JPS6292090A - Pattern recognizing device - Google Patents

Abstract

PURPOSE:To produce a dictionary having a learning function by applying a prescribed pre-processing to an input standard pattern and producing the correction and discrimination functions of the standard order logic according to said pre-processing. CONSTITUTION:A CPU 1 executes a program stored in a memory 4 and a standard pattern is supplied from a scanner 2 according to this program. Thus the pre-processing is carried out in the working area of a memory 5 such as the picture input processing, the broken line approximation processing and the standard order processing. Then a dictionary producing memory 6 corrects the standard order logic supplied from a console 3 based on the picture data pre-processed by the memory 5 and produces the discrimination function for the production of dictionaries I and II. Thus it is possible to produce a dictionary having a learning function.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to identification processing using sequential logic and identification functions in a pattern recognition device.

The present invention also relates to a pattern recognition device having a learning function.

The present invention relates to automatic creation of a dictionary when developing OCR.

The present invention is a pattern recognition device particularly having a learning function, and includes preprocessing such as pattern size normalization processing, thinning processing, and Cheno coding; This method consists of intermediate classification based on comparison with other people, and identification processing using a discriminant function.

The present invention relates to sample pattern input processing for learning.

[Prior art]

BACKGROUND ART Many studies have been conducted on character recognition, and various recognition methods and devices have been put into practical use. However, development such as creating a dictionary was difficult, and it was also difficult to eliminate misreading. However, since it does not have a learning function, similar misreadings may be repeated, resulting in lower recognition rates and lower device usage efficiency, which is one of the reasons why recognition devices are difficult to spread.

Also, one of the recognition methods is, for example, a method of analyzing the structure of the strokes of a pattern, but one of the other recognition methods is
Although this method has the advantage of achieving a higher recognition rate than other methods such as the similarity method, it has the disadvantage that it is difficult and time-consuming to create a dictionary. Therefore, automatic dictionary creation for the structural analysis method is important in terms of recognition rate, and research on learning functions is required.

Although the known technology and device are not related to the learning function,
For example, Hitachi, Ltd.'s handwritten character recognition device (H-8
959OCR), which performs the following processing on character patterns.

In other words, the character pattern is converted into a line diagram, represented by a directional code called a chain code, and based on that code,
Identification is made by checking the order logic of standard patterns in a dictionary. In addition, a restriction is placed on the number of times the state transition state is passed, thereby improving the recognition rate.

However, depending on the character, the starting point of the chain code cannot be uniquely determined, and there are many things that cannot be expressed using a single sequential logic.For example, it is said that approximately 100 sequential logics are required for handwritten digits. The disadvantage is that it takes a lot of time to create and recognize.

As is clear from the above, the creation of a dictionary with a learning function in character pattern recognition.

Currently, research on learning during recognition is desired.

[Problem that the invention seeks to solve]

As is clear from the above points, an object of the present invention is to eliminate the above-mentioned drawbacks.

Another object of the present invention is to provide a learning function in the development of OCR or the creation of a dictionary during pattern recognition.

Another object of the present invention is to provide a pattern recognition device with a learning function during recognition.

Another object of the present invention is to provide a pattern recognition device that incorporates the idea of sequential logic (automata) into a stroke structure analysis method and further performs a classification process using a classification function.

An object of the present invention is to automatically create a dictionary for online or offline character recognition.

An object of the present invention is to create a chain code for dictionary creation based on a chain code for pattern recognition.

The purpose of the present invention is to input standard sequential logic.

An object of the present invention is to modify the sequential logic, which is a dictionary, based on the chain code, and to learn better sequential logic.

Another object of the present invention is to create a discriminant function whose variable is the number of times each state of sequential logic, which is dictionary data, is passed.

[Means for solving problems]

As a means of solving this problem, for example, the pattern input device shown in the block diagram of FIG. , the dictionary information is corrected and created in the memory 6, and transferred to the memory 7. The memory 7 stores the sequential logic and the address of the discriminant function used in pair with the sequential logic.

Memory 8 is a part that stores identification functions.

The sequential logic of the specimen is input in the form of a pattern by the off-line input attachment of the scanner 2.

[For production]

In the configuration shown in FIG. 1, the CPUI executes a program stored in the memory 4, preprocesses image information read from the scanner 2 in the memory 5, and performs key input from the console 3 or external memory. The basic sequential logic from the optical disk or floppy disk is modified using the preprocessed data, and a discriminant function is created.

;Example〕 One embodiment of the present invention will be described below.

FIG. 1 is a block diagram of an apparatus for implementing the present invention. 4 is a memory in which a program for carrying out the present invention is stored.

