JP2001052112A - Recognizing processing method, information processor and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide highly accurate layout analytic processing by identifying an area in an original image, identifying the area in the original image again corresponding to information provided from an object to be recognized for each area defined by identification and performing recognizing processing for each area identified again. SOLUTION: A business card is read by a scanner and original image data are generated (S1). Character segmenting processing of the original image is performed. As layout analytic processing included in this character segmenting processing, area identifying processing of the original image data is performed (S2). It is discriminated whether each area defined by this area identifying processing is matched to prescribed merge conditions or not (S3). After discriminating processing is performed for each area, areas matched to the conditions are merged (S4). After merging processing, concerning each area defined again, character recognizing processing of the object to be recognized is executed (S5). As information corresponding to the respective items of a data base, the recognized result provided by character recognizing processing is registered (S6).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technology relating to recognition processing of a character recognition device or the like.

2. Description of the Related Art In recent years, as a device has become inexpensive and the processing speed has been improved, a special character recognition device (OCR: Optica) has been developed.
Character recognition processing using a personal computer (software that operates on the personal computer), a word processor, or the like, instead of a character reader, is becoming widespread.

For example, a function is realized in which description information such as a postcard or a business card is read by a scanner to perform a character recognition process, and the recognition result is automatically registered as name information or address information as each item of a database. ing.

As a result, it is possible to omit an input operation such as a character input by a user using a keyboard, and to reduce an operation burden on the user.

As described above, in order to automatically register a recognition result in a database without intervention of a user's operation, there is a demand for higher accuracy of a recognition process.

[0006]

2. Description of the Related Art In conventional character recognition processing, first,
A layout analysis process is performed on original image data of a document (business card) obtained by scanning with a scanner.

The layout analysis process is a process performed for the purpose of improving the accuracy of the character segmentation process at the time of the character recognition process, and is a process of setting an area of a block of information to be recognized which is assumed to be the same item. .

For example, assuming that the original image data is a binary image, this layout analysis process determines the layout structure of the document information from the connection state of black pixels and the interval between the clusters of black pixels. This defines the divided area.

After the layout analysis processing, for each defined area, a recognition target (character candidate) is cut out based on the range of black pixels in the area, its size, and interval. Then, a character recognition process using a dictionary is performed.

The character string of each area obtained in this manner is determined from the positional relationship to which item of the database, and registered as information of each item of the database.

[0011]

As described above, in the conventional character recognition processing, the layout analysis processing is performed to improve the accuracy of the character cutout processing.

In this layout analysis processing, the accuracy of character extraction processing is improved for the position of a recognition target such as a form and the size of a character to be recognized is predetermined, but the processing of a character such as a business card or a postcard is improved. For a free format or free layout document whose size and position are not determined, a phenomenon occurs in which a recognition target to be defined as one region is divided into another region.

This is because the layout analysis process defines an area mainly based on the proximity of the recognition target (character candidate).

Therefore, in the character recognition processing of a document in a free format or a free layout, the accuracy of the character extraction processing by the layout analysis processing is reduced, which causes the accuracy of the recognition rate to be reduced.

Since the recognition result of such erroneous recognition is automatically registered in the database as information, the user corrects the erroneous information in the database, and the user's operation burden increases. .

The present invention has been made in view of the above problems, and has as its object to provide a technique for realizing a layout analysis process with high accuracy. It is another object of the present invention to provide a high-accuracy character recognition processing technology by the high-accuracy layout analysis processing.

It is another object of the present invention to provide a technique for reducing a user's operation burden in automatically registering information in a database using a character recognition process.

[0018]

SUMMARY OF THE INVENTION According to the present invention, there is provided a layout for identifying a region in an original image and re-identifying the region in the original image in accordance with information obtained from a recognition target for each region defined by the identification. Analysis processing technology.

This is a recognition processing technique for performing recognition processing for each area defined by re-identification of the area by the layout analysis processing.

Further, the re-identification process may be a process of integrating regions according to information obtained from the recognition target.

Still another aspect of the present invention is a technique for identifying a database item corresponding to a recognition result obtained by the above-described recognition processing technique of the present invention, and registering the recognition result as information of a corresponding database item.

