JP2002123815A - Filing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the efficiency and accuracy of the correcting work of character errors by retrieving item data in the range of identity from a plurality of documents respectively and comparatively displaying the item data on a screen. SOLUTION: The document image data read by an image reading section 110 are stored in an image data storage section 101 and an image database 120. A structure analyzing section 102 conducts the character recognition of a layout in the document image data and stores a recognition result in a structural analysis result database 130. When an item specifying section 103 accepts the specification of an item name from an operator, an item data acquiring section 104 acquires the item data corresponding to the item name from the structural analysis result database 130. When the operator specifies the item data, an item data collating section 106 extracts the item data in the range of the specified identity from the item data acquiring section 104 and concurrently displays the item data on the screen.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a filing apparatus for storing a document such as a document as image data.

[0002]

2. Description of the Related Art A filing apparatus converts documents, forms, drawings, etc., converted into image data by an image reading apparatus.
This is a device that stores data in a storage medium such as a magneto-optical disk.
According to this filing device, a large amount of documents can be stored and managed by one device, so that space saving of the document storage facility can be achieved.

[0003] When documents are stored and managed using a filing device, it is necessary to search for a document intended by the user from a group of documents stored in a storage medium. In particular, when a large number of documents are stored, a function for efficiently performing this operation is required.

As a search function of stored documents, for example, a function using a keyword or a classification code is known. In this search function, an operator inputs a keyword or a classification code when storing a document in a filing device. Each document is stored in a storage medium in the apparatus in association with a keyword or a classification code input at the time of storage. And
When the operator inputs an arbitrary keyword or classification code during document search, a list of documents corresponding to the keyword or the like is created and displayed on the screen.

[0005] As a kind of search technique using a keyword, a technique for searching for an arbitrary character string from the entire text of a document stored in a filing device has been proposed (see Japanese Patent Application Laid-Open No. 8-87528). In this technique, a character rectangle is extracted from a document read by an image reading device, converted into text data using a character recognition technology, and stored in the device in association with the image data. At the time of document search, a character search is performed using the text data, and image data of the document including the search character string is displayed on the screen.

[0006]

When a character portion in a document is converted into text data and is used for search,
Incorrect character recognition can be a problem. For this reason, in the filing apparatus (described above) disclosed in Japanese Patent Application Laid-Open No. 8-87528, all the text data candidates during the character recognition processing are stored in the apparatus as search text data. For example, "-",
"One" and "-" are difficult to distinguish by character recognition technology. Therefore, in the filing apparatus disclosed in the above publication, when the image data “search” is converted into text data for search, “search”, “search”,
It is possible to search for any of "search". Thus, the filing apparatus disclosed in the above publication can perform a document search with a certain degree of accuracy even if there is an error in character recognition.

However, even in a filing apparatus having such a search function, there are cases where it is desired to correct an error in character recognition. In such a case, in the filing apparatus of the above publication, the document and the text data must be displayed on the screen, and the operator must visually check for a character error. Therefore, the efficiency of the correction work is low, and the accuracy of the correction work is limited.

For these reasons, there has been a demand for a filing apparatus capable of efficiently and accurately correcting a character error.

[0009]

A filing apparatus according to the present invention fetches document image data from the outside and temporarily stores the document image data, and stores the document image data in an external image data storage means. Extract the layout from the document image data stored in the image data storage means, determine the item name entry field and the item data entry field of the layout, and extract the character rectangle from the item name entry field and the item data entry field And a structure analysis unit that performs character recognition of these character rectangles and stores the item names and item data obtained by the character recognition in an external structure analysis result storage unit.
Item designation means for designating a type or a plurality of types; item data acquisition means for acquiring item data corresponding to the item name designated by the item designation means from the structural analysis result accumulation means; and item data acquisition means The operator designates one or more types of item data acquired by the operator, and extracts item data within a range of the same as the designated item data from the item data acquired by the item data acquiring means. And item data collation means for displaying the data.

According to such a configuration, the operator can specify an item name and item data, extract the item data corresponding to these specifications, and simultaneously display the item data on the screen. Therefore, item data of a plurality of documents can be compared on the screen, so that an error in character recognition can be corrected efficiently and accurately.

[0011]

Embodiments of the present invention will be described below with reference to the drawings. In the drawings, the size, shape and arrangement relationship of each component are only schematically shown to the extent that the present invention can be understood, and the numerical conditions described below are merely examples. .