1 is a CPU, which controls and operates other parts of the device according to a program stored in the memory 4.

Reference numeral 2 denotes a scanner, which converts image data into information that can be processed by a computer using optical reading means.

Reference numeral 3 denotes a console, which consists of a keyboard and a display, and includes means for interfacing with the CPU such as inputting data.

A preprocessing memory 5 is a memory area for storing image data read by the scanner 2 and performing preprocessing such as image processing and coding on the image data.

Reference numeral 6 denotes a memory for creating a dictionary, which stores basic sequential logic input from the console 3 or a floppy disk.

Based on the image data preprocessed in the memory 5, standard sequential logic (or
(You may also read out those registered in external memory such as a floppy), create a discrimination function, and create a dictionary I.
, create II.

A memory 7 stores the sequential logic that is part of the dictionary, and stores the sequential logic corrected in the memory 6 and the address of the discriminant function used in pair with it.
).

Reference numeral 8 denotes a memory for storing the discrimination function, which is another part of the dictionary.The discrimination function created in memory 6 and the recognition result identified using this are stored, and the rest of the dictionary for recognition section (dictionary ■).

[Operation wave diagram]

FIG. 2 is a flowchart showing the operation of the present invention.

A more detailed explanation of the operation will be given using FIG.

The flow of operations will be explained below.

In this embodiment, patterns belonging to the same category are used intensively, a dictionary is created for that pattern, and then dictionaries are created for patterns belonging to other categories.

First, sample patterns belonging to the same category are input from the scanner 2 described above. In the image input process of step 17, one pattern to be subjected to pattern recognition is cut out from the image read by the scanner 2, which is an image league such as a digital copying machine, hand scanner, word processor, facsimile, etc.

In the preprocessing of step 18, z-value processing is performed.

It consists of size normalization processing and line thinning processing.

In the broken line approximation process of step 19, the thinned image is subjected to the broken line approximation process using a predetermined finite number of line segments, for example, line segments having the directions of the 8-direction code shown in FIG. 321. At this time, the direction of each line segment of the pattern approximated by a broken line can be represented by an 8-direction code. In the standard sequential logic of step 5, an example of which is shown in FIG. 19, transition conditions connecting states corresponding to each line segment in the sample pattern approximated by a broken line are determined. 1st
A more detailed explanation will be given based on FIGS. 9 and 4.

Condition AI corresponds to five-way code 3°2.1.8 and 7 for each line segment in FIG.

A2. A3. A4. Provide A5.

Next, the transition conditions between each state are determined by the connection state of the line segments.
TERA1 and A2 are connected, and the direction code 2 causes a transition from A1 to A2.

Similar mA2 and A3 are from A2 to A3 by direction code 1.
Transition to. Same below.

In addition, in this embodiment, each state also allows a transition to itself. For example, in states AI and A2, each state transitions to itself with a direction code of 3.2. From 9 or above, the sample broken line approximation pattern of FIG. 19 is shown in FIG. The standard state transition diagram shown in is obtained.

The above image input processing, preprocessing, broken line approximation processing, and standard order processing are processed in the working area of the memory 5,
According to step 6, the standard sequential logic is stored in memory 7.

In step 1, image input processing, 1
Performs processing to cut out two patterns.

The preprocessing in step 2 consists of binarization processing, size normalization processing, thinning processing, and Cheno coding, and preprocessing that converts image data into encoded data suitable for CPU processing and pattern recognition processing. I do.

The above-mentioned image input processing and pre-processing are processed in the working area of the memory 5, and the encoded image is stored.

[Chen fist code]

An example of chain encoding is shown in FIG.

21 is an 8-way code, which expresses the relationship between adjacent pixels by 1
It is expressed using chords from 8 to 8. The image 20, which has been thinned so that the line width is one pixel, is subjected to Cheno coding from pixel S to pixel E using an eight-way code 21, resulting in a Cheno code 22.

Since the image 20 can obviously be reproduced from the cheno code 22, the cheno code 22 stores image information.

The memory 7 stores pattern names corresponding to the data of the previously created sequential logic.

By the way, Chen number code 2? , the adjacent pixels from the start point S to the end point E of the image 20 in FIG. For example, from pixel S to the next pixel, the direction is 3, and then the direction is 2.
It is oriented as follows.

In this embodiment, the creation of a dictionary regarding standard logic and the creation of a dictionary regarding discriminant functions are performed sequentially, and the image input processing in step 1 and the preprocessing in step 2 are executed at the stage of creating both dictionaries.

In the process determination process of step 3, a flag (discrimination function creation if flag (F)=1, which will be described later) is checked to determine whether it is a sequential logic creation stage or a discriminant function creation stage. If it is the step of creating a sequential logic (F=0), proceed to Y. In the step of creating a sequential logic, the next step is to modify the sequential logic.