The object is solved by a computer readable recording medium on which a method, an apparatus, and a program for realizing the technique of the present invention are recorded.

[0023]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention will be described in detail.

The following is an embodiment of the present invention for performing business card recognition processing and automatically registering the recognition processing result in a database.

FIG. 1 is a diagram showing a schematic processing flowchart of this embodiment.

First, a business card is read by a scanner to generate original image data. (Step S1) Next, character cutout processing of the original image is performed. The layout analysis process included in the character segmentation process is the main process of the present invention.

As a layout analysis process, first, a region identification process of the original image data is performed. (Step S2) It is determined whether or not each area defined by the area identification processing meets a predetermined integration condition. (Step S
3) After performing the above-described determination processing for each area, areas matching the conditions are integrated. (Step S4) After the integration process, a character recognition process of a recognition target is executed for each of the redefined regions. (Step S5) The recognition result obtained by the character recognition processing is registered as information corresponding to each item of the database. (Step S6) Next, the configuration of a general computer that executes the processing of the present invention will be described.

FIG. 2 is a diagram showing a schematic configuration of the computer.

The computer 1 mainly includes the CPU 2 and the RA
M3, HDD (hard disk drive) 4, CD-R
It comprises an OM drive 5, an FDD (floppy disk drive) 6, an NCU (network control unit) 7, a display 8, a keyboard 9, and a scanner 10. These components are connected by a bus.

In the figure, a CPU 2 executes various processes and controls by programs, and is also used in various processes and controls by the program of the present invention.

The RAM 3 is used to hold temporary data for various processes and controls by the CPU 2 and to hold data to be displayed on the display 8.

The hard disk drive 4 and the floppy disk drive 6 are devices for recording programs and control data on a non-volatile recording medium (hard disk and floppy disk 12) and for reading out programs and control data from the recording medium. .

The CD-ROM drive 5 is a CD-ROM 1
1 is a device for reading out a program and control data recorded in the device 1.

The network control unit 7 is connected to a network and exchanges data with other devices via the network, and can be used for downloading programs and data from a remote server. This network control unit 7 may be a modem or a LAN card.

The display 8 displays menus, processing results, and the like.

The keyboard 9 has character keys and cursor movement keys, and is used for character input and cursor movement.

The scanner 10 is for reading a document (such as a business card or postcard).

The program of the present invention is stored in the hard disk drive 4, and the program is read from the hard disk drive 4 in response to an execution instruction and read out from the hard disk drive 4.
It is loaded into M3 and controlled by the CPU 2 to cause the computer 1 to execute the processing of the present invention.

The program of the present invention is stored in a portable recording medium such as a CD-ROM 11 or a floppy disk 1.
2 is read using a drive (CD-ROM drive 5 or floppy disk drive 6) corresponding to a recording medium and stored in the hard disk drive 4.

Instead of using a portable recording medium, the program of the present invention may be downloaded via a network using the network control unit 7 and stored in the hard disk drive 4.

Furthermore, FIG. 2 and the above description show an example in which the program of the present invention is stored in the hard disk 4, but the program of the present invention is stored in the ROM,
The program may cause the computer 1 to execute the processing of the present invention.

Further, detailed processing of this embodiment will be described.

FIGS. 3 to 5 are flowcharts showing the detailed processing of this embodiment.

First, the user selects an item "register from business card to address book" from a menu (not shown) displayed on the screen, and instructs to start a business card recognition process. Also, when the instruction to start the recognition processing is given, the character direction of the business card (vertical writing /
Horizontal writing). (Step S11) In this example, it is assumed that horizontal writing is designated as the character direction of the business card.

After the instruction to start the recognition process in step S11, the process of reading the business card set in the scanner is performed to generate original image data. In this example, the original image data is a binary image. (Step S12) A circumscribed rectangle is set for each connected region of black pixels in the original image data generated in step S12. (Step S13) For each of the defined rectangles, scanning is performed from one side of the original image data to the other side in the direction in which the characters of the business card are written, and a unique serial number is set in the order closer to the scanning start side. In other words, when horizontal writing is specified as the character direction of the business card, scanning is performed from the left side to the right side of the business card, and when vertical writing is specified, scanning is performed from the upper side to the lower side of the business card and the serial number is assigned to each area. Is set.

