JP5054472B2 - Character recognition device and character recognition method in character recognition device - Google Patents

Description

  The present invention relates to a character recognition apparatus that recognizes characters by reading an image on the surface of a form, and more particularly to a character recognition apparatus that can read both a dropout color image and a non-dropout color image on a form. It is.

  Conventionally, a character recognition device has been used as a device for recognizing characters, symbols, and the like described in various forms such as postcards, slips, questionnaires, and test answer sheets. In general, a form that is read by a character recognition device is a portion that is printed in a dropout color that is visible to the human eye but cannot be read by the character recognition device, and a non-drop that is a color that can be read by the character recognition device. Parts printed in out color.

  The conventional character recognition apparatus conveys the form one by one, reads the image on the surface of the form one by one while conveying it, and is in a position determined from the edge of the form among characters, symbols, etc. in the read image. Images such as characters and symbols are cut out, and the cut out images are recognized as recognition targets.

  However, if a transport error such as a transport delay occurs during the reading of an image while transporting a form, it recognizes that it should be at the position determined from the end of the form by expanding the read image. The target character or the like is shifted and exists in another place. For this reason, there is a problem that even if the image at the position determined from the edge of the form is cut out, the character image to be recognized cannot be cut out correctly, so that the character to be recognized cannot be recognized well.

  “Character recognition device and character recognition method” (Patent Document 1) has been proposed as a character recognition device that corrects a conveyance error even if a conveyance error occurs during image reading and can correctly cut out a character image. .

  The character recognition device described in Patent Literature 1 is based on a form image having a plurality of character entry frames and preprint information corresponding to each of these character frames and provided adjacent to each other. It is characterized by cutting out and recognizing characters. Specifically, the character recognition device corrects a character frame position using preprint information, cuts out a character image based on the character frame whose position has been corrected, and extracts a character image from the cut out character image. Recognize.

As a result, even if a transport error occurs during image reading, the character image can be cut out correctly and the character can be recognized correctly.
JP 2002-109469 A

  In the conventional character recognition device, when a conveyance error occurs during reading of an image, the read image is expanded and the position of the character to be recognized is shifted. Therefore, the character image cannot be correctly cut out. Even if the character image was recognized, the characters could not be recognized correctly.

  In addition, even if there is a slippage deviation or printing slippage on the form, the character position on the form surface is misaligned, so the character image cannot be cut out correctly. Therefore, even if the cutout character image is recognized, the character is accurately I couldn't recognize it.

  Furthermore, the character recognition apparatus described in Patent Document 1 needs to print preprint information for correcting a conveyance error on a form in advance, and cannot use a conventional form. For this reason, a new form is required and the cost is high.

  Furthermore, the character recognition device described in Patent Document 1 has a problem that the processing speed becomes slow because the position correction processing is performed based on the preprint information even if no transport error occurs.

  The present invention has been made in view of such circumstances, and provides a character recognition device capable of correctly recognizing a character even when a character misalignment occurs and does not slow down the processing speed, and a character recognition method in the character recognition device. It is something to be offered.

The problems of the present invention can be solved by the following inventions.
That is, the character recognition device and the character recognition method of the present invention are character recognition devices for recognizing characters written in the character frame of a form having a character frame printed in dropout color,
All image reading means capable of reading both images of a dropout color and a non-dropout color on the form, and a character image capable of reading only a non-dropout color image on the form A reading means; a form edge detecting means for detecting the position of the edge of the form; and acquiring information relating to the edge position of the form; and the character frame position or the character position on the form based on the position of the edge of the form Format parameter storage means for storing format parameter designation positions, character cutout means for cutting out a character image from the dropout image read by the character image reading means, and recognition of characters from the character image cut out by the character cutout means Character recognition means for
The character cutout means is
A first step of calculating a character arrangement position and a character cutout reference position on the dropout image using information on the form edge detected by the form edge detection means and the format parameter designation position stored in the format parameter storage means. When,
A second step of comparing the character cutout reference position calculated in the first step with the arrangement position and determining whether or not there is a positional deviation between the reference position and the arrangement position;
A third step of cutting out characters based on the character cutout reference position when it is determined in the second step that there is no misalignment;
When it is determined that there is a displacement in the second step, the character frame position in the reading area is detected based on the character arrangement position calculated in the first step from the non-dropout image read by the whole image reading means. A fourth step;
Run a fifth step for cutting out a character from the dropout image based on the character frame position detected by said fourth step,
A main feature is that the character recognition means executes a sixth step of recognizing the character cut out in the third step or the fifth step.