Modification of the sequential logic in step 4 involves storing the standard sequential logic in step 5, which was input from the console corresponding to the category, into the sequential logic memory 6 in step 6, using actual data, and following the learning rules. , we add and modify the order of practical sequential logic, that is, modify the modification.

[Modification of sequential logic]

The sequential logic modification process will be described in detail below.

When learning and creating sequential logic.

1) Preservation of necessary information 2) It is necessary to establish rules for the learning method, and for this purpose, in the present invention, standard sequential logic, which is a sample sequential logic that ignores noise such as digitization noise, is stored on a console, floppy disk, etc. By inputting from , the necessary information is stored, and in particular, the direction that represents the characteristics of the stroke of the pattern is defined from the transition between states in standard sequential logic. It has characteristics.

A transition diagram of the standard sequential logic of pattern "7" is shown in FIG.

In FIG. 4, AI, A2. -----A
5 is a state, and the code attached to the arrow provides a transition condition between states.

For example, in state A1, when the code is 3, it is possible to transition to state A1, when the code is 2, it is possible to transition to state A2, and in other cases, transition is not possible. (That is, it does not match.) Here, when starting from state S and arriving at state E according to the allowed transitions, pattern "7" and the sequential logic match.

An example of a Cheno code obtained by input processing and preprocessing an image of the actual pattern ``7'' is 7E.Since the Cheno Φ encoding has been described above, it will be omitted here.

This code can transition from state S to state A5 in the standard state transition diagram shown in FIG. 4 (input from the aforementioned console 3 or external storage such as floppy), but code
Since transition to is not allowed, it will not be matched. The purpose of modifying the sequential logic in step 4 is to modify the state transition diagram to match this.

Definitions of directions are illustrated in FIGS. 5-1 and 5-2.

The forward direction of state B in FIG. 5-1 is determined by the transition conditions for states A and C, that is, the arrangement of direction codes 8 and 7 in FIG. The rotation direction when taking a shortcut from to 7 is the forward direction, and the reverse direction is the reverse direction. Similarly, regarding the transition state 771p, state C
Due to the transition conditions for and D, that is, the arrangement of codes 1 and 2, the arrow in FIG. 5-2 is in the forward direction.

[Correction and repair of sequential logic]

Figure 6 shows a flowchart of the sequential logic modification process.A detailed explanation of the sequential logic modification process will be given based on Figure 6.Step 30 is the process of converting the Cheno chord, and the subsequent process will be simplified. In order to
Converts a code into a code string that only represents changes. The original cheno chord will also be used later for pattern recognition.

For example, a cheno chord "5E" becomes 532187676767E through the conversion process.

[Verification process]

The subsequent processing is performed based on the converted code string.

The stored sequential logic is moved to a working area in memory 6 for modification by the storage process of step 38. Then, in the matching process in step 31, a code string representing only the changing points of the converted code string is compared with the sequential logic.

Let's take the code string S32187676767E and the state transition diagram of FIG. 4, that is, the sequential logic, as an example.

The correction process after the verification process will be explained.

First, code 3 is next to the code string S, and in FIG. 4, the code that is allowed to transition from state S is 3, and the transition is allowed for code 3, that is, it cannot be matched in the matching judgment at step 32. Ta. Therefore, a transition is made from state 8S to state A1. At the end of the collation process in step 33, it is determined whether the collation process is completed based on whether or not the code is at the last code E of the code string. is transferred to the location of the sequential logic stored in step 6, and the sequential logic is rewritten with the corrected one (step 6-A in FIG. 3).

If the process has not been completed, the process returns to step 31.

Then, for example, following the flow of the above-mentioned story, a process of matching code 3 with the next code, that is, code 2, is performed.

In FIG. 4, a transition is made from state S to state A1 due to code 3, but in state A1, transition is allowed for code 3 and code 2. Therefore, the following code 2 can transition from state A1 to state A2, that is, it verifies,
The state transitions to state A2.

Similar processing is repeated until state A3. A transition is made from A4 to state A5 by code 7 which appears first in the code string.

The next code after code 7 is code 6, but in state A5 transition is only allowed for codes 7 and E, and verification is not possible. Rules 1 to 5 for correction at this time
(described later), and the correction was possible.

This is fed back to the matching process in step 31. If correction is not possible, the correction process ends.

The sequential logic at the location stored in step 38 during the current modification is ignored. That is, step 6 of the above-mentioned FIG. 2 of the sequential logic stored in the process of step 6
Rewriting as explained in A is not performed.

