JP5844698B2 - Character recognition device - Google Patents

Description

  The present invention relates to a character recognition device, and more particularly to an item reading method.

  When the form is read by the scanner, the reading position of each item is calculated on the image (image) on the basis of the outer shape of the form paper, for example, the edge or end face, and the character recognition of each item is performed. However, the method based on the outline of the form is required to have a sufficiently high outline accuracy. Therefore, as will be described below, for example, in a camera type scanner, if a float occurs at the corner of a form, the reference position may be shifted and character recognition may not be performed correctly.

  FIG. 17 is a diagram illustrating an example in which an error occurs in the detection of item coordinates due to the occurrence of floating. FIG. 17A shows a part of an image (image) of the form 80 photographed in a state without floating, that is, in a normal state. The image of the form 80 is processed so that the periphery of the form 80 becomes a black background 90 so that the edges can be easily detected. From this image, the upper left end 80a of the form 80 is detected as a reference position. When the item A is read, the data is read by determining the reading position according to the distance d registered in advance from the end 80a up to the item 82 corresponding to the item A.

  FIG. 17C shows the form 80 placed on the scanner pedestal 54 in a state where the end portion is lifted from the side. When the reference end portion 80a is in such a floating state, only the end portion 80a approaches the camera of the scanner, so that the end portion 80a is enlarged and photographed compared to other portions. FIG. 17B is an image of the form 80 taken with the end 80a floating. As a result of enlarging the vicinity of the end portion 80a, the reference end portion 80a moves outside the original position on the image, so that the interval between the end portion 80a and the item position 82 on the image is greater than d. Furthermore, it extends by Δd. Since the item A is read at an incorrect position, correct character recognition cannot be performed.

  As one of countermeasures against such floating of a form, there has been proposed a distortion detection method for detecting the presence or absence of distortion of a taken form image and giving a warning or the like (Patent Document 1).

2007-272572

  In the method of using the outline of a form as a reference for reading out the form items, it is greatly affected by the cutting accuracy of the paper, the positional accuracy of printing with respect to the paper, or the floating of the form. In order to avoid such an influence, a reading method that does not use the outline of the form as a reference position is required.

  In view of the above problems, an object of the present invention is to provide a character recognition device capable of reading items without being affected by the outer shape of a form.

In order to achieve the above object, in a character recognition device that performs character recognition based on a photographed form image, a form ID that identifies the type of form and a form ID extraction unit that extracts the position of the form ID from the image A definition information acquisition unit that acquires definition information corresponding to the form based on the extracted form ID from a storage unit in which the definition information of the form is stored, and a ruled line information included in the acquired definition information A ruled line search unit that searches for a reference ruled line of the form from the image, and converts item coordinates included in the acquired definition information based on the outline of the form into coordinates based on the position of the form ID. An item coordinate correction unit that corrects the item coordinates corresponding to the image using the position of the reference ruled line on the image, and the item coordinate correction unit. Based on the item coordinates Tadashisa includes a character recognition unit for performing character recognition, the form ID extracting unit, a position of the form ID of the image, close to the center of the character of the document in the character of the form ID Extract based on

  ADVANTAGE OF THE INVENTION According to this invention, the character recognition apparatus which can read an item can be provided, without being influenced by the external shape of a form.

It is a figure for demonstrating the character recognition apparatus 100 and its system to which this invention is applied. It is a figure which shows an example of the specific form 80. FIG. 3 is a functional block diagram showing a specific configuration of the character recognition device 100. FIG. It is a figure which shows the relationship between form ID84 and the item 82. FIG. It is a figure which shows the relationship between form ID84 and the reference ruled line 86. FIG. It is a figure for demonstrating skew feeding information calculation. It is a figure explaining calculation of each expansion / contraction rate of a horizontal direction and a perpendicular direction. It is a figure for demonstrating expansion / contraction correction | amendment in item coordinate correction | amendment. It is a figure for demonstrating skew correction in item coordinate correction. It is a main flowchart of a character recognition process. It is a subroutine of form ID extraction processing. This is a subroutine for definition information acquisition processing. It is a subroutine for ruled line search processing. This is a subroutine for item coordinate correction processing. It is a figure explaining the example corrected without using the standard ruled line. It is an image figure explaining the influence of the cutting of the form 80 with respect to the reading reference | standard of an item. It is a figure explaining the example which produces an error in the detection of an item coordinate by generation | occurrence | production of floating.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a diagram for explaining a terminal device 1 equipped with a character recognition device 100 to which the present invention is applied and a schematic system thereof. This system is, for example, a system that is installed at a window of a financial institution and reads a predetermined deposit or withdrawal slip entered by a customer. A scanner 50, an operation unit 60, and a monitor 70 are connected to the terminal device 1.