Thus, if the positional deviation in the character described in the document does not occur, since not corrected character extraction reference position, speed character recognition processing is possible, if the positional displacement has occurred By cutting out the character image based on the character frame recognized from the non-dropout image, the character can be correctly recognized even if the character is misaligned.

  Further, according to the present invention, the character cutout unit passes a vertical center line on the rectangular screen and the character cutout reference position when the character image on the dropout image is surrounded by the smallest rectangle. The positional deviation amount is calculated by calculating the positional deviation amount with respect to a straight line parallel to the center line.

  Still further, in the present invention, the character cutout unit calculates the positional deviation amount by calculating a center of gravity of the character image on the dropout image and calculating a deviation amount between the center of gravity and the character cutout reference position. Do.

As described above, according to the character recognition device of the present invention, when a character is cut out even when the character position on the form is recognized to be misaligned due to conveyance error, form cut deviation, form print deviation, etc. Cuts out from the dropout image, so that the character frame printed in the dropout color is not cut out, and it is possible to prevent high-precision character recognition by preventing the cutting out of images other than the characters that hinder character recognition. Can do.

  Furthermore, when there is no positional deviation in the characters described in the form, the character extraction reference position is not corrected, so that high-speed character recognition processing can be performed.

The best mode for carrying out the character recognition device of the present invention will be described below in detail with reference to the accompanying drawings.
<Configuration>
A configuration of a character recognition device according to an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a configuration diagram of a character recognition apparatus according to an embodiment of the present invention.
As shown in FIG. 1, the character recognition apparatus of the present invention comprises an image reading means 1, a form edge detection means 2, a character cutout means 3, a format parameter storage means 4, and a character recognition means 5. Yes.

The image reading means 1 comprises an all image reading means 6 and a character image reading means 7.
The all image reading means 6 reads the character frame printed in the dropout color described on the form surface and the character printed in the non-dropout color by the image sensor. As the image sensor, a CCD image sensor, a CMOS image sensor, or the like can be used.

  The dropout color refers to a color that cannot be read when a color original is read by a monochrome image reading means, and usually the color of the light source used in the image reading means corresponds to the dropout color. If the image reading means can read a color image, any color of RGB can be set as a dropout color.

  For example, when red is set as the dropout color, the contrast difference between R of RGB, that is, red (Red), and the read portion is read as an image. As a result, the portion printed in red, which is the dropout color, has no difference in contrast with the red color, so that it cannot be read as an image, and the portion is read so that nothing is printed.

  That is, the all-image reading means 6 can also read a character frame printed in the dropout color by reading colors other than those set as the dropout color among RGB. Hereinafter, the image read by the whole image reading means 6 is referred to as a non-dropout image.

  The character image reading means 7 reads characters and the like written in the non-dropout color described on the surface of the form by the image sensor, but cannot read the character frame and the like printed in the dropout color. Hereinafter, the image read by the character image reading means 7 is referred to as a dropout image.

  The form edge detection means 2 detects the form edge based on the image read by the image reading means 1. Further, the form edge detecting means can recognize the form edge not by image recognition but also by an optical sensor or a mechanical sensor.

  The character cutout means 3 is based on the position information of the form edge detected by the form edge detection means 2 and the character position information with reference to the form edge stored in the format parameter storage means 4. The position described is calculated, and the character image at that position is cut out.

  The character cutout means 3 reads not only the position information of the form edge detected by the form edge detection means 2 and the character position information based on the form edge stored in the format parameter storage means 4, but also reads all images. Based on the image of the character frame read by the means 6, the position where the character is described can be calculated, and the character image at that position can be cut out.

