JPH04270480A - Character recognizing method - Google Patents

Abstract

PURPOSE:To reduce judgement for the validity of character string contents for each field executed by a recognition post-processing up to the moment by judging whether the character string as the recognized result of the field satisfies a stored condition or not. CONSTITUTION:Based on information stored in a slip format storage part 13, a character recognition part 14 segments a character part from the image information and executes a character recognition processing, and the results are collected as the recognized result character string of each field by a field recognized result synthesizing part 15. Afterwards, a field recognized result consistency judgement part 17 judges it while referring to a matching condition stored in a field recognized result matching condition storage part 16 whether the recognized result character string of each field satisfies the condition or not. Therefore, the recognized result character string of the field not to satisfy the recognized result matching condition registered in advance can be detected.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition method for recognizing a form based on format information registered in advance, and more particularly to a character recognition method that can reduce post-recognition processing.

0002

BACKGROUND OF THE INVENTION Conventionally, character recognition devices in financial institutions such as securities companies and banks read characters written on a form as a set of pixels when recognizing a form based on format information registered in advance. Character recognition was performed character by character using a dictionary and pattern matching hardware. For a field, which is the smallest unit of a form, the recognition result for that field is simply the concatenation of the recognition results for each character that makes up the field, and the validity of the recognition results as a field can be interpreted. had not been carried out. Therefore, the consistency check between the recognition result of each field and the field meaning had to be performed separately from the character recognition device by a program that performs post-recognition processing. Therefore, for example, even if the content of the amount field is a negative number, it may be output as a recognition result of the character recognition device. Regarding conventional character recognition methods, for example, Japanese Patent Application Laid-Open No. 61-243581, Japanese Patent Application No. 1-289716, and Japanese Patent Application No. 2-230927
It is stated in the number etc.

0003

[Problem to be Solved by the Invention] The above conventional technology does not interpret the validity of the character recognition result as a field, so the recognition result may contain an obvious error compared to the content that should be written in the field. Even if it was obtained, it could not be detected. Furthermore, it was not possible to detect that the recognition result was inappropriate for a field in which the field was filled in incorrectly, such as when a name is written in the amount field. As a result, character recognition devices may output recognition results for fields that contain free errors as is, and the validity of field recognition results had to be checked as post-recognition processing outside the character recognition device. . The purpose of the present invention is to improve such problems, store the conditions that should be satisfied by the recognition result character string of each field included in the field, and store the recognition result character string of each field obtained as a result of character recognition. An object of the present invention is to provide a character recognition method that can reduce post-recognition processing by detecting when a string does not satisfy the condition.

0004

[Means for Solving the Problems] In order to achieve the above object, the character recognition method of the present invention includes a means (reading section) for reading a form as image data, and a means (reading section) for storing image information read by the reading section. image information storage unit) and means for storing form format information (
A recognition result character string as a field is created from the recognition result for each character obtained by the character recognition unit. means (field recognition result combination unit), means (field recognition result matching condition storage unit) for storing matching conditions that the recognition result character string should satisfy for each field included in a form; means (field recognition result consistency determination unit) for determining whether or not the resulting character string satisfies the matching condition; A feature of this method is that it detects entry errors and character recognition errors included in the recognition results for each field by making a judgment and outputting the recognition result character string and the judgment result.

[0005]

[Operation] In the present invention, a document is read as image data by the reading section, the read image information is stored in the image information storage section, and the character recognition section uses the information stored in the document format storage section as the basis. Then, character portions are extracted from the image information, character recognition processing is performed, and the resulting character recognition results are compiled into a recognition result character string for each field by a field recognition result synthesis section. Thereafter, the field recognition result consistency determination unit refers to the conditions (consistency conditions that should be satisfied by the recognition result as a field) stored in the field recognition result consistency condition storage unit for each field, and Determine whether the column satisfies the condition. Therefore, it is possible to detect a recognition result character string in a field that does not satisfy the recognition result matching condition registered in advance. This allows the character recognition device to perform a validity check on the recognition result character string for each field, which has conventionally been performed in post-recognition processing, thereby reducing post-recognition processing.