Here, if the standard state transition diagram in FIG. 4 is modified to the state transition diagram in FIG. 7 according to rules 1 to 5, it becomes possible to match the codes after code 6 in the code string, and step 33 It is determined that the collation process has been read and finished, and the correction process for the data is completed. Then, in step 38, the modified sequential logic that has been modified is stored in the working area of the memory 6.

[Correction Processing, Rules 0 to 4] Hereinafter, Rules 1 to 5 for performing sequential logic modification processing will be explained.

Before explaining each rule, let us define the language used here.

The auxiliary state refers to a state added by modification to the standard state transition diagram, and is distinguished from the state in the standard state transition diagram (hereinafter referred to as standard state).

The state iAs in FIG. 7 is the auxiliary state.

A jump transition is a transition to a state other than the adjacent state, and the number of jumps is 0 calculated based on the standard state, for example, the fourth
In the figure, the transition from state A1 to state A3 is skipped by 1.
This is called the transition.

Suppose there is an auxiliary state adjacent to state A4,
The transition from the auxiliary state to state A5 is called a transition with jump number 1, and the transition to the adjacent state has jump number O.

What is common to Rules 1 to 5 below is [Rule 2
] Transitions to rules are not allowed except for transitions to auxiliary states.

In other words, although the state transitions in the directions of the respective arrows shown in FIG. 5-2, the state does not change, for example, from 8 to 1 or from 1 to 8.

[Rule 1] One auxiliary state is added from the state S (this is called the starting state), and then the auxiliary state is added if a transition to the standard state is possible with the number of jumps, for example, 2 or less.

The transition conditions allowed at this time are transition from the initial state to the auxiliary state,
Nine examples of transitions from the auxiliary state to the auxiliary state itself and from the auxiliary state to the standard state are shown in FIG. 8, and as is clear from the figure, the shape mao in the figure is the auxiliary state.

[Rule 2] When Rule 1 is not satisfied, if a transition from the initial state to the standard state can be made with a transition of 2 or less jumps, this transition is added.

An example is shown in FIG. 9, in which a and b are attached, indicating the added transition.

[Rule 3] When the state is between state S and state E, and the code is in the opposite direction, by adding an auxiliary state, it is possible to transition to the standard state with the number of jumps of 1 or less in the next code, When a transition can be made from the standard state to the standard state with a jump count of, for example, 1 or less, an example of adding this transition is shown in FIG. 10. The one with a attached in FIG.

[Rule 4] When Rule 3 is not satisfied, 0 examples of adding transitions from the auxiliary state and standard state to the auxiliary state, transitions from the auxiliary state to the auxiliary state itself, and transitions from the auxiliary state to the standard state are shown in the 11th example. State aQ in the 0 diagram shown in the figure is an auxiliary state.

In each determination process of rules 1 to 4 in steps 34 to 37 in FIG. 6, it is determined whether or not each of the previous rules is satisfied.

If it is satisfied, the order logic is modified according to the rules and fed back to the matching process. If the rules are not satisfied, judgment processing is performed in the order of rules 1 to 4.

Returning to the overall flowchart of FIG. 2, the steps after the sequential logic correction process in step 4 will be explained.

In step 7, it is determined whether or not the modification of the sequential logic using all the data in the l category has been completed.
Input various next pattern images belonging to the ``category'' and modify the sequential logic.

l When all the data in the category has been corrected,
The standard sequential logic stored in the memory 6 by the standard sequential logic processing in step 5 and the sequential logic storage processing in step 6 described above is modified in step 4, but the data is stored in the memory 6 in step 6-A. , is registered in the memory 7 as a dictionary (I) in step 10.

Note that the operation of step 6-A may be performed in step 4. In the determination process of step 8, it is determined whether or not the processing for all categories has been completed. If not, the process returns to step 1. At this time, step 5 mentioned above
The standard sequential logic of the next new category to be modified can be stored in the memory 5 through the standard sequential logic processing in step 6 and the sequential logic storage processing in step 6.

[Discrimination function creation process]

When it is determined that the modification of the sequential logic using all data of all categories is completed, the result is "Y" in step 8,
Proceeding to step 9, a flag is set to inform the process judgment process in step 3 that a discriminant function will be created in the following process. As a result, the subsequent processing through step 3 becomes the discriminant function creation processing from step 11 onwards from "No". Then, the image input processing of Step E and the preprocessing of Step 2 can be used for both modifying the sequential logic and creating the discriminant function.

Next, the discriminant function will be explained.

By using sequential logic in the present invention, pattern identification accuracy has become extremely high compared to conventional methods. Use functions.

This makes the features of the present invention even clearer.

1. By removing the limit on the number of times each state can be passed,
The number of required sequential logics has been greatly reduced.

