above Xi in the storage area corresponding to the position Qi where the above Xi is obtained for all the Xi in the one Gx, the standard as an aggregate on the memory M1 of W1 exhibiting a value of "1" is obtained. B to obtain the image Px
Execute for each type of Gx.

Then, in a fourth step V4, the sensor 10 measures the local brightness at the position Qi in the pattern on the surface 3a of the banknote 3 as the discrimination target pattern Gy placed on the observation surface 7. After obtaining the brightness data Yi as the measured value, if this Yi is greater than or equal to Ly set in step V1, the value of the second binary quantity W2 as the binary quantity in this case is set to "1". Yi
Immediately after is obtained, the value of binary quantity W2 is determined by pattern G.
The process of storing all Qi in the pattern Gy in the storage area corresponding to the position Qi where the above-mentioned Yi was obtained in the second image memory M2 corresponding to y is performed, and the memory M2 of W2 exhibiting the value "1" is stored. After forming the measurement image Py as a collection of Py, the number Ay of W2 value "1" constituting this Py is calculated, and the task of finding a pair of this Ay and the above Py is performed using the C type set in step V1. In the fifth step V5, select the number Ay0 as the number Ay closest to the number Ax set in step V1 from among the number Ay of convenience C type obtained in step V4, and then The measurement image Py0, which is the image Py that forms a pair with Ay0 in the image Py of type C, obtained in step V4, is compared with the reference image Px of type B, obtained in step V3, to determine pattern G.
It is determined that y is a pattern F that belongs to the reference pattern Gx that produced the image Px that is most similar to Py0. Then, the pattern discrimination method 8 shown in FIG.
It is composed of V5.

Discrimination method 8 is configured as described above, and the measurement image Pyo compared with the B type reference image Px in step V5 is an image whose number Ayo matches or approximates the number Ax in Px. In this pattern discrimination method 8, image comparison is performed after correcting the overall density of the discrimination target pattern Gy from which Pyo was obtained to match or approximate the overall density of the reference pattern Gx from which Px was obtained. It is clear that it is equivalent to the conventional method described above, and in addition, since the image sensor 10 is used in the discriminator 9 that adopts this discrimination method 8, the number of points at the position Qi is smaller than that in the discriminator 1 described above. It is clear that there are many more. Therefore, the discrimination device 9 can precisely and accurately discriminate the pattern on the side 3a of the banknote, but as is clear from the above, in the discrimination method 8, all the brightness data in the pattern Gy Since the formation of all C type measurement images Py is completed immediately after the measurement of Yi is completed, it is better to use the above-mentioned conventional discrimination method of measuring all the data Yi and then performing density correction to form the measurement image Pz. Also, pattern discrimination can be performed in a shorter time using discrimination method 8.

In determination method 8, the number Ay is obtained from the measurement image Py after forming the measurement image Py in step V4, but in the present invention, as soon as the data Yi is obtained, this Yi is binarized at the level Ly. Py There is no problem in performing the work in which the formation of and the calculation of Ay are completed at the same time for each of the C types of Ly, and if the pattern discrimination method is configured in this way, discrimination method 8 It is clear that pattern discrimination can be performed in an even shorter time than in .

[0019] Although the explanation of each of the above-mentioned embodiments was related to the discrimination of banknotes, it is clear that the method of the present invention can also be applied to the discrimination of patterns other than banknotes, and In the banknote validator 9, the observation surface 7 is moved relative to the fixed pattern 10, but in the banknote validator that adopts the pattern discrimination method of the present invention, the observation surface 7 is fixed and the sensor 10 is moved. It is also clear that the discriminator may be configured to do so.

[0020]