  As a result, even if the form image is expanded and read due to an error in carrying the form when reading the form image, the character frame image information is used to correct the misalignment of the character description position. A character image can be cut out.

  As described above, the format parameter storage unit 4 stores character position information with reference to the edge of the form, character frame position information, and the like. The character recognition unit 5 recognizes characters based on the character image cut out by the character cutout unit 3.

<Operation when correcting all characters>
Next, the operation at the time of all character correction of the character recognition device according to the embodiment of the present invention will be described with reference to the drawings. FIG. 2A is a diagram showing an example in which a form is read by all image reading means 6. FIG. 2B is a diagram showing an example in which a form is read by the character image reading means 7.

  First, a description will be given with reference to FIG. The all image reading means 6 reads the image on the form surface. As shown in FIG. 2A, a character frame 22 and a character 23 are read. Here, on the form, the character frame 22 is printed in a dropout color, and the character 23 is described in a non-dropout color.

  The form edge detecting means 2 detects the edge of the form by recognizing the edge of the form from the read image on the surface of the form. Further, the form edge detecting means can recognize the form edge not by image recognition but also by an optical sensor or a mechanical sensor.

  The character cutout means 3 uses the information related to the form edge detected by the form edge detection means 2 and the information related to the character position stored with reference to the form edge stored in the format parameter storage means 4 to perform non-dropout. The position where the character is arranged on the image is calculated. Next, the character cutout unit 3 recognizes the character frame 22 in the reading area 21 which is an area in a predetermined range based on the calculated character arrangement position, and based on the recognized character frame 22. The character cutout reference position 24 is determined. By recognizing the character frame 22 in the reading area 21, which is an area of a predetermined range, even if there are a plurality of character frames on the form, a necessary character frame can be selected without being erroneously recognized. Can be recognized.

Further, the character cutout means 3 is based on the information on the form edge position detected by the form edge detection means 2 and the information on the character frame position on the basis of the form edge stored in the format parameter storage means 4. It is also possible to recognize the character frame 22 within the reading area, which is an area within a predetermined range based on the character frame position, and to determine the character cutout reference position 24 based on the recognized character frame 22. By recognizing the character frame 22 in the reading area, which is an area of a predetermined range, even if there are a plurality of character frames on the form, a necessary character frame can be selected without erroneous recognition. It becomes possible to recognize.
Further, the character cutout unit 3 can also recognize the character frame 22 from the entire non-dropout image and determine the character cutout reference position 24 based on the recognized character frame.

  Next, description will be made with reference to FIG. The character cutout unit 3 cuts out characters based on the determined character cutout reference position 24, and the character recognition unit 5 recognizes characters from the cut out image. At this time, the character is cut out by the dropout image. As a result, since the character frame is not shown in the dropout image, the character frame is not shown in the clipped area, and the character frame does not become an obstacle to recognition in character recognition. Accurate character recognition is possible.

<Operation when correcting deviation by retry processing>
Next, the operation in the case of correcting the deviation in the retry process of the character recognition device according to the embodiment of the present invention will be described with reference to the drawings. FIG. 3A shows an example in which a form is read by the character image reading means 7. FIG. 3B is a diagram showing a deviation between the calculated character cutout reference position and the actual character position. FIG. 3C is a diagram showing the calculation character frame positions in the image read by the all-image reading means 6. FIG. 3D is a diagram in which characters are cut out from the image read by the character image reading means 7.

  First, a description will be given with reference to FIG. The all image reading means 6 and the character image reading means 7 read the image on the form surface. The image read by the all-image reading means 6 is a non-dropout image, and both character frames and characters are read. In the image read by the character image reading means 7, as shown in this figure, the character frame described in the dropout color is not read, but only the characters are read.

  The form edge detecting means 2 detects the edge of the form by recognizing the edge of the form from the read image on the surface of the form. Further, the form edge detecting means can recognize the form edge not by image recognition but also by an optical sensor or a mechanical sensor.