2. Instead of restricting individual states, we find a function that can express the stroke structure of the relationship between states, and use it as a function for identifying patterns, that is, a discriminant function. It becomes a wide recognition device.

Then, by defining a discriminant function whose variable is the number of times it passes through each state in the sequential logic, and finding a method for extracting coefficients, threshold values, and features, the automatic creation of the discriminant function was made convoluted.

Next, the principle of automatically creating a discriminant function will be explained.

Consider a case where, in arbitrary sequential logic, each state in the sequential logic represents something that identifies a pattern belonging to a different category compared with the sequential logic (i.e., a discriminant function).

In the simplest case, the number of times X passes through the first state
When n is more than 10 times in the case of pattern Pa, and less than 10 in the case of pattern Pb, it is possible to know whether the input image is pattern Pa or pattern Pb from the value of Xn. Expressing this mathematically, X
When n > I O, that is, Xn-10>0(7), the input image is pattern Pa, and Xn<10, that is, Xn-1o
<o, the input image becomes pattern Pb, and at this time F
(Xn)=Xn-10 is called the discrimination interval number, and the reference value 10 for separating pattern Pa and pattern Pb is called a threshold value.

Generally, it is necessary to use more variables for identification, and when expressed as a linear identification function, equation (1) is obtained.

Here, Xn is the number of times An and Bmn pass through the nth state. G is a function consisting of terms of order higher than the third order of Xn (this can be ignored when identifying a simple pattern). S belongs to two categories. Threshold value for identifying patterns (X11 is a set of xl, x2, --m=) For patterns Pa and Pb that yield to two categories, the number of times Pan and Pbn pass through the n-th state, respectively. For any pattern belonging to , F ((Panl , S) > O(2) F (
(Pbn) , S) < O(3) Coefficients An , B
Once mn and the threshold value S are determined, it is possible to identify whether it is the human power 17Iiii image, pattern Pa, or pattern Pb using the relationship between equations (2) and (3).

[Coefficients An, Bmn] Next, a case will be described in which the coefficients An and Bmn are calculated.

From equations (1)2 to (3), it becomes. (In the above, we will look at the G term in order to identify it with a simpler tH function and to make the explanation easier.) 1': This is Pan-Pbn and PamPan-Pbm.
Pbn represents the difference in the bead pattern Pb, that is, the distinguishing feature and feature.

In general, if you use something with a clear difference, that is, something with a large difference, it will be easier to distinguish and identify the difference.In order to improve the recognition rate, it is necessary to use Pan-Pbn and PamPa.
When n-PbmPbn is large, the coefficients Am and B
mn needs to be large.

From PnoF, An = C(Pan-Pbn) K
(5) Bmn=C(PamPan−PbmPbn)
K (6) Here, K is an appropriate odd number and C is a positive constant.

Equation (5) is the first and second term on the left side of Equation (4).

Substituting (6) gives (Pan-Pbn) K
”I.

(PamPan-PbmPbn)+'to4'''
This is to make the exponent an even number so that even if 1 is negative, the equation will be positive. The pattern is the seed/
Since we take A16 of /, it is necessary to use equations (5) and (6) as average values.

An=Can-bn
(7) Bmn=CPam an-bmPbn
(8) becomes. Here represents the average value.

Next, the threshold value S is determined.

In equations (1)2 to (3), F ((Panl , S) >0>F ((Pbnl
, S), ΣAnPan+ΣΣBmnPamPan>S>n
mn ΣAnPbn+ΣΣBmnPbmPbn (9
). The threshold value S that satisfies equation (9) is approximately.

■ 5=-(ΣAn(Pan+Pbn)+n ΣΣBmn(PamPan+PbmPbn)) (
9-1)mn. (Panl in formula (9-1).

(P b n) takes various values depending on the pattern, so
Need to average, 5=-(ΣAn an+ nΣ+n ΣΣBmn PamPan+PbmPbn)
(10)mn.

In equations (7), (8), and (10), C and K can be determined appropriately (for example, C=K=1)
, from pattern Pa and pattern Pb (Panl
and (P b n), the coefficient An,
Bmn and threshold S are determined.

In other words, the discriminant function is found.

Although the method for determining the discriminant function for patterns belonging to two categories has been described above, discriminant functions for patterns belonging to two or more categories can be similarly determined.

For example, if patterns belonging to three categories are Pa and P
b and Pc.

First, the discriminant function F for pattern Pa and pattern Pb is
l is found, Fl((Pan), Sl) > O(lt) Fl(
(Pbn), St)<O(12). Similarly, the discrimination amount @F2 is found for pattern Pa and pattern Pc, and F2 ((Pan), S2) >O(13)F2
((Pen), S2) <O(14). Similarly, the discriminant function F3 is found for pattern pb and pattern Pc, and F3 ((Pbn), S3)>O(15)F3 ((
Pcnl , S3)<O(16).

