JPH0737034A - Optical character reader - Google Patents

Abstract

PURPOSE:To provide an optical character reader which can recognize the characters printed or handwritten on a document at a high speed and with high accuracy. CONSTITUTION:An optical character reader consists of a printed character recognizing means 30 which recognizes the character data as the printed characters, a handwritten character recognizing means 40 which recognizes the character data as the handwritten characters, a character reading means 50 which optically reads the written characters and also converts these characters into the character data, a character data output means 10 which outputs the character data received from the means 50 to both means 30 and 40 at a time, and a recognizing result selecting means 20 which selects the recognizing result of the means 30 or 40.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an optical character reader, and more particularly to an optical character reader suitable for sorting mail.

[0002]

2. Description of the Related Art Optical character readers are used for speeding up sorting work of mail.

However, since there are a mixture of printed mail addresses and handwritten mail items, it is necessary to accurately recognize each of them.

Generally, the recognition method is different between printed characters and handwritten characters. That is, printed characters are recognized by pattern matching, and handwritten characters are recognized by feature extraction. In many cases, if the print character is recognized by the handwritten character recognition method, it is not correctly recognized, and if the handwritten character is recognized by the print character recognition method, it is often not correctly recognized.

Therefore, in order to sort the mail items with high accuracy, it is first necessary to judge whether the address is a printed character or a handwritten character.

Japanese Unexamined Patent Publication (Kokai) No. 57-193874 discloses a method of determining whether an address is a printed character or a handwritten character based on the average thickness of the stroke of the character and the size of the character.

In Japanese Patent Laid-Open No. 61-255490, a document having a printed key item and a handwritten data item is read, the key item is first recognized by a handwritten character recognition means and a type recognition means, and the recognition result is obtained. A method of recognizing a data item based on it is disclosed.

[0008]

However, in the above-mentioned conventional example, in Japanese Patent Laid-Open No. 57-193874, printed characters have a larger average stroke thickness of characters than handwritten characters and a smaller character size. Since the print character and the handwritten character are distinguished under the condition, there is a disadvantage that a recognition error is likely to occur.

Further, in Japanese Patent Laid-Open No. 61-255490, since a key item printed and a data item handwritten are required, there is no key item or the key item and the data item are not in a predetermined positional relationship. In that case, there was a problem that it could not be recognized.

Further, in Japanese Patent Laid-Open No. 61-255490,
The final recognition result is obtained because the recognition processing by the handwritten character recognition means and the recognition processing by the type recognition means are performed sequentially and the recognition results are compared to distinguish between printed characters and handwritten characters. There was also the inconvenience that it took time to get it.

[0011]

SUMMARY OF THE INVENTION It is an object of the present invention to provide an optical character reading apparatus which can improve the inconvenience of the conventional example and can recognize printed characters and handwritten characters written in a document at high speed and with high accuracy. To do.

[0012]

Therefore, according to the present invention, a print character recognizing means for recognizing character data as a print character, a handwritten character recognizing means for recognizing character data as a handwritten character, a document, etc. A character reading unit that reads the written characters and converts them into character data, a character data output unit that simultaneously outputs the character data from the character reading unit to the print character recognition unit and the handwritten character recognition unit, and a print character recognition unit. It is configured to include a recognition result selection unit that selects either one based on the recognition result and the recognition result of the handwritten character recognition unit. This aims to achieve the above-mentioned object.

[0013]

The character reading means scans the surface of the document on which the characters are written, converts the light and shade into binary data, and outputs it as read data to the character data output means.

The character data output means outputs the read data from the character reading means to the print character recognition means and the handwritten character recognition means at the same time.

The print character recognition means detects a character area and cuts out the character data based on the read data from the character data output means, and recognizes each character data assuming that it is a print character.

At the same time, the handwritten character recognition means detects the character area and cuts out the character data based on the read data from the character data output means, and recognizes each character data assuming that it is a handwritten character.

The recognition result selecting means compares the recognition result of the print character recognizing means with the recognition result of the handwriting character recognizing means, and selects one having a higher recognition accuracy.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to FIGS.

In the embodiment shown in FIG. 1, a print character recognizing means 30 for recognizing character data as a print character, a handwriting character recognizing means 40 for recognizing character data as a handwritten character, and a written character are optically recognized. And a character data output unit 10 for simultaneously outputting the character data from the character reading unit 50 to the print character recognizing unit 30 and the handwritten character recognizing unit 40. Recognition result selecting means 20 for selecting either one based on the recognition result by the print character recognizing means 30 and the recognition result by the handwriting character recognizing means 40.
Composed of and.