[Effects of the Invention] As mentioned above, in the present invention,

[
1) Each two-dimensional position Qi of a plurality of fixed points on a planar observation surface, and each brightness data as a measurement value of the local brightness of a pattern on the observation surface measured at each position Qi. The binarization level Ly for binarizing Xi at C types of binarization levels Ly representing multiple types for each Xi, and Xi is the unique binarization level Lx.
If the above is the case, the value of "1" obtained by performing binarization of Xi using the Lx, where the value of the first binary quantity as the binary quantity in this case is "1", is performed for all Xi in the pattern. a first step of setting the value Ax of the number α of the first binary quantities exhibiting the value; and a first step of setting the value Ax of the number α of the first binary quantities exhibiting the value; A second step of determining Lx of convenience B type so that the value of the number α in each reference pattern is equal to the Ax by finding Xi in each of Qi, and In the one reference pattern, the first binary amount obtained by binarizing with the set Lx is stored in the storage area corresponding to the position Qi where Xi was obtained in the first image memory corresponding to this reference pattern. All Xi
A third step is to obtain a reference image Px as an aggregate on the first image memory of the first binary quantity exhibiting a value of "1" by performing the operation for each of the B types of reference patterns, and observation. After measuring the local brightness at the position Qi in the discriminant symmetric pattern placed on the surface and obtaining the brightness data Yi as this measurement value, this Yi
is greater than or equal to Ly, the second value as a binary quantity in this case
Y
The process of storing the second binary quantity in the storage area corresponding to the position Qi where Yi was obtained in the second image memory corresponding to the discrimination target pattern, which is performed immediately after the determination of Yi, is carried out immediately after the determination target pattern is determined. The measurement image Py as a collection of second binary quantities exhibiting a value of "1" on the second image memory and the number of second binary quantities exhibiting a value of "1" constituting this Py. A fourth step of finding a pair with Ay for each of the C types of Ly, and selecting the number Ay0 as the Ay closest to Ax from the C types of Ay, and then selecting the same pair as this Ay0. a fifth step of comparing the measurement image Py0 as the Py formed by Py with the B types of the Px, and determining that the pattern to be determined is the pattern belonging to the reference pattern that produced the reference image Px that is most similar to Py0; The pattern discrimination method is configured to consist of
Also,

2) In the method described in item 1) above, the fourth step is performed by measuring the local brightness at the position Qi in the discrimination target pattern placed on the observation surface and obtaining brightness data Yi as this measurement value. Then, if this Yi is equal to or higher than the binarization level Ly, the value of the second binary quantity as the binary quantity in this case is set to "1", and Yi is binarized by Ly. Immediately after that, the second binary quantity is stored in the storage area corresponding to the position Qi where Yi of the second image memory corresponding to the discrimination target pattern was determined, and the above-mentioned "
1'' appears as the value of the second binary quantity.By performing the work of integrating this ``1'' as a number 1 for all Qi in the discrimination target pattern, the value ``1'' of the second binary quantity can be calculated. The work of simultaneously obtaining the measurement image Py as a collection on the second image memory and the number Ay of the value "1" of the second binary quantity as the result of the integration is performed in parallel for each of the C types Ly. The pattern discrimination method is configured to be a work step to be performed, and furthermore,

3) In the method described in either 1) or 2) above, a large number of brightness detectors each detect the position Qi of the fixed point and the local brightness of the pattern placed on the observation surface. each position of the brightness detector in a one-dimensional image sensor arranged linearly, and relative movement in a direction perpendicular to the linear arrangement of the brightness detector between the observation surface and the one-dimensional image sensor. The pattern discrimination method was configured to be set based on the following.

[0024] Therefore, with the above configuration, either the pattern of the discrimination method that does not perform the formation of the measurement image and the integration of the number of values "1" of the second binary quantity in parallel, or the discrimination method that does In the case of the discrimination method, C
The formation of all types of measurement images Py is completed, and in this case, the number A of the second binary quantity value "1" is determined from among these Py.
Py selected via y is the measurement image Pz in the conventional pattern discrimination method described above based on the determination target pattern Gy to which the overall density has been corrected by the coefficient K.
Since it is clear that the image matches or approximates Pz, the present invention has the effect of being able to identify the pattern Gy in a short time, and also has the advantage of being able to identify the pattern Gy in a short time. In this case, since the number of fixed points at the position Qi increases, the present invention has the advantage that precise pattern discrimination can be performed in a short time.

Furthermore, in the method of the present invention, as is clear from the above description, the reference pattern Gx is determined from the pattern F exhibiting an arbitrary overall density for each of the B types of patterns F. On the other hand, in the conventional pattern discrimination method described above, as explained in the steps T2 and T3 in the discrimination method adopted in the discrimination device 1 described above, the overall density is first determined from an extremely large number of patterns F. Since it is necessary to perform the very labor-intensive work of determining the reference value Sx of the overall density as the average value of The pattern discrimination method of the invention also has the effect that it is easy to determine the reference pattern Gx, which simplifies the preparation work for pattern discrimination.

[Brief explanation of the drawing]

FIG. 1 is a flowchart explaining the configuration of the method of the present invention.