[0006]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a functional configuration diagram of a character recognition device according to an embodiment of the present invention. The character recognition device of this embodiment uses a CPU
The functional configuration is shown in FIG. 1, and includes a processing device such as, a storage device such as a magnetic disk and an optical disk device, and an input/output device such as a keyboard, CRT, and OCR. In FIG. 1, 11 is a reading unit that reads a form as image data; 12 is an image information storage unit that stores image information read by the reading unit 11; 13 is a form format storage unit that registers a form format in advance;
14 is a character recognition unit which refers to the form format storage unit 13 and extracts and recognizes a character part from image information;
5 is a field recognition result synthesis unit that synthesizes recognition results for each field from the recognition results for each character obtained by the character recognition unit 14; 16 is a field recognition result that stores recognition result matching conditions for each field of a form; A matching condition storage unit 17 is a recognition result consistency determination unit that determines whether the recognition results for each field obtained from the field recognition result synthesis unit 15 satisfy the conditions stored in the recognition result matching condition storage unit 16. These constitute the character recognition device 1. In addition, this field recognition result matching condition storage unit 16
The configuration of is shown in FIG. 2 and 3 are explanatory diagrams of a field recognition result matching condition storage unit in an embodiment of the present invention. In FIGS. 2 and 3, (a) is a format of a form used for online automated transactions of financial institutions, and (b) is a field recognition result matching condition storage unit 16 that registers matching conditions for each field in the format of (a). This is a configuration example. In this embodiment, when recognizing a form filled in according to the format of (a), it is determined whether each field satisfies the matching condition shown in (b). This reduces the occurrence of obvious errors from the operator's point of view, such as English characters being mixed in the customer name or negative numbers being output as the amount, thereby reducing the burden on the operator. In the case of Figure 2, there is only one amount column on the form, but in Figure 3
In this case, a total column is added for summing the values of multiple fields, and the sum of the values of Amount 1 and Amount 2 is entered here. Therefore, the matching condition for the "total" field shown in FIG. 3(b) specifies that the sum of the "amount 1" and "amount 2" fields is the same. Next, the operation of the character recognition device 1 of this embodiment will be described. In this embodiment, first, recognition processing is performed on a character-by-character basis in the same way as before.
The results are collected for each field and a recognition result character string is synthesized. That is, when a form is inserted into the character recognition device 1 , it is read by the reading section 11 , and the read image information is written into the image information storage section 12 and sent to the character recognition section 14 . Character recognition part 1
In step 4, fields in which characters are written are extracted from the image information, recognition processing is performed character by character, recognition candidate characters are output for each character, and the results are sent to the fields by the field recognition result synthesis unit 15. and the recognition result character string as a field is synthesized. Furthermore, the recognition result character string is judged and output based on matching conditions. That is, the recognition result character string of each field output from the field recognition result synthesis section 15 is sent to the field recognition result consistency determination section 17. The field recognition result consistency determination unit 17 refers to the conditions that the recognition result character string of the field stored in the field recognition result consistency condition storage unit 16 should satisfy, and determines whether the recognition result character string satisfies the conditions. Determine whether Thereby, for each field, the character recognition device 1 combines the recognition result character string synthesized by the field recognition result synthesis section 15 with the matching condition for that field (stored in the field recognition result matching condition storage section 16). It outputs both the results as to whether the conditions (conditions) are met or not.

[0007]

According to the present invention, the character recognition device includes a field recognition result matching condition storage unit that stores the conditions that the recognition result should satisfy for each field included in a form, and By providing a field recognition result consistency determination unit that determines whether the conditions stored in the field recognition result consistency condition storage unit are satisfied, it is possible to match the recognition result character string and the meaning of the field for each field. It is possible to determine whether or not the requirements are met. Therefore, if the recognition result for each field contains an obvious error, it can be detected by the character recognition device, and the validity of the content of the recognition result character string for each field, which has been performed in post-recognition processing, can be improved. Judgment can be reduced.

[0008]

[Brief explanation of the drawing]

FIG. 1 is a functional configuration diagram of a character recognition device in an embodiment of the present invention.

FIG. 2 is an explanatory diagram of a field recognition result matching condition table in an embodiment of the present invention.

FIG. 3 is an explanatory diagram of a field recognition result matching condition table in an embodiment of the present invention.

[Explanation of symbols]

1 Character recognition device 11 Reading section 12 Image information storage section 13. Form format storage section 14 Character recognition section 15 Field recognition result synthesis section

Claims (1)

[Claims] 1. A means for reading a form as image data, a means for storing image information read by the reading means, a means for storing format information of the form, and a character portion cut out from the image information for recognition. and a means for creating a recognition result character string as a field from the recognition result for each character obtained by the recognition means, wherein the recognition result is generated for each field included in a form. A means for storing a matching condition that a character string should satisfy, and a means for determining whether or not the recognition result character string satisfies the matching condition are provided. A character recognition method characterized by determining whether a matching condition is satisfied and outputting the recognition result character string and the determination result.