JPH0210480A - Character deciding method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a character discrimination method in a character recognition device comprising a character reading sensor, a central processing unit, and a character recognition section.

[Conventional technology]

FIG. 3 is a block diagram showing the configuration of an ordinary character recognition device. In the same figure, ■ is an optical character reading sensor, 2 is a C
PU (central processing unit) 3 is a character recognition unit.

After inputting and storing the character image data read by the character reading sensor 1, the CPU 2 takes out the character image data and passes it to the character recognition unit 3, and the character code and shape feature data as the recognition result are used for the character recognition. Character discrimination is performed by receiving the information from the department.

In character discrimination using such a character recognition device, when the target character is a number zero (0), an alphabetic character O (0), a circle mark (○), etc., it is possible to identify the character based on the similarity in shape. was so confusing that it was mistakenly identified as the number zero (0) even though it was the alphabetic letter O (0), or vice versa.

As a way to improve this and increase the recognition rate, if the target character is a number zero (0), an alphabetic character O (0), a circle mark (○), etc., the character is not identified by itself. There is a known method that takes into account the characters located before and after the character.

That is, when the character code that the CPU 2 receives from the character recognition unit 3 is a number zero (0), an alphabetic character O (0), a circle mark (○), etc., the code is not scattered and is not determined. If the characters before and after the undetermined code are numbers, for example, the character before (to the left) is a number and the character after (to the right) is also a number, it is possible that the undetermined code is also a number. Since it has a strong gender, it is decided that it is the number zero (0). If one of the adjacent characters is a number and the other character is a kanji, hiragana, symbol, etc., there is a strong possibility that the undetermined code is a number, so it is determined to be the number zero (0).

Similarly, if the character before (adjacent to the left) of the undetermined code is an alphabetic character and the character after (adjacent to the right) is also an alphabetic character, there is a strong possibility that the undetermined code is also an alphabetic character. (0
). If one of the characters on both sides is an alphabetic character and the other character is a kanji, hiragana, symbol, etc., there is a strong possibility that the undetermined code is an alphabetic character, so it is determined to be the alphabetic character O (0).

Furthermore, if the characters on both sides are kanji or symbols, the undetermined code is determined to be a circle mark (0).

In this way, when the target character is a number zero (0), an alphabetic character O (0), a circle mark (○), etc., the recognition rate is higher than when the character is identified alone. It is possible to improve the

[Problem to be solved by the invention]

However, in the above-mentioned conventional technology, when one adjacent character of the undetermined code is a numeric character and the other adjacent character is an alphabetic character, there is a possibility that the undetermined code is a numeric character or an alphabetic character. It's a 50/50 split, so it's impossible to say for certain. Therefore, as a result, we have no choice but to distinguish the character by itself (
Therefore, there was a problem that the recognition rate decreased.

The purpose of the present invention is to reduce the number zero (0) from the relationship between the preceding and following characters.
), the alphabetic character 0 (0), the circle mark (○), etc. Even when it is difficult to determine the unidentified code, it is possible to determine the character and improve the recognition rate. The purpose is to provide a method.

[Means to solve the problem]

In order to achieve the above object, the present invention includes a character reading sensor, a central processing unit, and a character recognition section, and the central processing unit is configured to:
Character recognition: After inputting and storing character image data read by a character reading sensor, the character image data is extracted and passed to a character recognition unit, and character codes and shape feature data as recognition results are received from the character recognition unit. In the device, the central processing unit selects a code (hereinafter referred to as undetermined code) representing any of the number zero (0), the alphabetic character O (0), or the circle mark (○) among the character codes received from the character recognition unit. code), it is determined whether the characters located before and after the undetermined code are numbers, alphabets, kanji, hiragana, katakana, or symbols such as periods, and the undetermined code is When the characters located before, after, or before and after are predetermined specific characters, the number zero (0), the alphabetic character O (0),
It determines which of the circles (○) it is, and if the characters located before, after, or before and after the undetermined code are not predetermined specific characters, the characters received from the character recognition unit are Based on the shape characteristic data of the undetermined code character, it is determined whether the undetermined code is a number zero (0), an alphabetic character O (0), or a circle mark (○). did.

[Effect]

In this way, the unidentified code for any of the numbers zero (0), alphabetical characters O (0), circle marks (○), etc. can be determined based on the combination of characters located before and after it. In addition to making a determination, if the determination cannot be made, the determination is made using the shape feature data of the undetermined code character (specifically, the ratio of the height dimension to the width dimension of the character).