Using equations (l l) and (16), Fl((Xnl , St)>o or F 2 ((Xn
l, S2)>0 (17), the input image is identified as the putter yPa.

Similarly, F+ ((Xn), Sl)<O and F3 (fXn)
, S3) >0 (18), the input image is identified as pattern pb.

F2 ((Xn), S2) < o or F3 ((xnl
, S3) When <O(19), the input image is identified as pattern Pc.

FIG. 12 abstractly represents the above relationship. Each pattern P
The category spaces of a, Pb, and Pc are divided by discriminant functions Fl, F2. It is F3, (17) 2~
Each category space is defined by equation (19).

For example, if the input image is within the category space of pattern Pa, the input image is identified as pattern Pa.

Even when the number of categories of patterns matched by sequential logic is larger, the discriminant function can be found in the same way.

From the above, from equations (7) and (8) for calculating the coefficient An and coefficient Bmn, and equation (10) for calculating the threshold value S,
Equation (1) giving the discriminant function is found. It is possible to identify multiple categories that match any one order logic, but by using this discriminant function and the discriminant conditions of equations (2) and (3), it is possible to identify categories that are effective for discrimination, that is, to identify The features improve recognition rate and processing speed.

[Feature extraction method]

Next, a feature extraction method will be explained.

From formulas (1) to (2), formulas (7), formulas (8), and formulas (10), F ((Pan), 5) (2o) F (1Pbn), 5) is obtained.

In equations (20) and (21), what gives stable characteristics is one that is always positive for all patterns Pa and always negative for all patterns Pb.

For example, for the number of times the nth state is aIIiS to be a stable feature.

must hold true for all patterns of Pan and Pbn.

This condition may also be true for the sum of multiple terms in equations (20) and (21).

For example, in An(Pbn--'s completed an + completed qT, "<0 (25) is all patterns of Pam, Pan, Pbm and
If it holds true for bn, the linear sum AmXm+AnXn of the number of times of passing through the m-th and n-th states becomes a stable feature.

Generally, in equations (20) and (21), 1
The linear sum of a term or multiple terms has a high probability (more than 90% in practice), etc., is positive for the pattern Pb, and is positive for the pattern Pb. If it is negative, it can be used as a feature for identification.

In the method created using the uninvented modified iTE method of sequential logic, if there are states that transition with the same code and the number of jumps between states is 1 or less, these states (hereinafter referred to as common states) are The probability that the linear sum becomes a stable feature is higher than that of the number of passes alone. For example, in the state transition diagram of pattern "7" in FIG. 13, state A1 and state a1
The number of jumps between states is l, and the state transitions by code 3. In the discrimination between pattern "7" and pattern "wa", the discriminant function is −7,6(X1+xl). can do.

Here, Xl and Xl are distinguished by the number of times they pass through state A1 and state a1, respectively.

This is because the state 1QtA1 and the state a1 share information on the line segment in the stroke.

The next most likely feature is the term (X1+X1)'' (X1+xt corresponds to the distance between state 8A1 and state a1) among the quadratic terms of Xn in equation (1). ) term (corresponds to the product of the number of passes through adjacent states - the curvature of a line segment) where X2 is the number of times xl passes through state A1 and adjacent state A2, and X2 is the number of times xl passes through state A2. These tend to be features because the former is the length of the line segment, and the latter is the sum of the products of adjacent lines, which is related to the curvature of the line segment.

Next, the features that make a big difference between the patterns, that is, the features that make them easy to identify, are the features that make them easy to identify. This is a term with a large absolute value in (An) and (Bmn), or a term with a large absolute value of the average value of the sum of a plurality of coefficients.

In practice, stable and easily distinguishable features are used as the discriminant function.

[Feature extraction procedure]

A procedure for extracting such features and determining a discriminant function will be explained.

1. Find the number of times each state in the sequential logic is passed through the process of matching with the sequential logic.

2. Find the sum of the number of times the common state is passed.

3. Find the square of the values found in 1 and 2.

4. Find the product of things passing through adjacent states in each state in the sequential logic and in a common state.

5. When the matching process for each category is completed, calculate the average value of the values obtained in step 4 from l.

6. In equations (7) and (8), K=l, C
= 1, and the coefficient (A n
), (B m n ) are determined.

7.6 Number the obtained coefficients (An) and (Bmnl) in ascending order of absolute value, for example, up to 10.

8. In the order of the numbers assigned in step 7, find those that satisfy equations (23) 2 to (25) with a probability of 90% or more, and find, for example, up to about 5 in descending order of probability. This is the feature used in the discriminant function.