  The character cutout means 3 uses information on the form edge detected by the form edge detection means 2 and information on the character frame position 33 or the character position based on the form edge stored in the format parameter storage means 4. The character arrangement position on the dropout image and the character cutout reference position 31 are calculated.

  Next, a description will be given with reference to FIG. The character cutout means 3 checks the positional deviation on the dropout screen between the calculated character cutout reference position 31 and the character 23 on the dropout screen. The positional deviation can be determined by determining the amount of positional deviation between the character cutout reference position 31 and the character 23.

  For example, it can be obtained as follows. That is, when the character 23 is surrounded by the smallest rectangle, the vertical and horizontal center lines on the plane of the rectangle are obtained. A positional deviation amount between the vertical center line of the rectangle and a straight line passing through the character cutout reference position 31 and parallel to the vertical center line is obtained. Similarly, an amount of positional deviation between the horizontal center line of the rectangle and a straight line passing through the character cutting reference position 31 and parallel to the horizontal center line is obtained. In this way, the positional deviation amount can be obtained.

  Further, the amount of displacement can be obtained by calculating the position of the center of gravity of the character 23 and using the position of the center of gravity. Specifically, the center of gravity of the character 23 is obtained, and a horizontal line passing through the center of gravity on the image plane and a straight line passing through the center of gravity and perpendicular to the horizontal line (hereinafter referred to as a vertical line) are obtained. A positional deviation amount between the horizontal line and the straight line parallel to the horizontal line through the character cutout reference position 31 is obtained. Similarly, the amount of positional deviation between the straight line passing through the character cutout reference position 31 and parallel to the vertical line and the vertical line is obtained. In this way, the positional deviation amount can be obtained.

  Furthermore, a deviation amount between the position of the center of gravity and the character cutout reference position 31 can be obtained and used as a positional deviation amount.

  When the obtained positional deviation amount is within a predetermined range, the character cutout unit 3 determines that there is no character positional deviation, and based on the character cutout reference position 31, an image of the character 23 from the dropout image. Cut out. Further, the image cut-out of the character 23 can be cut out based on the information on the character position stored in the format parameter storage means 4. The character 23 is recognized by the character recognition means 5.

  As a result, when there is no character misalignment, the character recognition processing speed can be increased because the misalignment correction is not performed. Further, since characters are cut out from the dropout image, the character frame is not cut out, and the cut image does not include a character frame that becomes an obstacle to recognition, so that highly accurate recognition is possible. If any one of the two obtained misregistration amounts exceeds the predetermined range, the character cutout unit 3 determines that there is a misalignment of the characters.

Next, description will be made with reference to FIG. When it is determined that there is a character misalignment, the character cutout unit 3 performs misalignment correction as follows. In the non-dropout image read by the whole image reading means 6, the character cutout means 3 applies the information related to the character arrangement position calculated previously, and is a predetermined area based on the character arrangement position. A character frame 22 is recognized in a certain reading area 21. By recognizing the character frame 22 in the reading area 21, which is an area of a predetermined range, even if there are a plurality of character frames on the form, a necessary character frame can be selected without being erroneously recognized. Can be recognized.
The character cutout unit 3 determines the character cutout reference position 32 based on the recognized character frame 22.

Further, the character cutout means 3 is based on the information on the form edge position detected by the form edge detection means 2 and the information on the character frame position 33 on the basis of the form edge stored in the format parameter storage means 4. The character frame 22 can be recognized in the reading area 21, which is an area in a predetermined range based on the character frame position, and the character cutout reference position 32 can be determined based on the recognized character frame 22. .
Further, the character cutout unit 3 can also recognize the character frame 22 from the entire non-dropout image and determine the character cutout reference position 32 based on the recognized character frame.

  Next, description will be made with reference to FIG. The character cutout unit 3 cuts out the character based on the determined character cutout reference position 32, and the character recognition unit 5 recognizes the character from the cut out image. At this time, the character is cut out by the dropout image.

  As a result, since the character frame is not shown in the dropout image, the character frame is not shown in the clipped area, and the character frame does not become an obstacle to recognition in character recognition. Accurate character recognition is possible.

