JPH04109379A - Ocr system - Google Patents

Abstract

PURPOSE:To easily and efficiently correct a misread character by providing the OCR system with a reliability index setting means and a means for retrieving a character to be corrected. CONSTITUTION:A recognition part 14 segments the character pattern of each character in each field of a document in accordance with an instruction from an OCR control device 20 through a scanner 12, collates the segmented pattern with a reference pattern (dictionary), finds out similarity value between both the patterns, decides the shape (e.g. reference character, deformed character, blurred character, dirty character, or continued/projected character) of the entered character, and finds out a reliability index in each character and field based upon the similarity value and the shape of the character. A control part 22 in the control device 20 checks the existence of a misread character and executes the correction (edition) or the like of the character based on the results of character recognition and reliability index obtained by the OCR device 10, and displays the processed result on a CRT 24. The specification of processing execution and the input of data are executed through a keyboard 26.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Application Field) The present invention relates to an optical character reader (OCR) system having a function of correcting misread characters in the result of character reading by an OCR.

(Prior Art) In general, a check method for checking the results of reading characters written on forms etc. using an optical character reading device is as follows.
A check digit method and a total check method are used. In these checking methods, it is checked whether or not the reading result contains misread characters in each field and each form. When the user is shown that the text has been misread, the user visually checks each character in each field and each form, finds the misread character, and corrects the misread character.

(Problems to be Solved by the Invention) As described above, in the conventional check method, it is necessary to visually check each character in each field and each form, and it is necessary to visually check each character in each field and in each form. It takes a lot of time to find out which characters are misread characters, and it is a heavy workload.

The present invention has been made in view of the above-mentioned points, and it is possible to easily and efficiently correct misread characters.
The purpose is to provide a CR system.

[Structure of the Invention] (Means for Solving the Problems) The present invention provides an OCR system that has a function of correcting misread characters for character reading results, and provides a reliability index that calculates a reliability index indicating the certainty of character reading results for each predetermined unit. The apparatus includes a setting means and a search means for searching for a character to be corrected for misread characters based on the confidence index determined by the confidence index setting means.

(Operation) According to this configuration, characters to be corrected are searched based on the confidence index, so characters with a high possibility of error are prioritized and corrected, or a check is made to see if they have been misread. can be done.

(Example) Hereinafter, an example of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of an OCR system according to the same embodiment. The OCR system shown in FIG. 1 is composed of an OCR that performs character recognition processing.

The OCR device 10 includes a scanner unit 12 that inputs a form image, a recognition unit 14 that performs character recognition processing on the form image input by the scanner unit 12, and an OC that controls the interface with the OCR control device 20.
It is composed of an RC interface 6. The recognition unit 14, in response to instructions from the OCR control device 20,
Character recognition processing is performed for each character in each field provided in the form.

In the character recognition process, a character pattern is cut out for each character, and the character pattern is compared with a standard pattern (dictionary) to obtain a similarity value. Also, based on the cut out character pattern, the shape of the characters written on the form (for example, standard characters, modified characters, faded characters, smeared characters, continuation/extrusion characters) is determined. Further, the recognition unit 14 obtains a confidence index indicating the probability that the character reading result is correct for each character and each field based on the similarity value and the shape of the character.

The OCR control device 20 includes a control unit 22 that controls the entire OCR system including control of the OCR device 1o, user interface (I10 device) control, editing of character reading results, etc., and a CRT that is a user interface.
24. It is composed of a keyboard 26 and an OCRC open control device interface 8 that controls the interface with the OCR device IO. The control unit 22 checks whether or not there are misread characters, corrects (edits) the characters, etc., based on the character recognition result obtained by the OCR device 1o and the reliability index. The CRT 24 displays various types of information such as character recognition results. The keyboard 28 is used by the user to instruct the OCR system to execute processing, input data, and the like.

Next, the operation of this embodiment will be explained.

First, a document whose characters are to be read in this embodiment will be explained. FIG. 2 shows an example of the form. It is assumed that the form shown in FIG. 2 is provided with a plurality of fields (a, -f) for entering characters, and numbers may be entered therein. In particular, it is assumed that field f is specified to contain the sum of the numbers entered in fields a to g.

