JP2001022883A - Character recognizing system and recording medium for realizing function for the same - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the accuracy of character recognition and erroneous recognition detection in a character recognition processing. SOLUTION: In a character recognizing part 20 to which the image information on an original 10 is inputted, preprocessing/segmenting processing 21 is performed and optical character reader(OCR) recognizing processing is performed by three OCRs (A, B and C) 22, 23 and 24. Merging processing 25 of the plurality of recognized results provided by the respective OCRs is performed and the final recognized result 30 is outputted. In merging processing 25, language processing is performed by using candidate characters only to the part where the recognized results provided by the respective OCRs are not matched. When all the recognized results are different for the same object, language processing is performed while considering the interpolation of candidate characters. When displaying the merged recognized result on a screen or the like, a display method is changed corresponding to the recognized results of the respective OCRs.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition system, and more particularly to a character recognition system applicable to image processing in general, and particularly applicable to a character recognition unit and an erroneous recognition detection unit of an optical character reader (OCR). .

[0002]

2. Description of the Related Art When a large number of originals are input in a short period of time, the accuracy of OCR recognition and the accuracy of detection of erroneously recognized parts are very important in order to make the recognition rate after correcting the recognition result as close to 100% as possible. Come. Conventionally, there has been a method of preparing a plurality of types of OCR engines having completely different methods, comparing the distance values of the OCR engines, and determining a final recognition result by majority decision. However, depending on the OCR engine, there are problems that the distance value is not output, and the recognition result cannot be simply compared because the distance value is not normalized. In addition, for example, when all the recognition results are different using three types of OCR engines, there is a problem that it cannot be determined even by a majority decision.

As a conventional technique, for example, Japanese Patent Application Laid-Open
The character recognition device disclosed in Japanese Patent No. 39430 uses distance values output from a plurality of character recognition units. However, the above publication has a problem that, depending on the OCR, only a candidate for a recognition result is output, a distance value is not output, and a distance value may not be normalized.

[0004]

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and it is an object of the present invention to provide a character recognition system in which the accuracy of character recognition and erroneous recognition detection is improved.

[0005]

According to the first aspect of the present invention, there is provided an input means for inputting character image data of a document, a plurality of character recognizing means for recognizing the character image input to the input means, and Integrated means for integrating recognition results obtained by a plurality of character recognition means.

According to a second aspect of the present invention, in the first aspect of the present invention, the integration means performs language processing using candidate characters only on portions where the recognition results obtained by the plurality of character recognition means do not match. It is characterized by the following.

According to a third aspect of the present invention, in the second aspect of the present invention, the integration means considers interpolation of candidate characters only in a portion where the recognition results obtained by the plurality of character recognition means are all different. Is performed.

According to a fourth aspect of the present invention, in the first aspect of the present invention, the integrated recognition result is obtained according to information obtained when integrating the recognition results obtained by the plurality of character recognition means. It is characterized by changing the display method.

According to a fifth aspect of the present invention, in the fourth aspect of the present invention, the information obtained when the recognition results are integrated is information indicating the degree to which the plurality of recognition results obtained by the plurality of character recognition means are different. It is characterized by the following.

A sixth aspect of the present invention is characterized in that, in the fourth aspect of the invention, information obtained when integrating the recognition results is processing content information of language processing.

According to a seventh aspect of the present invention, in any one of the first to sixth aspects of the present invention, a computer-readable recording medium recording a program for realizing a function of a character recognition system is provided. is there.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The basic idea of the present invention is to use a plurality of OCR engines (character recognition units). FIG. 1 is a diagram for explaining the processing operation of the character recognition system according to the present invention. In the character recognition unit 20 to which the character information of the document 10 has been input, the pre-processing / cutout processing 21 is performed, and the OCR recognition processing is performed by a plurality of OCRs. In this embodiment, three OCRs (A,
B, C) 22, 23 and 24 are prepared, and the recognition result integration process 25 is performed on the plurality of recognition results obtained by each OCR by comparison in consideration of line shift and character shift, and the final recognition result 30 is obtained. Is output. FIG. 2 is a block diagram showing a configuration example of a character recognition device to which the above-described character recognition method is applied. Character image data is stored in the scanner 41
From multiple OCR storage ROM or RAM4
A plurality of recognition results are obtained by the third function.

A processing example of the integration processing 25 shown in FIG. 1 will be described below. [Processing Example 1] In integrating recognition results obtained by a plurality of OCR engines, if even one of the plurality of recognition results does not match, language processing is performed using each candidate character.

[0014] (the recognition result example) a positive solution: Weather forecast for today, it was out. OCR (A) of the results: husband care forecast of today, out. OCR (B) of the results: arrow care forecast of today, out. OCR (C) results: Weather forecast for today, was out.

For example, in the case of the above example, the recognition results of "heaven" of "weather" are all different. In this case, only the parts having different recognition results are subjected to language processing. Candidates for language processing are the first of each OCR (A, B, C).
Only the candidate characters may be included, or the second and subsequent candidate characters may be included. Here, description will be made assuming that the former process is performed. In this case, since the recognition result of only the “heaven” part is different, the input data to the language processing is as follows.