<Character extraction flow when correcting all characters>
Next, with reference to FIG. 4, the character cutout flow when all characters are corrected will be further described. FIG. 4 is a diagram showing a character cutout flow when all characters are corrected.
The character cutout means 3 includes information on the form edge detected by the form edge detection means 2 and information on the character position or character frame position based on the form edge stored in the format parameter storage means 4 (format parameter designation position). ) Is used to detect a character frame on the non-dropout image (S41).

Next, the character cutout unit 3 calculates the position of the character frame 22 on the screen from the detected image of the character frame 22, and acquires the position information of the character frame 22 (S42).
The character cutout unit 3 determines the character cutout reference position 32 based on the acquired position information of the character frame 22, and cuts out the character from the dropout image (S43).
The character recognition means 5 selects a character component to be read from the cut image and recognizes the character (S44).

<Character cutout flow for correcting misalignment in retry processing>
Next, with reference to FIG. 5, the character cutout flow in the case of correcting the deviation by the retry process will be further described. FIG. 5 is a diagram showing a character cutout flow in the case where deviation correction is performed by retry processing.

The character cutout means 3 uses information related to the form edge detected by the form edge detection means 2 and information related to the character position (format parameter designation position) stored in the format parameter storage means 4 as a reference. Then, the character arrangement position on the dropout image and the character cutout reference position 31 are calculated (S51).
The cutout unit 3 checks the positional deviation between the calculated character cutout reference position 31 and the character 23 and determines the presence or absence of the positional deviation (S52).

  Next, when there is no position shift, the character cutout means 3 cuts out a character based on the character cutout reference position (S58), and when there is a position shift, executes S54 (S53).

In the non-dropout image read by the whole image reading means 6, the character cutout means 3 applies the information relating to the character arrangement position calculated previously, and is a predetermined area based on the character arrangement position. The character frame 22 is detected and recognized in a certain reading area 21 (S54), and information on the position of the character frame 22 is acquired based on the information (S55).
The character cutout unit 3 cuts out characters from the dropout image based on the acquired information on the position of the character frame 22 (S56).
The character recognizing means 5 selects a character component to be read from the clipped image and recognizes the character (S57).

  As mentioned above, although the specific embodiment was shown and demonstrated about the character recognition apparatus of this invention, this invention is not limited to these. Those skilled in the art can make various changes and improvements to the configuration and functions of the character recognition device in each of the above embodiments without departing from the scope of the present invention.

  For example, the processing performed by the form edge detection means 2, the character cutout means 3, and the character recognition means 5 of the present invention can be executed by a program stored in a storage device of a computer. Similar effects can be obtained.

  The character recognition apparatus according to the present invention is realized by an OS, an application, an optical sensor, and the like built on hardware resources including a computer CPU, memory, auxiliary storage device, display, and the like, and is described on a form. Since the process of recognizing the written characters is specifically realized by using the above hardware resources, it corresponds to the technical idea using the laws of nature, and the characters described on the form are It can be used in various fields that need to be recognized, such as mail sorting, courier sorting, questionnaire counting, test counting, and the like.

It is a block diagram of the character recognition apparatus which concerns on one Embodiment of this invention. (A) It is a figure which shows the example which read the form with all the image reading means 6. FIG. (B) It is a figure which shows the example which read the form with the character image reading means 7. FIG. (A) It is a figure which shows the example which read the form with the character image reading means 7. FIG. (B) It is a figure which shows the shift | offset | difference of the calculated character position and an actual character position. (C) is a diagram showing a calculation character frame position in the image read by the all image reading means 6 of the form. (D) It is the figure which extracted the character from the image which the form read with the character image reading means 7. FIG. It is a figure which shows the character extraction flow at the time of all the character corrections. It is a figure which shows the character cutout flow in the case of carrying out deviation correction by retry processing.