After the circumscribed rectangle is defined in step S13, the adjacent relationship between the defined rectangles is determined, and a pointer is set between the adjacent rectangles. That is, each rectangle is set as a target rectangle, and pointers to adjacent rectangles at the top, bottom, left, and right of the target rectangle are obtained. (Step S
14) After setting the pointer in step S14, it is determined whether or not the pointer to the adjacent rectangle associated with each rectangle satisfies the release condition. (Step S15) When the horizontal writing is specified as the character direction of the business card, the cancellation condition is that the distance between the rectangle and the adjacent rectangle is larger than a predetermined multiple of the height of the most frequent rectangle, or The height value of the rectangle is large. When vertical writing is specified as the character direction of the business card, the height of the rectangle of the above-described cancellation condition is set to the width of the rectangle.

As a result of the determination in step S15, the pointer that matches the release condition is released. (Step S16) After the pointer release processing in step S16, the same numbers are assigned to rectangles that are adjacent to each other in the character description direction. In this processing, a rectangle serving as a starting point is obtained in the scanning direction when the serial number is given, and the starting point rectangle is set in each of the rectangles whose adjacency is set by the pointer in the character description direction with the rectangle as the starting point. The numbers are sequentially propagated.
(Step S17) After the process of step S17, a rectangle including the rectangle assigned the same number is set, and the rectangle is defined as a character string rectangle. (Step S18) Next, character recognition processing is performed on each character string rectangle defined in step S18. (Step S19) In this character recognition process, the recognition target is divided based on the connection state of black pixels, the intervals between black pixel regions, and the like, and the characteristics of the characters stored in the dictionary and the characteristics of the recognition target are compared. The character having the closest distance is output as a recognition result for the recognition target. Since this character recognition processing is a well-known technique, its details are omitted.

By this character recognition processing, a character string line can be recognized within each character string rectangle. In this example, the case where the character direction of the business card is horizontal writing is described, so the line is recognized. However, in the case of vertical writing, the column of the character string is recognized.

Next, the center coordinates of each character string rectangle defined in step S18 are calculated. When only one line of the character string is recognized in the character string rectangle, the center axis coordinate of the line is calculated from the position coordinate information of the line of the character string and defined as the center coordinate of the line. Further, the height of the area including the row is calculated, and the height of the area is defined as the height of the row.
(Step S20) When calculating the center coordinates and the height of this line, if a plurality of lines are recognized in the character string rectangle, the center coordinates and the bottom position of the top line existing in the character string rectangle are determined. Is defined as the center coordinate of the row, and the height of an area including the plurality of rows is defined as the row height.

In this example, the center coordinates and height of a row are calculated because horizontal writing is specified. However, when vertical writing is specified, the center coordinates and width of a column are calculated and a character string rectangle is set. When a plurality of columns are recognized, the center coordinates of the rightmost column and the center between the leftmost columns are set as the column center coordinates.

The center coordinates of the rows obtained in step S20 are sorted in ascending order (step S21), and it is determined whether or not each row satisfies the integration condition. (Step S2
2) This integration condition targets a line of interest and a line having the next largest central coordinate value after the central coordinate of the line of interest, and the central coordinates of those lines are close to or below a predetermined threshold value and are close to each other. In this case, those having a difference in height between the rows equal to or less than a predetermined value are subjected to the integration processing.

In step S22, lines matching the integration conditions are integrated, and a character string rectangle including the lines is defined.

By performing the processing of steps S22 and S23 on each line, a character string rectangle to be finally recognized is obtained.

The character recognition process is performed on each of the character string rectangles thus obtained. (Step S2
4) This character recognition process is also a well-known process similar to the above, and a description thereof will be omitted.

When the character recognition result (line) of each character string rectangle is obtained in step S24, knowledge processing is performed on which item of the database corresponds to each line. (Step S2
5) In order to perform the knowledge processing in step S25,
Various kinds of information relating to place names, personal names (first and last names), company relations, kana-kanji conversion, and description patterns are provided in advance, and the following determination is made using these various kinds of information.