FIG. 2 is an explanatory diagram of the configuration of the main parts of a banknote validator that adopts the pattern discrimination method whose flowchart is shown in FIG.

[Fig. 3] An explanatory diagram of the configuration of the main parts of a conventional banknote validating machine

[
FIG. 4 is a flowchart illustrating the configuration of a conventional pattern discrimination method employed in the banknote validator whose main configuration is shown in FIG. 3

[Explanation of symbols]

7 Observation surface 8 Pattern discrimination method 10 One-dimensional image sensor

Claims (3)

[Claims] 1. Two-dimensional positions Qi of each of a plurality of fixed points on a planar observation surface, and each value as a measurement value of the local brightness of a pattern on the observation surface measured at each of the positions Qi. The binarization level Ly for binarizing the brightness data Xi at C types of binarization levels Ly representing a plurality of types for each Xi, and if the Xi is equal to or higher than the unique binarization level Lx The first as a binary quantity in case
The number α of the first binary quantities exhibiting the value of “1” obtained by performing binarization of the Xi by the Lx with the value of the binary quantity as “1” for all the Xi in the pattern a first step of setting the value Ax, and determining the Xi at each of all the positions Qi of the reference pattern for each of the B types of reference patterns as the B types of patterns representing a plurality of types. a second step of determining the Lx of convenience B type so that the value of the number α in each of the reference patterns is equal to the Ax; and the Lx in which the Xi in one of the reference patterns is set to this reference pattern. The operation of storing the first binary amount obtained by binarizing in the storage area corresponding to the position Qi where the Xi was obtained in the first image memory corresponding to this reference pattern is performed for all the operations in the one reference pattern. The above "1" by performing for the above Xi
a third step of obtaining a reference image Px as an aggregate on the first image memory of the first binary quantity exhibiting the value of for each of the B types of reference patterns; After measuring the local brightness at the position Qi in the placed discrimination target pattern and obtaining the brightness data Yi as the measured value, if this Yi is greater than the Ly, the binary quantity in this case is determined. Immediately after the Yi is obtained, binarization of the Yi by the Ly is performed to set the value of the second binary quantity to "1", and the second binary quantity is converted into a second image corresponding to the discrimination symmetric pattern. Y of memory
The operation of storing in the storage area corresponding to the position Qi where i was obtained is performed for all the Qs in the discrimination symmetric pattern.
i, the measurement image Py as a collection on the second image memory of the second binary quantity exhibiting the value of “1” and the measurement image Py exhibiting the value of “1” constituting this Py. 2. A fourth step of obtaining a pair with the number of binary quantities Ay for each of the C types of the Ly, and the C type of the A.
The number A as the Ay closest to the Ax from y
y0 is selected, and then the measurement image Py0 as the Py forming the same group as this Ay0 and the P of the B type are selected.
a fifth step of comparing the discrimination symmetric pattern with x and determining that the pattern belongs to the reference pattern that produced the reference image Px that most approximates the Py0;
A pattern discrimination method characterized by comprising: 2. In the method according to claim 1, the fourth step is performed by determining the position Q in the discrimination target pattern placed on the observation surface.
Measure the local brightness at point i to obtain brightness data Yi as this measured value, and then if this Yi is equal to or higher than the binarization level Ly, the second binary quantity as the binary quantity in this case is determined. Immediately after the Yi is determined, binarization of the Yi by the Ly with a value of "1" is performed, and the second binary amount is determined by the Yi of the second image memory corresponding to the discrimination target pattern. All operations in the discrimination target pattern include storing the data in a storage area corresponding to the position Qi, and adding up this "1" as a number 1 each time the "1" appears as the value of the second binary quantity. By performing this on the Qi of the second binary quantity, the measurement image Py as a collection of the second binary quantity value "1" on the second image memory and the value of the second binary quantity "1" as the result of the integration are obtained. 1. A pattern discrimination method characterized in that the task of simultaneously determining the number Ay of "1" is performed in parallel for each of C types of Ly. 3. The method according to claim 1, wherein a plurality of brightness detectors each detecting the position Qi of the fixed point and the local brightness of the pattern placed on the observation surface are arranged in a straight line. each position of the brightness detector in a one-dimensional image sensor arranged in and relative movement in a direction perpendicular to the linear arrangement of the brightness detector between the observation surface and the one-dimensional image sensor; A pattern discrimination method characterized by setting based on.