FIG. 2 is an explanatory diagram showing a comparison of the height and width ratios of the alphabetic character O (0) and the number zero (0).

As seen in the same figure (a), in the alphabet O (0),
When the height dimension is 1, the width dimension is 0°9, and with the number zero (0), when the height dimension is 1, the width dimension is 0°9.
．． It is 6. Therefore, the threshold value is 0.7 or 0.8.
If you select something like , you can distinguish both characters.

However, what is shown in FIG. 2 is just an example, and since this ratio changes depending on the type of typeface, font, etc., it is necessary to correct the threshold value for each document. In that case, if it can be determined that it is a number zero (0) or an alphabetic letter O (○) based on the combination of characters located before and after it, then the shape feature data (height dimension of the character) and the width dimension) can be memorized and used to determine the threshold value.

〔Example〕

FIG. 1 is a flowchart showing the flow of operations performed by the CPU when implementing the present invention.

As shown in the figure, the CPU extracts character image data in step S1) and recognizes the target character in step S2.
), this is repeated to obtain recognition results for one block.

Next, the character recognition results for that one block are searched to find characters such as the number zero (0), the alphabetic character O (O), and the circle mark (○), which are then marked as unidentified characters. Find the combination of character types before, after, and on both sides of a character. Number zero (0)
, the alphabetic character O (0), a circle mark (○), etc., are consecutive, the entire character group is considered as one character group, and the combination of character types on both sides is determined (step S4).
.

It is determined whether the obtained combination is a combination of an alphabetic character and a number (step S5), and if it is a combination of an alphabetic character and a number, the dimensional ratio of the width and height of the undetermined character is set as a threshold. Step S6. ), if it is high, it is determined to be the letter O (○) (step S8), and if it is low, it is determined to be the number zero (0) (step S7>
.

If it is determined in step S5 that it is not a combination of alphabetic characters and numbers, the process proceeds to step S9 to further examine the combination, and if it is a combination of alphabetic characters and alphabetic characters (or kanji), step Proceed to S13 and press the letter O (
0). If it is determined in step S9 that it is not a combination of alphabetic characters and alphabetic characters (or kanji), step 3
Step S10 further examines the combination, and if it is a combination of numbers and numbers (or kanji), step S12 determines that it is the number zero (0); otherwise, It is determined that the circle is marked (○) (step 511).

If it is determined in step S13 that it is an alphabetic character O (0), or if it is determined that it is a numeric character zero (0) in step S12, the determined width-to-height ratio of the character is memorized, and step The threshold value used in S6 is corrected (step 514).

〔Effect of the invention〕

As explained above, according to the present invention, the number zero (0), the alphabetic character O (0), and the circle mark (○)
Even when it is difficult to determine which of ) has the advantage of improving the recognition rate of characters such as characters.

[Brief explanation of the drawing]

Fig. 1 is a flowchart showing the flow of operations performed by the CPU when implementing the present invention, and Fig. 2 shows the height and width dimensions of the alphabetic character O (○) and the number zero (0). FIG. 3 is a block diagram showing the configuration of an ordinary character recognition device. Explanation of symbols 1... Optical character reading sensor, 2... CPU (Central Processing Unit), 3... Character recognition department agent Patent attorney Akio Namiki

Claims (1)

[Claims] 1) Consisting of a character reading sensor, a central processing unit, and a character recognition section, the central processing unit inputs and stores character image data read by the character reading sensor, and then processes the character image data. In a character recognition device that takes out a character code and passes it to a character recognition unit, and receives the character code and shape feature data as a recognition result from the character recognition unit, the central processing unit includes a numeric character in the character code received from the character recognition unit. Zero (0), English letter O (O), circle mark (○
) (hereinafter referred to as an undetermined code), the characters located before and after the undetermined code are numbers, alphabets, kanji, hiragana, katakana, or symbols such as periods. If the character located before, after, or before and after the undetermined code is a predetermined specific character, the number zero (0) or the alphabetic character (O), circle mark (
○), and if the character located before, after, or before and after the undetermined code is not a predetermined specific character, the unidentified character received from the character recognition unit is determined. Based on the shape feature data of the judgment code character,
The undetermined code is the number zero (0) and the alphabetic letter O (O).
, a character discrimination method characterized by determining which one of the circles (○) it is.