If there is a postal mark in the line to be processed, it is determined that the description following the postal mark corresponds to the postal code item. Alternatively, it is determined that a character string corresponding to a predetermined description pattern representing a postal code (for example, a three-digit number + a hyphen + a four-digit number) also corresponds to the postal code item.

If the line to be processed includes a character string relating to a place name, it is determined that the character string corresponds to the address item.

If there is a telephone number mark in the description of the line to be processed, it is determined that the description following the mark corresponds to the item of telephone number. Alternatively, it is determined that a character string corresponding to a predetermined description pattern representing a telephone number (for example, “TEL” or three digits + hyphen + four digits + hyphen + four digits) is also a telephone number item. These determinations are the same for the FAX item.

In the description of the line to be processed, a character string relating to a job title (title) or a predetermined description pattern representing a company relationship (for example, “company”, “—section”, “—section”, “manager”, If there is a character string including “section manager”, the character string including the character string is determined to correspond to the item of work place, affiliation, and post, and is classified into each item by using the various information prepared in advance.

Similarly, a character string corresponding to each item of the database is obtained using the above various information, and finally, a process of identifying a line corresponding to the item of the name of the database is performed.
The determination as to whether or not the line corresponds to the item of the name is performed in the following order, and the identification processing ends when the determination is satisfied.

[0061] 1. 1. Line where only one remains at the time of name determination processing 2. A line containing a character that matches the surname and first name information provided in the personal name dictionary and matches both information. 3. Lines whose character size is larger than other lines 4. A line in which the space between each character included in a line is larger than the space between characters included in other lines. 5. Lines with title (title) 7. Line next to line determined to be company name 7, Line next to line determined to be address The line following the line determined to be the telephone number The character string obtained by the knowledge processing in step S25 is registered as information on each item of the corresponding database.
(Step S26) Next, the data structure of the database will be described. In the present embodiment, the data structure of the card database is exemplified.

FIG. 6 is a diagram showing the data structure of the database. FIG. 6A shows the format of card data, and FIG. 6B shows the format of character data.

As shown in FIG. 6A, one card data in the present embodiment is composed of the number of items, the item offset, and the character data.

The number of items indicates the number of types of items provided in the database. In this example, since there are 20 items from item 1 to item 20, the number of items “20” is recorded.

Although not described in detail, for example, the items of this database include a name, a work postal code, a work address, a work telephone number, a facsimile, and an affiliation + post.

The item offset records the offset value from the head of the card to the character data corresponding to each item in the character data area.

The character data area is a area in which character data corresponding to each item is continuously recorded, and the recording area is changed according to the number of characters of character data for the item and the presence or absence of character data. This is the data area.

In this character data area, character data is recorded in the format shown in FIG.

FIG. 6B is a diagram showing a character data format for one item.

This character data is composed of a character number area and a real data area. The character number area records the number of characters of the real data, and the real data area records real data (characters).

FIG. 7 is a diagram showing an example of area (line) identification obtained by the character recognition processing.

When the present invention is not applied, as shown in FIG. 7A, the space between "Fuji" and "Gaku" is wide, so that another area (line) is set. A recognition process is performed.

Since "Fuji" is composed of two characters, there is a high possibility that "Fuji" and "shi" will be correctly recognized from the character size and character spacing.

However, since "gaku" is one character, there is no reference in the recognition process, and one character is determined to be a plurality of characters, and there is a high possibility that an incorrect recognition result will be obtained.

Moreover, since the part recognized as one item as the item "name" in the database is divided into two parts, the user must correct the information on the automatically registered name.

On the other hand, when the present invention is applied, as shown in FIG. 13B, one area of "Fuji" and "Gaku" are integrated by the integration processing of rows (areas).

Therefore, the object to be recognized corresponding to “Gaku” is character-recognized with reference to the object to be recognized corresponding to “Fuji”, so that the possibility of being correctly recognized increases. Also,
Since it is recognized as one line, it is automatically registered as information of the item “name” in the database. Therefore, no correction work by the user is required.

The embodiments of the present invention have been described above.
The present invention is not limited to this, and various modifications can be made without changing the gist of the present invention.