First, the scanner unit 12 of the OCR device 10 scans the second
Enter the form image for the form shown in the figure. When this form image is transferred from the scanner section 12 to the recognition section 14, character recognition processing is executed according to instructions from the OCR control device 20. The recognition unit 14 cuts out a character pattern for each character in each field provided in the form and performs character recognition processing. At this time, for each character, the similarity value with the standard pattern (dictionary) and character properties (character shape) are determined. Then, the recognition unit I4 determines the reliability of each character as a numerical value (reliability index) as an answer of the character recognition result based on the similarity value and the shape of the character.

Generally, when reading characters using an optical character reading device, some characters tend to be more likely to be misread. Furthermore, the level of difficulty in reading generally differs depending on the character shape, for example, standard characters, modified characters, blurred characters, dirty characters, continuations and overflowing characters. Things like this,
In the form of a comprehensive and statistically reliable index of character recognition results,
By obtaining it as additional data to the recognition results,
If the presence of a misread character is found in the misreading check on the OCR control device 20 side, the misread character can be easily searched.

In this embodiment, it is assumed that the recognition unit 14 obtains a confidence index indicated by the ]O level for each character. As a result, characters with a confidence index of 1 to 4 are treated as diect characters.
For characters with a confidence index of 5 to ↑0, the first candidate (the one with the highest similarity value) is output as a character code as a character recognition result.

Furthermore, the reliability index is output for each character as additional data of the recognition result. FIG. 3 shows the reliability index corresponding to each character written in the form shown in FIG. 2.

Furthermore, after each character in the field is recognized, a confidence index for the field is determined based on the confidence index of each character, and is output as additional data for each field. Note that here, the value of the reliability index for each field is assumed to be the minimum value of the reliability index for each character in the field. Figure 4 shows the correspondence between the characters entered in each field, the characters (#), the result, and the reliability index for each field.

In this way, recognition processing is performed for each character in the recognition unit 14, and when a reliability index is obtained, the result is
The data is transferred to the control section 22 via the CCD interface 6 and the OCR control device interface 28.

The control unit 22 determines whether the reading results for fields a to e (a, b, c, d, e) match the reading results for field f (f). That is, a + b + c + d 10e −f
- Calculate whether or not (1) holds true. As a result, if the formula (1) is satisfied, it is assumed that the reading has been performed correctly, and the control section 22 gives the recognition section 14 a feed command instructing the reading of the next form.

On the other hand, if the above formula (1) does not hold, the reading result for one sheet is displayed on the CRT 24.

At this time, the control unit 22 sets the cursor position to the character with the lowest reliability index in the field with the lowest reliability index in the reading results for one sheet.

Then, it waits for input of a character correction instruction from the keyboard 26 for the character at the cursor position.

Characters to be corrected (characters indicated by the cursor) are set in order of decreasing confidence index. In other words, characters to be corrected are displayed in descending order of probability of error.

In this way, by providing the reading status in the OCR device 10 to the OCR control device 2o in the form of a reliability index, misread characters can be easily found without visually checking each character. It is possible to significantly reduce the work burden of correction and to perform the work efficiently.

In the embodiment described above, the reliability index is determined for each character and each field, but it is also possible to determine and utilize the reliability index for each sheet.

In a system having a total patch check function, it is possible to easily infer in which sheet there is a high possibility that a misread character is present from the reliability index for each sheet. In other words, there is a high probability that the sheet containing the character with the lowest reliability index is the sheet that is misread, and based on this, the order of correcting the misread characters for each sheet can be determined.

Although the confidence index is determined based on the degree of similarity and the shape of the characters, it may be set based on only the degree of similarity, for example. Also, other factors may be added.

[Effects of the Invention] As described above, according to the present invention, a confidence index is obtained for the recognition result and misread characters are searched based on this confidence index, so misread characters can be corrected easily and efficiently. It is possible.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of an OCR system according to an embodiment of the present invention, FIG. 2 is a diagram showing an example of a form to be processed, and FIG. 3 shows the actual results and field units in FIG. FIG. 10...OCR device, 12...scanner section, 14.
...Recognition section, 1B...OCR interface section, 2
0...OCR control device, 22...control unit, 24...
- CRT, 26...Keyboard, 28...OCR control device interface.

Claims (1)

[Claims] OC having a function of correcting misread characters for character reading results
In the R system, a confidence index setting means for obtaining a confidence index indicating the certainty of character reading results for each predetermined unit, and a character to be corrected for misread characters based on the confidence index obtained by the confidence index setting means. An OCR system characterized by comprising a search means for searching, and the following.