FIG. 1 is a block diagram conceptually showing the configuration of a filing apparatus according to this embodiment.

As shown in FIG. 1, the filing apparatus 100 includes an image reading unit 110, an image database 12
0, connected to the structural analysis result database 130 and the image display unit 140. The filing apparatus 100 includes an image data storage unit 101, a structure analysis unit 10
2, an item specifying unit 103, an item data obtaining unit 104, an item data storage unit 105, an item data collating unit 106, and an item data post-processing unit 107.

The image reading section 110 optically reads a document such as a form, and converts optical information obtained by the reading into a quantized electric signal (ie, image data).

The image database 120 has a storage medium such as an optical disk, a magneto-optical disk, and a hard disk. The image database 120 receives the image data and the ID from the image data storage unit 101 and stores the image data and the ID in the storage medium (described later).

The structural analysis result database 130 has a storage medium such as an optical disk, a magneto-optical disk, and a hard disk. Structural analysis result database 13
No. 0 receives the structure analysis result from the structure analysis unit 102 and stores it in this storage medium (described later).

The image display unit 140 is a CRT (Cathode Ray).
Tube) etc., and the item data collating unit 1
06 and the data input from the item data post-processing unit 107 and the like.

The image data storage unit 101 has an image memory (not shown) inside. The image data storage unit 101 receives image data from the image reading unit 110, temporarily stores the image data in the image memory, and assigns an ID to each image data. Each image data and ID are sent to the image database 120.

The structure analysis unit 102 includes the image data storage unit 1
For example, a structural analysis of the document image data stored in the storage unit 01 is performed.
Through this analysis, data such as “item name” and “item data” can be obtained. These data are text data obtained by character recognition in the case of a character, and are coordinate data indicating a position on a document in the case of not a character. For example, in the case of image data of a form as shown in FIG. 2, "address", "name", "age", "sex", and "seal" are obtained as text data as item names, and these items are obtained. As item data corresponding to each name, "XX
Prefecture △△ City □□ Town XX Address ”,“ Nippon Taro ”,“ 26 ”,
The text data of "male" and the coordinate data indicating the position of the seal image are obtained. As a result of character recognition, if there are a plurality of candidate characters, the plurality of candidate characters and the candidate order of these candidate characters become a structural analysis result. If there are no candidate characters, all the characters are treated as candidate characters. The details of the structural analysis will be described later. The data obtained by this structural analysis is sent to the structural analysis result database 130.

The item designation section 103 accepts designation of an item name by an operator during a search. The specified item name is
There may be more than one.

The item data acquisition unit 104 is provided with the item designation unit 1
03, the item data corresponding to the item name received in
A search is made from all the structural analysis results stored in the structural analysis result database 130, and all relevant item data is extracted. When there are a plurality of item names accepted by the item designation unit 103, the same search processing is performed for all of the designated item names.

The item data storage unit 105 has an image memory (not shown) inside. When the item data obtained by the search by the item data acquisition unit 104 is coordinate data, the item data storage unit 105 stores the partial image data corresponding to the coordinates in the image database 1.
20 and temporarily stored in the image memory. For example, in the case of the example of FIG.
In the case where the item name received in “03” is “seal”, coordinate data indicating the position of the seal is obtained by searching the item data acquisition unit 104, and the image data of the seal is temporarily stored in the image memory of the item data storage unit 105. Is stored.

The item data collating unit 106 receives designation of item data by an operator. The operator selects the item data corresponding to the item name received by the item designation unit 103 (that is, the item data acquired by the item data acquisition unit 104 or the item data stored in the image memory of the item data storage unit 105). , One or a plurality of item data are designated. Item data collation unit 106
Extracts the item data (text data and image data) within the range of the same as the designation from the item data acquisition unit 104 and the item data storage unit 105 by performing the collation processing. All of the extracted item data are displayed on the image display unit 140. When there are a plurality of candidate characters for the text data, the character having the first candidate rank is displayed. Details of the matching process will be described later.

The item data post-processing unit 107 performs a process of correcting the item data displayed on the image display unit 140.
If the operator determines that the character displayed on the image display unit 140 is incorrect, the operator changes the candidate order of the character using input means (not shown) or the like and sets the correct character to the first candidate order. I do.

Next, the filing apparatus 10 shown in FIG.
The operation of 0 will be described. FIG. 3 is a flowchart illustrating the operation of filing apparatus 100.