The following layout analysis processing described as the character recognition processing of the present invention is also the present invention. (Invention 1) A layout analysis processing method in recognition processing, in which an area in an original image is identified, and an area in the original image is re-identified according to information obtained from a recognition target for each area defined by the identification. Layout analysis processing method to be performed. (Invention 2) In the layout analysis processing method according to Invention 1, the re-identification processing is an integration processing of regions according to information obtained from the recognition target. (Invention 3) An information processing apparatus for recognizing information in an original image, comprising: an identification unit for identifying an area in the original image;
An information processing apparatus comprising: a re-identification unit that re-identifies an area in the original image in accordance with information obtained from a recognition target for each area defined by the identification. (Invention 4) In the information processing apparatus according to Invention 3, the re-identification unit re-identifies an area in the original image by integrating areas according to information obtained from the recognition target. (Invention 5) A computer executes a procedure for identifying an area in an original image and a procedure for re-identifying an area in the original image according to information obtained from a recognition target for each area defined by the identification. A computer-readable recording medium on which a program for causing a computer to record is recorded. (Invention 6) A computer-readable recording medium on which the program of Invention 5 is recorded, wherein the procedure for re-identification is integration of areas according to information obtained from the recognition target.

[0080]

As described above, according to the present invention, even if an object is erroneously divided as another object to be recognized (area), it is set as a correct object to be recognized by integrating the areas by re-identification. be able to.

Therefore, the character recognition rate can be improved, and even when the character recognition result is automatically registered in the database, the correction work by the user is unnecessary, and the operation burden on the user can be reduced. It becomes possible.

[Brief description of the drawings]

FIG. 1 is a diagram showing a schematic processing flowchart of an embodiment.

FIG. 2 is a diagram illustrating a schematic configuration of a computer.

FIG. 3 is a diagram (part 1) illustrating a detailed processing flowchart of the embodiment;

FIG. 4 is a diagram (part 2) illustrating a detailed processing flowchart of the embodiment;

FIG. 5 is a diagram (part 3) illustrating a detailed processing flowchart of the embodiment;

FIG. 6 is a diagram showing a data structure of a database.

FIG. 7 is a diagram illustrating an example of area (line) identification obtained by a character recognition process.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Computer 2 CPU 3 RAM 4 Hard disk drive 5 CD-ROM drive 6 Floppy disk drive 7 Network control unit 8 Display 9 Keyboard 10 Scanner 11 CD-ROM 12 Floppy disk

Claims (8)

[Claims] 1. A recognition processing method for recognizing information in an original image, comprising: identifying an area in the original image; and determining a region in the original image according to information obtained from a recognition target for each area defined by the identification. A recognition processing method for re-identifying an area and performing recognition processing of a recognition target for each of the re-identified areas. 2. The recognition processing method according to claim 1, wherein the re-identification processing is processing for integrating regions according to information obtained from the recognition target. 3. An information processing apparatus for recognizing information in an original image, comprising: identifying an area in the original image; and determining an area in the original image according to information obtained from a recognition target for each area defined by the identification. An information processing apparatus comprising: a layout analysis unit that re-identifies an area; and a recognition unit that performs recognition processing of a recognition target for each of the re-identified areas. 4. The information processing apparatus according to claim 3, wherein the layout analysis unit re-identifies an area in the original image by performing area integration processing according to information obtained from the recognition target. 5. An apparatus according to claim 3, further comprising: an item identification unit that identifies an item corresponding to a recognition result obtained from the area; and a registration unit that registers the recognition result in a database as information on the corresponding item. Information processing device. 6. A computer, comprising: a step of identifying an area in the original image; a step of re-identifying an area in the original image according to information obtained from a recognition target for each area defined by the identification; A computer-readable recording medium that records a procedure for performing recognition processing of a recognition target for each re-identified region, and a program for executing the procedure. 7. The computer-readable recording medium according to claim 6, wherein the procedure for re-identification is integration of areas according to information obtained from the recognition target. 8. The method according to claim 6, further comprising: a step of identifying an item corresponding to a recognition result obtained from the area, and a step of registering the recognition result in a database as information of the corresponding item. A computer-readable recording medium recording the above program.