JP2973898B2 - Character recognition method and device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a character recognition device for automatically reading input characters.

[0002]

2. Description of the Related Art In the case of automatically recognizing characters with high precision, candidates for recognition results output from individual character recognition results are:
By performing post-processing such as word matching, even if all characters cannot be read completely, they may be able to be read in word units, and in situations where knowledge such as names and addresses can be used, aggressively improve reading accuracy It is used for

Hereinafter, two specific examples of address reading will be described. First, in Japanese Patent Application Laid-Open No. 2-109187, "Post-processing method for character recognition of solid address", the number of matches between the character recognition result and the address dictionary is calculated for matching the individual character recognition result with the address dictionary, A large number of address names are extracted from the address dictionary and are used as recognition results.

Japanese Patent Laid-Open Publication No. Sho 62-103785 discloses a character reading device, in which a candidate output from an individual character recognition result is set as a collation range, and collation with an address dictionary is performed within the collation range. In this case, the matching result is regarded as the recognition result.

[0005]

However, the above-mentioned 1
As shown in FIG. 6, for example, as shown in FIG.
If "Higashimachi" and "Honmachi" are in the same area by chance and cannot be read by individual character recognition, the number of matches is 1 for both "Higashimachi" and "Honmachi" for the individual character recognition result "? It becomes the same, and the address cannot be specified. This is because the number of matches becomes don't care for an unreadable portion of the individual character.

[0006] On the other hand, in the above-mentioned second, the collation range is raised as a candidate for outputting the individual character recognition result, if there is a candidate to be collated.
Compared to the first example, the matching accuracy can be expected to be improved by searching for a part that cannot be read from candidates lower in the individual character recognition result. For example, as shown in FIG. 7, even if "Higashimachi" is the correct answer and "East" does not appear in the first place of the individual character recognition result as shown in FIG. The possibility of "Higashimachi" is quite high, and the matching can be performed with higher accuracy than the first example. However, even in this case, if both “East” and “Book” are present in the lower candidate, even if the upper candidate is “East”, the reliability of the rank of the lower candidate is generally unstable. If it is not certain, and "book" appears at the top, or if there is no correct character in the candidate, a correct collation result cannot be obtained as in the first case. This is particularly noticeable in the case where there are a plurality of candidate characters that can be embedded and these are similar to each other. Even if the characters are crushed, the feature extraction is not performed well and the character becomes unstable.

[0007] This is because if a character pattern that cannot be read by the individual character recognition unit can be complemented as a result of collation, it is not confirmed by any means whether or not the character can be complemented.

It is an object of the present invention to provide a character recognition device which solves the above-mentioned problems with the collation method and can perform post-processing with higher accuracy.

[0009]

A character recognition device according to the present invention.
Is a photoelectric conversion means for converting an input character pattern into an electric signal
Image data storage means for storing the signal obtained by
And the image data stored in the image data storage unit.
Binarize data, absorb fluctuations within the same character type, and
Binarization / normalization means that emphasizes differences between character types
And an input sentence processed by the binarization / normalization means
A feature for extracting feature values effective for character discrimination from character patterns
Character extraction means and characters extracted by the feature extraction means
Feature data storage means for storing feature values effective for discrimination
And the feature amount stored in the feature data storage means.
To determine which character type the input character pattern corresponds to
Multiple candidates in descending order of possibility as the judgment result.
Character recognition means for outputting, and said image data storage means
Extract the graphic information of the input character pattern stored in the
Graphic information extracting means, and the character information obtained by the character recognizing means.
The plurality obtained as the determination result for the input character pattern
For the candidate, using the input character pattern knowledge, the input character
Matching is performed based on the judgment results for the patterns before and after the pattern.
And recognizes one character from the plurality of candidates for the input character pattern.
Post-processing means for determining the result
If the recognition result could not be determined in
1. ~ 3. Is performed. 1. The graphic information extracted by the graphic information extracting means is changed.
Until the binarization / normalization means
2 Ru by changing the threshold. Binarization at the time when the change of the graphic information is detected;
For the normalized character pattern, the character recognition means
2. Recognize more . 1. ~ 2. In the post-processing means
Repeat until the recognition result is determined

[0010]

[0011]

The operation of the present invention will be described with reference to FIG.