From the features obtained in 9.8, use equation (10) to find the threshold value. Here, the coefficient (An
l, (O for Bmnl, and find the threshold value.

10. - Perform steps 6 to 9 above between different categories checked against the same sequential logic to obtain a discriminant function.

The process for determining the discriminant function of J-1 will be further explained based on the flowchart of FIG.

In the flowchart of FIG. 2, the processing after the dictionary search/verification process in step 11 in which the creation of a dictionary related to sequential logic is finished and the creation of a discriminant function is started will be described.

In the dictionary search/verification process in step 11, step 10
Search for sequential logic from the dictionary related to sequential logic prepared separately, and perform matching processing.

At the stage of the matching process, the number of times each state is passed is determined.

Step 12 determines the matching result, and if not matching, performs search matching for the next sequential logic. When the comparison is made, a process is performed in which the sequential logic and result of the comparison are stored in step 13. A detailed flowchart of this process is shown in FIG.

Step 50 is to perform step 2 of feature extraction to obtain a feature extraction and a discriminant function. For example, in the state transition diagram of pattern "7" in FIG. 13, state A1 and state a1. Condition A
2 and state a2 and state A4 and state a4, the sums passing through each state are calculated.

Step 51 is step 3 of extracting features and determining the discriminant function.
This process performs the following processing.

Step 52 is step 4 of extracting features and determining a discriminant function.
Process. For example, in the state transition diagram of pattern "7" in FIG. 13, the sum of state A1 and state a1, state bO1, state A1
and the sum of state a1, the sum of state A2 and state a2, the sum of state A4 and state a4, and the state A3. The product of the sum of state A4 and state a4 and the number of times each state of state A5, state A5, and state a6 is passed is determined.

In step 53, the number of times each state is passed and the results of steps 50 to 52 are sorted by sequential logic and category and stored.

The above process is performed by the judgment process in step 14 of FIG.
This will continue until all data have been collected.

After data collection is completed, a discriminant function is created in step 15. A detailed flowchart of this process is shown in FIG.

The process of creating a discriminant function in step 15 is performed for each sequential logic between categories matched with the sequential logic. For example, if the number of matched categories is one, the process of creating a discriminant function is not performed. [165 If there are two categories that match, create a discriminant function to identify these categories. qr (combined category is rA),
rB'', rc'', rA, 1 and
"B", "A" and "C"
and r]3'' and rcj, respectively. The same procedure is applied to cases 4 to 5.

The flowchart for creating the discriminant function shown in FIG. 15 will be described below.

In the average value processing in step 60, the values stored in step 53 are averaged for each category of each sequential logic. Step 61 is a process of calculating coefficients, and from each average value obtained in the average value process of step 60, formula (5) or (
6) Determined from the formula.

In the coefficient ranking process of step 62, the coefficients found between different categories of each sequential logic are numbered in descending order of absolute value.

In the feature extraction process in step 63, the coefficients in step 62 are ranked between different categories of each sequential logic according to the numbers assigned in the process, so that each coefficient is between (22)2 and (25).
) that satisfies the conditional expression with a probability of 90% or more, extract at most five items with a high probability and a low rank in step 62 (that is, the absolute value of the coefficient is large), and use this as a feature. shall be.

In the threshold processing in step 64, threshold values are determined according to equation (10) from the features extracted in the feature extraction processing in step 63 between different categories of each sequential logic.

Through the above processing, the constants in the discriminant function are adjusted, that is, the discriminant function can be created.

The discriminant function created here exists between different categories compared with each sequential logic, so step 1 in Figure 2
In registering the discriminant functions in the dictionary in step 6, the sequential logic stored in the memory 7 is made to correspond to the discriminant function that identifies the category matched with each sequential logic. It also stores the results identified by the discriminant function, that is, the recognition results.

Although the dictionary necessary for pattern recognition can be created through the above processing, each of the memories 4 to 8 in FIG. 1 and the phase relationship will be explained in detail below.

The memory 4 stores programs for performing preprocessing, sequential logic creation processing, discriminant function creation processing, memory management, etc., and these programs are executed by the CPUI.

Memory 5 and memory 6 are work areas for preprocessing and dictionary creation processing, and the memory contents are constantly changed.

The memory 7 is divided in advance, and each of the divided sections stores each sequential logic. This simplifies the sequential logic modification process and search process.

[Memory structure of sequential logic]

FIG. 16 shows the notation tα structure of sequential logic.