Here, as shown in FIG. 2, the print character recognizing means 30 has a print character data storage section 35 for storing character data of a typeface which is supposed to be used as a print character in advance, and a print character. Data character reading means 5
A print destination area detection unit 31 that detects a destination area based on read data from 0, a print character cutout unit 32 that extracts character data from the destination area detected by the print destination area detection unit 31, and a print character cutout unit A print character recognition unit 33 that compares the character data extracted in 32 with the character data stored in the print character data storage unit 35 to perform individual character recognition, and the character data recognized by the print character recognition unit 33. A print address recognition unit 34 that recognizes as an address by
It has and.

In the print character recognition unit 33, the extracted character overall shape is compared with the overall shape of all the printed characters stored in advance in the print character data storage unit 35 by the pattern matching method, and the highest similarity is obtained. Output characters as recognition result.

However, the method of recognizing print characters is not limited to this.

As shown in FIG. 3, the handwritten character recognition means 40 has a handwritten character feature storage section 45 for storing the characteristics of characters which are supposed to be used as handwritten characters in advance, and a character reading means 50. A handwritten address area detection unit 41 for detecting an address area based on the read data from the handwriting and a handwritten character cutout unit 4 for extracting character data from the address area detected by the handwritten address area detection unit 41.
2, and a handwritten character recognition unit 43 that performs feature recognition by extracting features from the character data extracted by the handwritten character cutout unit 42 and comparing the feature data stored in the handwritten character feature storage unit 45 with each other. The handwritten character recognition unit 43 is provided with a handwritten address recognition unit 44 that recognizes the character data as an address.

In the handwritten character recognizing unit 43, the complicated characters are decomposed into "hen" and "making", and the features thereof are extracted and stored in the handwritten character feature storage unit 35 in advance. "Hen" and "Tsukuri"
Character data is compared by comparing with the characteristic data of and the obtained "hen" and "make" are combined as in the original.

However, the method for recognizing handwritten characters is not limited to this.

As the character reading means 50, C
An image sensor using a CD is used.

Next, the operation of this embodiment will be described.
Here, for the purpose of explanation, as an address, "1-1, Nisshin-cho, Fuchu-shi"
It is assumed that “0” is written in print characters.

.. The character reading means 50 scans the surface of the postal matter on which the address is written, converts the light and shade into binary data, and sequentially outputs the binary data to the character data output means 10.

.. The character data output means 10 simultaneously outputs the read data from the character reading means 50 to the print destination area detection unit 31 of the print character recognition unit 30 and the handwritten destination area detection unit 41 of the handwritten character recognition unit 40.

.. Based on the read data from the character data output unit 10, the print address area detection unit 31 considers the area where the area of the high density portion exceeds a preset value as the address area.

Similarly, the handwritten address area detection unit 41 also detects the handwritten address area based on the read data from the character data output means 10.
A region in which the area of the high-density portion exceeds a preset value is regarded as the address region.

.. The print character cutout unit 32 extracts character data from the address area detected by the print address area detection unit 31 and outputs it to the print character recognition unit 33. That is, the read data in the address area is subjected to image processing such as Fourier calculation processing to remove dirt and noise.

Similarly, the handwritten character cutout unit 42 extracts character data from the address area detected by the handwritten address area detection unit 41 and outputs it to the handwritten character recognition unit 43. That is, the read data in the address area is subjected to image processing such as Fourier calculation processing to remove dirt and noise.

.. The print character recognition unit 33 compares the overall shape of each character extracted by the print character cutout unit 32 with the overall shape of all print characters stored in advance in the print character data storage unit 35 by a pattern matching method, The character with the highest degree of similarity is output as the recognition result.
That is, since the entire shape of the first character extracted by the print character cutout unit 32 is most similar to the character “fu” stored in the print character data storage unit 35 in advance, the character code of “fu” is set. Is output to the print address recognition unit 34.

The handwritten character recognizing section 43 puts each character extracted by the handwritten character slicing section 42 into a "creation" or a "creation".
The extracted "features" and "features" are compared with the feature data of "features" and "features" stored in the handwritten character feature storage unit 45 in advance. Character recognition is performed by combining as in the original. That is, since the first character extracted by the handwritten character cutout unit 42 is a combination of "square", "a", and "dimension", it is determined that the character is "fu".
The character code of “fu” is output to the handwritten address recognition unit 44.