The graph of FIG. 4 is a density histogram of the character pattern "east" which has not been verified in the example cited in FIG. 6 or 7 at the time of word verification, the horizontal axis represents the density level, and the vertical axis represents the frequency. .

A threshold value for binarization in the first individual character recognition is s1, and a binary pattern obtained at that time is shown in FIG.
Shown in Since the crushing has occurred, “East” or “Book” cannot be accurately determined at this time by the individual character recognition means. Therefore, the threshold value of the binarization is set such that the line width becomes smaller in FIG.
Perform the change to 2. FIG. 5B shows the binary pattern obtained at that time. In this pattern, since the strokes are clear and the characters are not crushed, a stable feature is extracted. As a result, the individual character recognition means can correctly determine "east".

As described above, according to the present invention, for a character that has not been collated during word collation, a pattern in which the threshold for binarization is changed is generated again, individual character recognition is executed, and the discrimination obtained therefrom is performed. Since the collation is performed by referring to a new candidate of the result, the conventional technology enables the collation of the correct character even when the erroneous collation is performed.

[0015]

Next, a first embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram of a character recognition apparatus showing an embodiment of the present invention, and FIG. 2 is a flowchart showing the operation thereof.

The character recognition device according to the present invention comprises an image data storage means 301 for storing a signal obtained by a photoelectric conversion means for converting an input character pattern into an electric signal, and an image data stored in the image data storage means 301. The binarization / normalization unit 101 performs a process of binarizing and absorbing variations within the same character type and emphasizing a difference between different character types. A feature extraction unit for extracting a feature amount effective for character discrimination from the input pattern;
Feature data storage means 302 for storing a feature amount effective for discriminating the character extracted by the above, and the feature data storage means 3
Character recognition means 103, which determines which character type the input pattern corresponds to based on the feature amount stored in the input pattern 02, and outputs a plurality of candidates as a determination result in a descending order of possibility.
Graphic information extracting means 10 for extracting graphic information of the input pattern stored in the image data storing means 301
5 and the plurality of candidates obtained as the determination result for the input pattern obtained by the character recognizing means 103 are compared from the determination results for patterns before and after the input pattern using knowledge of the input pattern, Post-processing means 1 for determining one character from the plurality of candidates of the input pattern as a recognition result
And a data bus 401 for exchanging data among the binarization / normalization means 101, the image data storage means 301, the feature extraction means 102, the feature data storage means 302, the character recognition means 103, and the graphic information extraction means 105.
The binarization / normalization means 101, the image data storage means 301, the characteristic means means 102, the characteristic data storage means 302, the character recognition means 103, and the graphic information extraction means 10
5 is constituted by control means 201 for controlling the control unit 5.

First, the photoelectrically converted character pattern is input and stored in the image data storage means 301 (step 21).

Next, the binarization / normalization means 101 performs binarization and further normalization (step 2).
2). Binarization may use the same threshold value for the entire image of one character. As a threshold value determination method at this time, for example,
"Nobuyuki Otsu: Mathematical Study on Feature Extraction in Pattern Recognition, AIST Research Report No. 818 (1991.
7), and can be realized using a discriminant analysis method. In addition, for normalization, in order to facilitate identification, fluctuations within the same character type may be suppressed and differences between different character types may be clarified. For example, “J. Tsukumo,
et al. : Classification of
handprinted Chinese Chara
cters Using Non-linear No
rmalization and Correlati
on Methods, Proc. of9th ICP
R (1990.7) ".