  The terminal device 1 includes a CPU 10, a DRAM 12, an HDD 14, an IF unit 18, and an image processing unit 20. As the terminal device 1, a general personal computer can be used. The CPU 10 is a control unit that comprehensively controls the terminal device 1, the scanner 50, the operation unit 60, and the monitor 70. The CPU 10 reads a predetermined program and executes each process of calculation, external control, and character recognition based on an instruction from the operation unit 60. Further, the CPU 10 has a character recognition device 100. When the document 50 is read by the scanner 50, the CPU 10 reads a character recognition program from the HDD 14 and functions as the character recognition device 100. Therefore, it is convenient for the CPU 10 to have a function unit called the character recognition device 100. Express.

The DRAM 12 is a temporary storage memory, and is used as a working area for storing image data read by the scanner 50 and performing image processing on the image data.
The HDD 14 is a storage memory, and the HDD 14 stores various control programs including a character recognition program and necessary data. The HDD 14 stores form definition information 16 necessary for reading a form as data.

  The definition information 16 refers to the position (coordinate) information of each item on the form and information necessary for character recognition of each item, and is also called a definition body. Therefore, the definition information 16 is different for each type of form 80. The definition information 16 includes definition information related to the form ID and information related to the reference ruled line. The form ID is information indicating the type of form, and is printed at a fixed position on the form 80 regardless of the type of form 80. The form ID definition information is information for recognizing the form ID such as the coordinates (position) of the form ID on the form and the number of digits. The reference ruled line is a ruled line used as a reference for correcting the influence of inclination and enlargement / reduction at the time of photographing the form 80, and details will be described later.

  FIG. 2 is an example of a specific form 80. A form 80 shown in FIG. 2 is a horizontally long form and is a deposit process form used in a bank. “2000” printed on the upper left is the form ID 84. A box surrounded by a frame is an area of each item in which numbers and characters are entered by the customer. A form 80 enclosed by a customer in a frame and filled with numbers and characters in each item is captured as an image by the scanner 50, and the numbers and characters entered by the customer are recognized as characters from this image. Normally, each frame is printed in a red or blue color that is a dropout color, and the form ID 84 is printed in black.

  The IF unit 18 is an interface unit that transmits and receives data between a peripheral device (here, the scanner 50 and the operation unit 60) connected to the outside and the CPU 10. For example, the image processing unit 20 creates a screen for various operations to be displayed when the form 50 is read by the scanner 50 based on an instruction from the CPU 10, and displays the created screen on the connected monitor 70. .

  The scanner 50 includes a camera unit 52 that captures the form 80 and a base 54 on which the captured form 80 is placed. The camera unit 52 is disposed on the upper portion of the scanner 50 and is attached so as to photograph the form 80 placed on the pedestal 54 in a non-contact manner from above. The camera unit 52 includes a photographic lens, an imaging device, an imaging unit, and the like, and outputs digitized color image data of the form 80 to the IF unit 18. The scanner 50 may be a line sensor type that reads out the form 80 while moving, instead of the camera type on which the two-dimensional sensor as described above is mounted. In addition, the scanner 50 may be of a type provided with an automatic feeder that automatically feeds a form, instead of a type that reads and puts the forms 80 one by one.

  The operation unit 60 is, for example, a mouse or keyboard that receives an instruction for a form reading process from an operator and notifies the CPU 10 of the content of the instruction via the IF unit 18. The monitor 70 is, for example, an LCD panel, and displays various operation screens created by the image processing unit 20 when the form 80 is read. The monitor 70 may display an image of the form 80 read by the scanner 50 or a read character.

  FIG. 3 is a functional block diagram showing a specific configuration of the character recognition device 100. FIG. 3A is a diagram illustrating an overall configuration of the character recognition device 100. The character recognition device 100 includes a form ID extraction unit 110, a definition information acquisition unit 120, a ruled line search unit 130, an item coordinate correction unit 140, and a character recognition unit 150. Each unit of the character recognition device 100 performs processing while referring to the definition information 16. 4 is a diagram showing the relationship between the form ID 84 and the item 82, and FIG. 5 is a diagram showing the relationship between the form ID 84 and the reference ruled line 86. This will be described with reference to FIGS.