First, when an operator operates the image reading unit 110 to read a document, image data of the document is sent to the image data storage unit 101. Image data storage unit 1
01 receives this image data and performs image storage processing (S301). In this image storing process, first, the received image data is stored in an internal image memory,
Further, the image data is stored in the image database 120 with an ID.

Next, the structure analysis unit 102 executes the following structure analysis processing (S302).

First, the structure analysis unit 102 projects the image data stored in the image data storage unit 101 in the horizontal direction and the vertical direction, and obtains a contour circumscribed rectangle of a black pixel block obtained as a projection result. When the horizontal length and the vertical length of the outline circumscribed rectangle exceed predetermined values, it is determined that the outline circumscribed rectangle is a table rectangle. On the other hand, when these lengths are equal to or smaller than the predetermined value, it is determined that the outline circumscribed rectangle is a character rectangle. For example, in the case of the image data of FIG. 2, the “book”, “vote”, and “A” portions are determined to be character rectangles, and the portion surrounded by the table frame is determined to be a table rectangle.

When the outline circumscribed rectangle is a character rectangle, the character rectangle is cut out and character recognition is performed. In this character recognition, zero or more candidate characters are selected for each character rectangle. These candidate characters and candidate rankings are the result of a structural analysis on the form name. Then, of the candidate characters for each character rectangle, a combination of the candidate characters having the first candidate rank is adopted as the text data of the form name. For example, in FIG. 2, the candidate characters of the character rectangle “book” are the first place “book” and the second place “zhang”, and the character rectangle “vote”
Is the first place “vote” and the second place “necessary”, and the candidate character of the character rectangle “A” is the first place “A” and the second place “△”, the text data of the form name is “form A ". As will be described later, the form name text data “Form A” and 2
The candidate characters “Zhang”, “Necessary”, “$” and the candidate rank of each candidate character below the order are stored in the structure analysis result database 130 as the structure analysis result. If the number of candidate characters is 0, data indicating that is the structure analysis result. For example, if there are no candidate characters for the character rectangle “vote”, “book? A” becomes the text data of the form name (? Indicates an unspecified character). .

If the outline circumscribed rectangle is a table rectangle, the table rectangle is projected in the horizontal and vertical directions. And
If the length of the black pixel block obtained as a projection result is equal to or greater than a predetermined value, the block is determined to be a ruled line, and
An area surrounded by two ruled lines in the horizontal / vertical directions is determined to be a table frame rectangle. Next, a table frame structure is obtained from the vertex coordinates of the table frame rectangle. Then, among the table frame rectangles included in this table frame structure, it is determined that each table frame rectangle in the leftmost column is an item name rectangle, and each table frame rectangle in the other columns is an item data rectangle.

Next, character recognition is executed for each item name rectangle. In this character recognition as well, as in the case of the above-mentioned form name, zero or more candidate characters are selected for each character rectangle, and these candidate characters and the candidate order become a structural analysis result. Then, of the candidate characters for each character rectangle, a combination of the candidate characters having the first candidate rank is adopted as the text data of the item name. In the case of FIG. 2, for example, "address", "name", "age", "sex" and "seal" are the item names. As will be described later, these item names, the candidate characters in the second and lower positions and the candidate rank of each candidate character are stored in the structure analysis result database 130 as the structure analysis result. If the number of candidate characters is 0, data indicating that is the structure analysis result. For example, if there are no candidate characters for the character rectangle “place”,
“House?” Is the item name.

Subsequently, processing of each item data rectangle is executed. In the processing of the item data rectangle, if the item data rectangle is related to characters, the same character recognition as described above is performed, and the combination of the candidate characters having the first candidate rank is set as the item data. On the other hand, if the item data rectangle does not relate to characters, the ID of the document (image data) corresponding to the item data rectangle (see S301) and the vertex coordinate values of the item data rectangle are used as the item data. . For example, if the ID of the corresponding document is AA, the upper left vertex coordinates of the item data rectangle are (X ₁ , Y ₁ ), and the lower right vertex coordinates are (X ₂ , Y ₂ ), the item data is “ID = AA” ,
Coordinates = (X ₁ , Y ₁ , X ₂ , Y ₂ ) ”. In the case of FIG. 2, for example, “XX prefecture XX town XX address”, “Nippon Taro”, “26”, and “male” are item data obtained by character recognition, and the stamped portion is the ID and coordinates. Item data by value. If there are no candidate characters,
The data to that effect is the result of the structural analysis. For example, if there are no candidate characters for the character rectangle “thick”, “Japan?
Is the item data.