Next, feature extraction is performed on the character pattern that has been binarized and normalized (step 23). It is sufficient that a feature amount effective for character discrimination can be extracted. For example, "Hamanaka et al .: Consistency between nonlinear normalization and feature extraction in handwritten kanji recognition," IEICE Technical Report, PRU91-85 (199)
1.11)). The feature for the extracted character pattern is stored in the feature data storage unit 302.

Next, the character recognizing means 103 determines the character using the characteristic amount stored in the characteristic data storing means 302 (step 24). From the extracted feature data, it is determined which character it is. Assuming that the feature amounts exemplified in step 23 are used, for example, “Jun Tsugumo: Improvement of Direction Pattern Matching Method and Application to Handwritten Kanji Recognition,” IEICE Technical Report, PUR90-20 (199)
0.7) "is appropriate. A plurality of recognition results, for example, 10 candidates are output in the order of higher reliability.

Next, in the post-processing means 104, the recognition result is compared with the knowledge (step 25). Here 1
The recognition result for each word is verified using knowledge, and the individual recognition result for each character is corrected. At this time, a character that does not match the knowledge is detected, and finally, a collation result is output (step 26).

An example of the processing procedure in step 25 will be described with reference to FIG. FIG. 3 is a flowchart showing an example of realizing the collation between the individual recognition result for each character and the knowledge.

First, the candidate output from the individual recognition result is collated with the knowledge (step 31).

Next, it is determined whether all of the collation characters are present in the candidates output as the individual character recognition results (step 32).
That is, when collating based on knowledge, it is checked whether all the characters are among the candidates of the determination result of the individual character recognition. Here, if all the characters are among the candidates of the determination result of the individual character recognition, the process is terminated. Conversely, as a result of matching,
That is, if there is no character to be corrected by knowledge among the candidates of the recognition result, the graphic information is extracted from the character pattern of the corresponding portion (step 33). As this graphic information, for example, the line width of a character is extracted.

Next, the threshold for binarizing the character pattern of the corresponding portion is changed according to the character line width (step 3).
4). The line width is extracted by changing the line width in the direction of becoming thinner if the line width is large, and conversely, the line width is extracted in the direction of becoming thicker if the line width is small (step 35), and the line width is detected again (step 36). This is repeated until the line width changes (loop 1). The step of changing the threshold value may be optional. If a change in line width is detected, the pattern is subjected to individual character recognition again (step 37), candidates for the determination result are output (step 38), word matching is performed again (step 31), and matching is desired among the candidates. Repeat until a character appears (loop 2). If the character to be collated appears, the collation result is output. If the character to be collated until the end does not appear in the candidate of the decision result even if the threshold value is changed, the part is output together with the part which can be collated as unrecognizable. (Step 39).

As the graphic information, the number of connected line segments, the number of loops, or a combination thereof may be used regardless of the line width of the character.

Next, a second embodiment of the present invention will be described. In the first embodiment, if all the characters collated based on the knowledge in step 32 exist in the candidates for the determination result of the individual character recognition, the collation is terminated and the result is output. On the other hand, in the second embodiment, the above-described binarization and re-recognition of the pattern are repeated until all the collated characters reach the first place in the determination result of the individual character recognition. As the graphic information detected to change the binarization threshold when individual character recognition is repeated, as in the first embodiment,
Any of the line width of the character, the number of connected components, the number of loops, and a combination thereof may be used.

Next, a third embodiment of the present invention will be described. In the present embodiment, in step 32, when a plurality of characters that can be collated by knowledge appear in the candidates, the above-described binarization of the pattern is performed until one of the characters rises to the first place in the determination result of the individual character recognition. And re-recognition is repeated. As the graphic information to be detected in order to change the binarization threshold when individual character recognition is repeated, as in the first embodiment, any one of the character line width, the number of connected components, the number of loops, and a combination thereof is used. May be used.