In this example, in order to make it easier to modify the sequential logic, we will introduce the code for transitioning to a state, the (relative) address at which transition from one state to the next state is allowed, and the number of times to pass through each state. A fixed length memory capacity 9 (for example, 1 byte) is allocated. In addition, by storing the starting state, final state, and sample state in the sample transition diagram at predetermined addresses, and providing a place to add one auxiliary state between the sample states,
Made it easier to modify the sequential logic.

After the final state, code E, a predetermined address for storing the discriminant function is stored, and the sequential logic and the discriminant function are associated with each other.

FIG. 16 is a memory map representing the standard state of pattern "7" in FIG. 4 for S seconds, and FIG. 17 represents a modified state 8a transition of FIG.

The state interval in Figure 7 is modified by adding auxiliary letter 8A6 to state A5 during the standard state transition in Figure 4, and at this time, the transition from state A5 to state A6 with code 6 is made. To the newly added state A6, a transition to state 6 due to code 6 and a transition to state A5 due to code 7 are added.

On the memory map, as you can see from the difference between Figure 7316 and Figure 17, 6 is placed at the code (state).
(A6) has transition destination 6 (As) at transition condition 6 of 7(As) and transition destination 6 (A6) and 7(A5) at transition condition 6 and 7 of 6(As). The address will be memorized.

The memory 8 is divided in advance, and each division is written with a discrimination function corresponding to the sequential logic, a conditional expression using the discrimination function such as equations (20) and (21), and a discrimination result. .

[Recognition processing]

A case in which recognition processing is performed using a dictionary of sequential logic and discriminant functions created using this method will be explained.

At this time, the configuration of the device remains unchanged except that a program for recognition processing is used instead of a program for creating a dictionary. FIG. 18 shows a flowchart of the recognition process.

The operation will be explained below.

Image input processing in step l, i'11 processing in step 2, collation processing during search of the sequential logic dictionary in step 11,
Then, perform the same process as in Step 12, which is the judgment of collation and the creation of the dictionary in step 1-).

In step 70, it is determined whether or not the search for word 3 of the sequential logic is completed 15. If the input image does not match all of the sequential logics, that is, if it cannot be recognized, the process is terminated and recognition is rejected. issue.

In the identification process using the identification function in step 71, each state y, ! in the sequential logic obtained at the stage of the matching process. The value of the discriminant function is determined from the number of times one ligature passes, the discriminant result is obtained from the above-mentioned conditional expression, and the recognition process is terminated.

[Eri]

As described in detail above, according to the present invention, it is possible to provide a learning function in the development of 0CRO or in the creation of words -) during pattern recognition.

Furthermore, it is possible to provide a pattern recognition device which incorporates the idea of sequential logic (automaton) into the stroke structure analysis method and further performs classification processing using a classification function.

Additionally, it has become possible to automatically create a dictionary for online or offline character recognition.

Furthermore, it has become possible to create a chain code for dictionary creation based on a chain code for pattern recognition. By modifying the sequential logic that is a dictionary by starting with the chain code, it became possible to learn better sequential logic. Further, according to the present invention, it has become possible to create a discriminant function whose variable is the number of times each state of sequential logic, which is dictionary data created by code data indicating five directions, is passed through. Furthermore, the present invention has made it possible to provide claims.

[Brief explanation of drawings]

FIG. 1 is an explanatory block diagram of the present invention. FIG. 2 is a flowchart showing the operation of the present invention. FIG. Figures 5-1 and 5-2 are diagrams showing the definition of directions, Figure 6 is a flowchart of correction processing for sequential logic, Figure 7 is a diagram showing state transitions after modification, and Figure 8 is an auxiliary state. FIG. 9 is a diagram showing the state transition for explaining rule 1 which has an auxiliary state. FIG. 10 is a diagram showing the state transition for explaining rule 2 which has an auxiliary state. Figure 11 is a diagram showing a state transition for explanation of Rule 4 with auxiliary states, Figure 12 is a diagram showing identification of category space, and Figure 13 is a diagram showing pattern "7" Figure 14 is a flowchart related to data collection. Fig. 15 is a flowchart for creating a discriminant function, Figs. 16 and 17 are explanatory diagrams of memory maps of sequential logic, Fig. 18 is a flowchart for recognition processing, and Fig. 19 is a diagram showing a sample broken line approximation pattern. . 1 is a CPU 7 is a memory for storing sequential logic 8 is a memory for storing discriminant functions Figure 5-7 7 Kayama” '1' Yes, it's a fumimo?・Shifu'ryuha name 75 layer identification leap number group lifetime income A

Claims (1)

[Scope of Claims] Input means for inputting a sample pattern; sample pattern processing means for converting the sample pattern input by the input means into standard sequential logic; A pattern recognition device characterized by also performing conversion.