  The form ID extraction unit 110 searches the form ID 84 from the image of the photographed form 80, and performs character recognition on the searched form ID 84. As described above, the position of the form ID 84 based on the upper left end 80a of the form 80 is constant regardless of the type of form. As shown in FIG. 4A, a form ID 84 (9999) is extracted at a position that is a predetermined amount away from the end 80a of the form 80 based on the definition information 16. Only the search for the form ID 84 uses the end 80a of the form 80.

  Each item is read below based on the searched form ID 84, and an ID reference 85 that is a specific reference position of the form ID is detected from the form ID 84. The ID standard 85 is a circumscribed rectangular center of the number closest to the center in the form ID 84 (FIG. 4B). The center coordinate of “9” surrounded by the rightmost square of the form ID 84 “9999” is set as the ID reference 85. (FIG. 4 (A)). The reason why the number closest to the center of the form ID 84 is used as a reference is that the number closer to the center is less affected by the edge of the form 80. FIG. 4C shows a state in which the item 82 is read from a position separated by a predetermined amount with the ID reference 85 as a reference. The reference position of the item 82 is, for example, the upper left corner of the item.

  The definition information acquisition unit 120 reads definition information corresponding to the form ID obtained from the character recognition result of the form ID from the definition information prepared and stored in advance for each form ID in the HDD 14. For example, when the type of the form 80 is the deposit form shown in FIG. 2, the deposit form definition information 16 is read from the HDD 14.

  The ruled line search unit 130 extracts ruled line information that is information related to the reference ruled line from the read definition information 16, searches for the reference ruled line from the image, and calculates the inclination and enlargement / reduction ratio at the time of capturing the form 80 based on the reference ruled line. . Please refer to FIG. FIG. 5A shows the positional relationship between the form ID 84 and the reference ruled line 86 on the definition information 16, and FIG. 5B shows the form ID 84 and the reference ruled line 86 on the image. .

  The ruled line information is attribute information of the reference ruled line and is information regarding the position, direction, and length of the reference ruled line. As the reference ruled lines for correction, the ruled lines suitable for use in the form 80 are selected in advance for each type of form 80 and registered in the definition information 16. The reference ruled line is preferably a straight line having a certain length or more, a line width larger than a certain value, and a certain distance from the form ID 84 as shown in FIG. Further, the reference ruled line may be plural, for example, two in the horizontal direction and the vertical direction.

  FIG. 3B is a diagram illustrating a specific configuration of the ruled line search unit 130. The ruled line search unit 130 includes a ruled line detection unit 132, a skew feed information calculation unit 134, and an enlargement / reduction information calculation unit 136. The ruled line detection unit 132 calculates the position of the reference ruled line 86 based on the form ID 84 (specifically, the ID reference 85). When the position of the reference ruled line 86 is the definition information 16 and is based on the end 80a of the form 80, it is converted into a position based on the form ID 84 (FIG. 5A). Then, based on the position of the reference ruled line 86 calculated based on the form ID 84, the reference ruled line 86 is detected on the image as shown in FIG.

  Based on the detected reference ruled line 86, the skew information calculation unit 134 calculates a skewness θ that is skew information of the form 80 on the image. Here, the inclination θ of the form 80 placed on the pedestal 54 is calculated on the basis of the horizontal and vertical axes on the image. FIG. 6 is a diagram for explaining the skew information calculation, and calculates the inclination θ of the reference ruled line 86 with reference to the horizontal axis 92 on the image indicated by the one-dot chain line. In FIG. 6, the horizontal reference ruled line 86 is taken as an example, but even when the reference ruled line 86 is vertical, the inclination θ can be calculated by calculating the inclination with respect to the vertical axis.

  The enlargement / reduction information calculation unit 136 compares the distance from the form ID 84 to the reference ruled line 86 in the definition information (FIG. 5A) and the distance from the form ID 84 on the image to the reference ruled line 86 (FIG. 5B). By doing so, the enlargement / reduction ratio which is the enlargement / reduction information of the image is obtained.