[0030]

As described above, according to the present invention, when a character to be collated does not appear in a candidate for the judgment result of individual character recognition, graphic information of a pattern of a corresponding portion is extracted, and the pattern information is changed until it changes. By changing the threshold of the binarization and repeating the recognition until the result becomes a determination result candidate, it is possible to completely cope with the knowledge, and there is an effect that the accuracy of character recognition is remarkably improved.

[Brief description of the drawings]

FIG. 1 is a block diagram of a character recognition device according to an embodiment of the present invention.

FIG. 2 is a flowchart showing an example of the operation of the character recognition device shown in FIG.

FIG. 3 is a flowchart showing an example of an operation at the time of knowledge collation of the character recognition device shown in FIG. 1;

FIG. 4 is an explanatory diagram for explaining the principle of the present invention.

FIG. 5 is an explanatory diagram for explaining the principle of the present invention.

FIG. 6 is a diagram illustrating a determination method of a conventional character recognition device.

FIG. 7 is a diagram showing another determination method of the conventional character recognition device.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 101 Binarization / normalization means 102 Feature extraction means 103 Character recognition means 104 Post-processing means 105 Graphic information extraction means 201 Control means 301 Image data storage means 302 Feature data storage means 401 Data bus

Claims (4)

(57) [Claims] 1. An image data storage means for storing a signal obtained by a photoelectric conversion means for converting an input character pattern into an electric signal; Binarization that absorbs fluctuations in the text and emphasizes differences between different character types.
A normalization unit, a feature extraction unit that extracts a feature amount effective for character discrimination from the input character pattern processed by the binarization / normalization unit, and a character extraction unit that discriminates the character extracted by the feature extraction unit. A feature data storage unit that stores an effective feature amount; and a character type corresponding to the input character pattern is determined based on the feature amount stored in the feature data storage unit, and these can be determined as a determination result. Character recognition means for outputting a plurality of candidates in descending order of likelihood, graphic information extraction means for extracting graphic information of the input character pattern stored in the image data storage means, and the input character obtained by the character recognition means The plurality of candidates obtained as the determination result for the pattern are collated from the determination results for the patterns before and after the input character pattern using knowledge of the input character pattern, and the input is performed. And a post-processing means for determining a recognition result a character from the plurality of candidates of character patterns, it could not determine the recognition result in the post means situ
In this case, the following 1. ~ 3. Characterized by performing the processing of
Character recognition device. 1. The graphic information extracted by the graphic information extracting means is changed.
Until the binarization / normalization means
2 Ru by changing the threshold. Binarization at the time when the change of the graphic information is detected;
For the normalized character pattern, the character recognition means
2. Recognize more . 1. ~ 2. In the post-processing means
Repeat until the recognition result is determined 2. When the post-processing means adopts a judgment result output by the character recognition means other than the first candidate, the following 1. ~ 3. Characterized by performing
The character recognition device according to claim 1. 1. The graphic information extracted by the graphic information extracting means is changed.
Until the binarization / normalization means
2 Ru by changing the threshold. Binarization at the time when the change of the graphic information is detected;
For the normalized character pattern, the character recognition means
2. Recognize more . 1. ~ 2. In the post-processing means
Repeat until the recognition result is determined, and
The candidate is processed until the post-processing means determines the recognition result.
U 3. The method according to claim 1, wherein when the post-processing means adopts a candidate for the judgment result output by the character recognition means, if there are a plurality of collatable characters in the candidate, the following 1. ~ 3. Processing
The character recognition device according to claim 1, wherein the character recognition is performed.
1. The graphic information extracted by the graphic information extracting means is changed.
Until the binarization / normalization means
2 Ru by changing the threshold. Binarization at the time when the change of the graphic information is detected;
For the normalized character pattern, the character recognition means
2. Recognize more . 1. ~ 2. In the post-processing means
Repeat until the recognition result is determined, and
The candidate is processed until the post-processing means determines the recognition result.
U 4. A character recognition apparatus according to claim 1 , wherein said graphic information extracting means extracts the line width, the number of connected components or the number of loops of the image data as the graphic information.