.. The print address recognition unit 34 checks whether the recognition result by the print character recognition unit 33 is valid as an address, and if it is not valid, the print character recognition unit 33.
Request re-recognition. On the other hand, if the recognition result in the print character recognition unit 33 is a valid address, the recognition result is output to the recognition result selection means 20. Here, for the sake of explanation, the recognition result by the print character recognizing unit 33 is "1-1 Fushin-shi Nisshin-cho 1-1.
It is assumed that it is "0".

Similarly, the handwritten address recognition unit 44 checks whether or not the recognition result by the handwritten character recognition unit 43 is valid as an address, and if it is not valid, requests the handwritten character recognition unit 43 to re-recognize. . On the other hand, handwritten character recognition unit 4
If the recognition result in 3 is a valid address, the recognition result is output to the recognition result selecting means 20. Here, for the sake of explanation, it is assumed that the recognition result by the handwritten character recognition unit 43 is “Fuschu City Nisshin Town”.

.. The recognition result selection means 20 compares the recognition result from the print address recognition unit 34 with the recognition result from the handwritten address recognition unit 44, and selects the one with higher recognition accuracy. That is, in the present embodiment, the recognition result from the print address recognition unit 34 is "1-10 Nisshincho, Fuchu-shi", whereas the recognition result from the handwritten address recognition unit 44 is only "Nisshin-cho, Fuchu-shi". Therefore, the recognition result from the print address recognition unit 34 is selected.

Further, even if print characters and handwritten characters are mixed, it is possible to recognize them at high speed and with high accuracy.

[0040]

Since the present invention is constructed and functions as described above, according to this, it is possible to simultaneously perform the recognition process as a print character and the recognition process as a handwritten character, which is why it is described in the document. It is possible to provide an unprecedented excellent optical character reader capable of recognizing printed characters and handwritten characters at high speed and with high accuracy.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing an embodiment of the present invention.

FIG. 2 is a detailed configuration diagram showing details of a print character recognition unit in FIG.

FIG. 3 is a detailed configuration diagram showing details of a handwritten character recognition unit in FIG.

[Explanation of symbols]

 10 Character Data Output Means 20 Recognition Result Selection Means 30 Printed Character Recognition Means 31 Printed Address Area Detection Unit 32 Printed Character Cutout Unit 33 Printed Character Recognition Unit 34 Printed Address Recognition Unit 35 Printed Character Data Storage Unit 40 Handwritten Character Recognition Unit 41 Handwritten Addressee Area detection unit 42 Handwritten character cutout unit 43 Handwritten character recognition unit 44 Handwritten address recognition unit 45 Handwritten character feature storage unit 50 Character reading means

Claims (3)

[Claims] 1. A print character recognizing means 30 for recognizing character data as a print character, a handwriting character recognizing means 40 for recognizing character data as a handwritten character, and a character written in a document or the like as well as a character. The printer includes a character reading unit 50 for converting into data, and a character data output unit 10 for simultaneously outputting the character data from the character reading unit 50 to the print character recognition unit 30 and the handwritten character recognition unit 40. An optical character reader equipped with a recognition result selecting means (20) for selecting one of them based on the recognition result by the means (30) and the recognition result by the handwritten character recognition means (40). 2. The print character recognizing means 30 is based on read data from a print character data storage section 35 in which character data of a typeface used as a print character is stored in advance, and read data from the print character data character reading means 50. A printing address area detecting unit 31 for detecting a destination address area, a printing character cutting unit 32 for extracting character data from the address area detected by the printing address area detecting unit 31, and a printing character cutting unit 32 for extracting the character data. A print character recognition unit 33 that compares character data with the character data stored in the print character data storage unit 35 to recognize each character, and the character data recognized by the print character recognition unit 33 is used as an address. The optical character reader according to claim 1, further comprising a print address recognition unit (34) for recognizing. 3. The handwritten character recognizing means 40 is based on the handwritten character characteristic storage section 45 which stores the characteristics of the characters which are supposed to be used as handwritten characters in advance, and the read data from the character reading means 50. A handwritten address area detection unit 41 for detecting an address area, a handwritten character cutout unit 42 for extracting character data from the address area detected by the handwritten address area detection unit 41, and a character extracted by the handwritten character cutout unit 42. A handwritten character recognition unit 43 that extracts a feature from the data and compares it with the feature data stored in the handwritten character feature storage unit 45 to perform individual character recognition.
The optical character reading device according to claim 1, further comprising: a handwritten address recognition unit 44 that recognizes an address based on the character data recognized by the handwritten character recognition unit 43.