The structure obtained by the above structure analysis processing
The analysis results are stored in the structural analysis result database 130.
It is. FIG. 4 is a structural analysis result for the image data of FIG.
Form name, item name and item data
You. In this example, the item name “address” and the item data “
△△ city □ town XX address ”, item name“ name ”and item data
"Taro Nippon", item name "age" and item data "26",
Item name "sex", item data "male" and item name "mark"
Kan ”and item data“ ID = AA, coordinates = (X₁, Y ₁,
X_Two, Y_Two) "Are stored in association with each other.
In addition to these, the structural analysis results include character recognition
Other candidate characters obtained and the candidate order of these candidate characters
Rank is included. These candidate characters and candidate rankings are also applicable.
Is stored in association with the first candidate character.

Next, the item specification section 103 receives the item specification (S303). The operator specifies one or a plurality of item names using input means or the like (not shown). In the following description, a case will be described in which the operator designates “seal” and “name” as item names.

Subsequently, the item data obtaining unit 104 stores the item data corresponding to the item name specified in S303.
Acquired from the structural analysis result database 130 (S30
4). Here, when a plurality of structural analysis results including the designated item name are stored, all item data corresponding to the item name is obtained from all the structural analysis results. For example, a large number of text data items other than “Nippon Taro” may be acquired as item data for the item name “name”. Similarly, a plurality of types of item data (“ID = AA, coordinate =
(X ₁ , Y ₁ , X ₂ , Y ₂ ) ”in some cases).

Next, the item data storage unit 105 stores the S
Of the item data obtained in step 304, item data that is not text (that is, “ID = AA, coordinate =
(X ₁ , Y ₁ , X ₂ , Y ₂ ) ”in the image database 1.
20 and stored in the internal image memory (S305). For example, when there are a plurality of structural analysis results including the item name “seal”, each partial image data is taken in from each document.

Next, the item data collating unit 106 executes an item data collating process (S306).

In this item data collation processing, first, designation of item data is received. The operator selects, from the item data acquired by the item data acquiring unit 104,
Specify one or more. This designation can be performed, for example, by displaying these item data as text data or partial image data on the image display unit 140 and allowing the operator to select them.

When the item data is specified, the item data collating unit 106 performs a search process.

When the designated item data is partial image data (that is, non-character item data), in this search processing, the partial image data is sequentially converted into the partial image data stored in the item data storage unit 105 and the partial image data. The matching is performed, and matching partial image data is searched for. Then, the item data included in the same structural analysis result as the partial image data determined to match and corresponding to the item name “name” is
It is extracted from the item data acquisition unit 104. Each of the partial image data obtained by the collation processing and the item data corresponding to the partial image data are displayed on the image display unit 140 in a format that makes the correspondence clear. FIG. 5 is a conceptual diagram showing an example of this screen display.
In the example shown in FIG. 5, four pieces of partial image data are obtained by the matching process, and the item data (item name “name”) in the same structural analysis result is associated with these partial image data.
) Are displayed.

On the other hand, when the designated item data is text data, in this search processing, the item data is sequentially collated with each item data simultaneously acquired by the item data acquiring unit 104, and within the range of the identity. Find the item data in. In this specification, "within the scope of identity"
This means that the same is applied to the case where the candidate character including the second or subsequent candidate character or the candidate character including 0 candidate character is included. For example, when the designated item data is “Nippon Taro”, for example, the item data in which the candidate ranking of “Large” is first and the candidate ranking of “Fat” is second (for example, this item data is “ "Nippon Dairo") is treated as item data within the range of the identity. Also, "Japan?"
(? Indicates that there are no candidate characters) is also treated as item data within the sameness range. Next, partial image data included in the same image data as the item data determined to be within the sameness range and corresponding to the item name “seal” is extracted from the image memory of the item data storage unit 105. Each item data obtained by the matching process,
The partial image data corresponding to these item data is displayed on the image display unit 140 in a format that makes the correspondence clear.