First candidate character: Today's husband affair forecast is off. Second candidate character: arrow Third candidate character: heaven

Actually, when the language processing (morphological analysis) is performed, there is no word "moe" or "arrow", so that "weather" is correctly selected. In other words, the processing of the present embodiment is to determine a part having a different recognition result as a part where the recognition is suspicious, apply language processing only to that part, and automatically correct the recognition result.

[Processing Example 2] In integrating recognition results obtained by a plurality of OCR engines, linguistic processing taking into account interpolation of candidate characters is performed only in a portion where a plurality of recognition results do not all match.

[0019] (the recognition result example) a positive solution: Weather forecast for today, it was out. OCR (A) of the results: husband care forecast of today, out. OCR (B) of the results: arrow care forecast of today, out. OCR (C) results: calamity care forecast of today, out.

For example, in the case of the above example, the recognition results of "heaven" of "weather" are all different, and there is no correct answer. In this case, only the parts having different recognition results are subjected to language processing. Candidates for language processing are OCR
Although the first candidate character of (A, B, C) may be inserted,
Language processing may be performed using the “heaven” part as a wild card.

In this case, since the recognition result of only the “heaven” part is different, the input data to the language processing is obtained by using a wild card (here, *) in the “heaven” part. The wildcard performs language processing on condition that any character may be entered. Further, the language processing may be performed by inserting the candidates of A, B, and C, and the language processing may be performed by a wild card only when a plausible result cannot be obtained.

Next, a description will be given of a process of changing a method of displaying the integrated recognition results on a screen or the like based on information obtained when integrating a plurality of recognition results.

[Processing Example 3] When an integrated recognition result is displayed on a screen or the like, a portion where a plurality of OCR recognition results are different is displayed differently from other portions. (The recognition result example) a positive solution: Weather forecast for today, was out. OCR (A) of the results: husband care forecast of today, out. OCR (B) of the results: arrow care forecast of today, out. OCR (C) results: calamity care forecast of today, out. The end result: Weather forecast for today, was out.

For example, in the case of the above example, since only "heaven" of "weather" has a different recognition result, by displaying only this portion differently from other portions, a character which may be erroneously recognized is displayed. Can be detected and displayed. For example, if a portion where all the OCR recognition results match is displayed in black and a portion where the recognition result is different is displayed in red, a misidentified character can be easily found.

[Processing Example 4] The method of displaying the integrated recognition results on a screen or the like is changed depending on the degree of the recognition results obtained by the plurality of OCRs.

[0026] (the recognition result example) a positive solution: weather forecast for today is out. OCR (A) of the results: now eyes of the husband care forecast was off. OCR (B) of the results: arrow care forecast of today is out. OCR (C) results: calamity care forecast of today is out. The end result: weather forecast for today is out.

For example, in the case of the above example, the recognition results of "day" of "today" and "heaven" of "weather" differ depending on the OCR. In the case of “day”, the recognition result of only one OCR is different, so that the recognition result is displayed in yellow, and “heaven” is displayed in red because the recognition results of all OCRs are different, so that they are displayed in red. It becomes possible.

[Processing Example 6] The method of displaying the integrated recognition result on a screen or the like is changed depending on the content of the language processing. In other words, if the result of the interpolation is performed according to the content of the language processing as in Processing Example 2 above, the result is displayed in red.
It is also possible to change the display, such as displaying the result selected from the candidates as shown in yellow in yellow.

[0029]

As is apparent from the above description, according to the present invention, the accuracy of character recognition and erroneous recognition can be improved by using a plurality of OCRs (character recognition units).

[Brief description of the drawings]

FIG. 1 is a diagram for explaining a processing operation of a character recognition system according to the present invention.

FIG. 2 is a block diagram illustrating a configuration example of a character recognition device to which a character recognition method according to the present invention is applied.

[Explanation of symbols]

Reference Signs List 10: manuscript, 20: character recognition unit, 21: preprocessing / cutout processing, 22: OCR (or Company A) OCR, 23: B
Method (or Company B) OCR, 24 ... C method (or C
OCR, 25: Integration processing of recognition results, 30: Final recognition result, 41: Scanner, 42: Central processing unit, 4
3 ROM or RAM for storing a plurality of OCRs, 44 display, keyboard, mouse, 45 memory, 46 dictionary, 47 work RAM, 48 CD-ROM,
FD drive.

Claims (7)

[Claims] An input unit for inputting data of a character image of a document, a plurality of character recognition units for recognizing the character image input to the input unit, and a recognition result obtained by the plurality of character recognition units A character recognition method, comprising: integration means for integrating a character string. 2. The character according to claim 1, wherein the integration unit performs the language processing using the candidate character only on a portion where the recognition results obtained by the plurality of character recognition units do not match. Recognition method. 3. The integration means according to claim 1, wherein only the parts where the recognition results obtained by said plurality of character recognition means are different from each other.
The character recognition method according to claim 2, wherein language processing is performed in consideration of interpolation of candidate characters. 4. The display method according to claim 1, wherein a display method of the integrated recognition result is changed according to information obtained when integrating the recognition results by the plurality of character recognition units. Character recognition method described in. 5. The character according to claim 4, wherein the information obtained when the recognition results are integrated is information of a degree at which a plurality of recognition results by the plurality of character recognition means are different. Recognition method. 6. The character recognition method according to claim 4, wherein information obtained when integrating the recognition results is processing content information of language processing. 7. A computer-readable recording medium on which a program for realizing the function of the character recognition method according to claim 1 is recorded.