DESCRIPTION OF SYMBOLS 1 Image reading means 2 Form edge detection means 3 Character extraction means 4 Format parameter storage means 5 Character recognition means 6 All image reading means 7 Character image reading means 21 Reading area 22 Character frame 23 Character 24 Character extraction reference position 31 Character extraction reference position 32 Character cutout reference position 33 Character frame position based on the form edge stored in the format parameter storage means

Claims (6)

A character recognition device for recognizing characters written in a character frame of a form having a character frame printed in dropout color, wherein the dropout color image on the form is also a non-dropout color image In addition, all image reading means capable of reading both images, character image reading means capable of reading only non-dropout color images on the form, and the edge position of the form are detected, and the edge position of the form is detected. A form edge detecting means for acquiring information on the form, a format parameter storage means for storing a format parameter designation position relating to the character frame position or the character position on the form with reference to the position of the edge of the form, and the character Text that cuts out a character image from a dropout image read by the image reading means A character recognition means for recognizing a character from the character image cut out by the character cut-out means, and the character cut-out means stores information about the form edge detected by the form edge detection means and a format parameter storage a first step of calculating a character position and the character extraction reference positions on dropout image using the format parameters specified position stored in the unit, and position the character extraction reference position calculated by said first step comparing a second step of determining whether the positional deviation between the arrangement position and the reference position, when it is determined that there is no positional deviation by said second step, cutting out characters based on the character extracting reference position When it is determined that there is a displacement in the third step and the second step, all image reading is performed. A fourth step of detecting the character frame position of the reading region from the non-dropout image read by viewing means relative to the character position calculated by said first step, the character frame position detected by the fourth step And a fifth step of cutting out characters from the dropout image, and the character recognition means executes a sixth step of recognizing the characters cut out in the third step or the fifth step. Character recognition device.   The character cut-out means is parallel to the center line through the character cut-out reference position and the vertical center line on the rectangular screen when the character image on the drop-out image is surrounded by the smallest rectangle. The character recognition device according to claim 1, wherein the positional deviation amount is calculated by calculating a positional deviation amount from a straight line.   The character cutout means calculates the positional deviation amount by obtaining a center of gravity of the character image on the dropout image and calculating a deviation amount between the center of gravity and the character cutout reference position. The character recognition device according to claim 1. A character recognition device for recognizing characters written in a character frame of a form having a character frame printed in dropout color, wherein the dropout color image on the form is also a non-dropout color image In addition, all image reading means capable of reading both images, character image reading means capable of reading only non-dropout color images on the form, and the edge position of the form are detected, and the edge position of the form is detected. Form edge detection means for acquiring information on the form, format parameter storage means for storing information on the character frame position and the character position on the form with reference to the position of the edge of the form, and the character image reading means Character cutout means for cutting out a character image from the dropout image read by A character recognizing section for recognizing characters from the character image extracted by means Eject and, a character recognition method in the character recognition apparatus having a, in the character segmentation means, the detected form end by the form edge detecting means The first step of calculating the character placement position and the character cutout reference position on the dropout image using the information on the position and the format parameter designation position stored in the format parameter storage means, and the character cutout calculated by the first step A second step of comparing the reference position and the arrangement position, and determining whether or not there is a positional deviation between the reference position and the arrangement position; and when determining that there is no positional deviation in the second step, Based on the third step of cutting out characters based on the second step and the second step, it was determined that there was a misalignment The fourth step of detecting the character frame position in the reading area based on the character arrangement position calculated by the first step from the non-dropout image read by the whole image reading means, and the detection by the fourth step Performing a fifth step of cutting out characters from the dropout image based on a character frame position, and causing the character recognition means to recognize a character cut out in the third step or the fifth step. The character recognition method in the character recognition apparatus characterized by making it perform . When the character cutout means surrounds the character image on the dropout image with the smallest rectangle, it is parallel to the centerline through the vertical centerline on the rectangular screen and the character cutout reference position. The character recognition method in the character recognition device according to claim 4, wherein a seventh step of calculating the positional deviation amount is executed by calculating a positional deviation amount from a straight line. It said character segmentation means obtains a center of gravity of the character image on the dropout image, and the center of gravity by calculating the deviation between the character extraction reference position, executes an eighth step of performing the positional deviation amount calculating The character recognition method in the character recognition apparatus of Claim 4 characterized by the above-mentioned.

Images