Next, the item data post-processing unit 107 executes the item data post-processing (S307). The operator looks at the display (see FIG. 5) on the image display unit 140 and searches for an error such as character recognition. In the example of FIG. 5, among the item data corresponding to the item name “name”, only one is “Nippon Taro” and the other is “Nippon Taro”. Therefore, the operator:
It is easy to notice that "large" in "Nihon Dairo" is an error of "thick". The operator corrects "large" of "Nihon Daro" to "thick" using input means (not shown) or the like.
Upon receiving this correction, the item data post-processing unit 107 changes the candidate ranking of “thick” to first and converts the candidate ranking of “large” to, for example, second.

As described above, according to the filing apparatus of this embodiment, item data within the same range can be searched for from a plurality of documents and displayed in comparison. Therefore, the correcting operation of the character error can be efficiently and accurately performed.

Further, according to the filing apparatus according to the present embodiment, search / collation can be performed even when the item data is image data, and in this regard, the correction work can be performed efficiently and accurately. it can.

In the structure analysis processing of this embodiment, ruled lines are determined using the lengths of black pixel clusters obtained by horizontal / vertical projection. For example, horizontal / vertical run lengths are determined. The ruled line may be determined using the number of connected black pixels in. In the structure analysis processing of this embodiment, the area surrounded by the ruled lines is determined to be a table frame rectangle. However, the table frame rectangle is determined using document-specific format information obtained in advance. Is also good.

Further, in this embodiment, after the determination of the ruled line and the determination of the table frame rectangle as described above,
Although the structural analysis is performed on the premise that the item data is located to the right of the item name, the structural analysis may be performed using the layout rule of the document structure as a document model (for example, Japanese Patent Application No. 7-23).
No. 2285).

In addition, in this embodiment, the item data matching section is made to perform the matching of the partial image data and the matching of the text data. However, it is clear whether the item data is the partial image data or the text data. When used in an environment, the filing apparatus may be configured to match only one of the partial image data and the text data.

[0048]

As described in detail above, according to the present invention, matching item data can be searched for from a plurality of documents and displayed in a contrasting manner, and a search using image data can be performed. Correction of character errors can be performed efficiently and accurately.

[Brief description of the drawings]

FIG. 1 is a block diagram schematically showing an overall configuration of a filing apparatus according to an embodiment.

FIG. 2 is a conceptual diagram showing an example of a document to be read into a filing device according to the embodiment.

FIG. 3 is a flowchart for explaining the operation of the filing apparatus according to the embodiment.

FIG. 4 is a conceptual diagram for explaining the operation of the filing apparatus according to the embodiment.

FIG. 5 is a conceptual diagram showing a screen display example of the filing apparatus according to the embodiment.

[Explanation of symbols]

 Reference Signs List 100 filing device 101 image data storage unit 102 structure analysis unit 103 item specification unit 104 item data acquisition unit 105 item data storage unit 106 item data collation unit 107 item data post-processing unit 110 image reading unit 120 image database 130 structure analysis result database 140 Image display

Claims (6)

[Claims] 1. An image data storage means for fetching and temporarily storing document image data from the outside, and storing the document image data in an external image data storage means; Extracting a layout from the document image data, determining an item name entry field and an item data entry field of the layout, extracting a character rectangle from the item name entry field and the item data entry field,
And a structure analysis unit that performs character recognition of these character rectangles and stores the item names and the item data obtained by the character recognition in an external structure analysis result storage unit. Item designation means for designating one or more kinds; item data acquisition means for acquiring the item data corresponding to the item name designated by the item designation means from the structural analysis result accumulation means; Causing the operator to specify one or more types of the item data obtained by the item data obtaining unit;
And item data collating means for extracting the item data within the specified range of the identity from the item data acquired by the item data acquiring means and displaying the extracted item data on a screen. Filing device. 2. When the description of the item data is not a character rectangle, the structure analysis means generates coordinate data for specifying the position of the description and stores the coordinate data in the structure analysis result storage means. The filing apparatus according to claim 1, wherein 3. When the description of the item data is not a character rectangle, the item data matching means displays an image extracted from the image data storage means using the coordinate data on the screen. The filing apparatus according to claim 2, wherein 4. As a result of character recognition by said structure analysis means, when there are a plurality of candidate characters for the same character rectangle,
4. A method according to claim 1, wherein said plurality of candidate characters and candidate ranks are stored in said structural analysis result storage means.
The filing apparatus according to any one of the above. 5. The filing apparatus according to claim 4, wherein the collating means displays the character having the first candidate rank on the screen as the item data. 6. The filing apparatus according to claim 5, further comprising item data post-processing means for changing the candidate order of the candidate character.