  FIG. 7 is a diagram for explaining the calculation of the enlargement / reduction ratios in the horizontal direction and the vertical direction. The enlargement / reduction information calculation unit 136 sets the reference ruled line in both the horizontal direction and the vertical direction, and compares the distance between the horizontal reference ruled line 86 and the vertical reference ruled line 86a, thereby making the horizontal and vertical directions You may make it obtain | require each expansion / contraction rate. When the image capturing unit as in the present embodiment is a camera, there is a possibility that enlargement / reduction may occur in both the horizontal direction and the vertical direction, so it is desirable to calculate enlargement / reduction information in both directions. Further, when the image capturing unit is a line sensor, the expansion / contraction in the line direction is sufficiently small. Therefore, the expansion / contraction information may be calculated only in the direction orthogonal to the line.

  The item coordinate correction unit 140 converts coordinates indicating the position of the “character recognition item” (hereinafter referred to as item coordinates) on the definition information into coordinates based on the position of the form ID 84 (hereinafter referred to as ID coordinates). To do. FIG. 3C is a diagram illustrating a specific configuration of the item coordinate correction unit 140. The item coordinate correction unit 140 includes an image coordinate conversion unit 142, an enlargement / reduction correction unit 144, and a skew correction unit 146. FIG. 8 is a diagram for explaining the item coordinate correction. FIG. 8A shows the relationship between the form ID 84 on the definition information 16 and the item 82, and FIG. 8B shows the form ID 84 on the image. And the item 82 are shown. In the definition information, the image coordinate conversion unit 142 converts the coordinates of the item 82 which is based on the end 80a of the form 80 into image coordinates based on the form ID 84.

  The enlargement / reduction correction unit 144 performs the enlargement / reduction correction of the coordinates of the item 82 by using the enlargement / reduction ratio calculated by the enlargement / reduction information calculation unit 136. As a result, as shown in FIG. 8B, the position of the item 82 based on the definition information is set to the item 82a, and the enlargement / reduction correction unit 144 moves, for example, Δx in the horizontal direction and Δy in the vertical direction. It is corrected to the position 82b.

  FIG. 9 is a diagram for explaining the skew correction of the item coordinate correction. FIG. 9A shows an example in which the form 80 is inclined on the image. As described above, the inclination θ of the reference ruled line 86 is calculated based on the horizontal axis 92 on the image.

  FIG. 9B shows the relationship between the form ID 84 and the item 82 on the definition information 16. In the skew correction, it is desirable that the ID reference 85 is set at a position where the distance from the upper side of the form 80 and the distance from the left side are equal distance f.

  FIG. 9C shows how correction is performed on the image. The skew correction unit 146 performs skew correction of the coordinates of the item 82 using the skewness θ calculated by the skew information calculation unit 134. The skew correction unit 146 performs correction by rotating the item 82 by θ degrees around the ID reference 85 of the form ID 84.

  The character recognition unit 150 cuts out a part of the image at the coordinates on the image of each item corrected by the item coordinate correction unit 140 and performs character recognition corresponding to each item. Refer to the definition information 16 for handwritten / printed characters / character types (numerals, kana, kanji, etc.) necessary for character recognition.

  FIG. 10 to FIG. 14 are flowcharts showing the procedure of the form character recognition processing described so far. Such character recognition processing is executed by the CPU 10, particularly the character recognition device 100. FIG. 10 is a main flowchart of the character recognition process. From the captured form 80 image, form ID extraction process (step S10), definition information acquisition process (step S12), ruled line search process (step S14), item coordinate correction process (step S16), character recognition process (step S18) In order. Each process of form ID extraction processing, definition information acquisition processing, ruled line search processing, and item coordinate correction processing will be described with reference to the subroutines of FIGS. However, the form ID extraction processing, definition information acquisition processing, ruled line search processing, and item coordinate correction processing are performed by the above-described form ID extraction unit 110, definition information acquisition unit 120, ruled line search unit 130, and item coordinate correction unit 140. Since the content is equivalent to the function to be executed, the overlapping part is omitted.

  FIG. 11 is a subroutine of form ID extraction processing. Character recognition of the form ID 84 is performed from the captured image of the form 80 (step S20). The form ID 84 is extracted at a position away from the end 80a of the form 80 by a predetermined amount. The position of the form ID 84 is printed at a fixed position from the end 80a of the form 80 regardless of the type of the form, and the position from the end 80a is described in the definition information 16. Then, character recognition is performed on the extracted image of the form ID 84, the number of the form ID 84 is detected, and the form ID 84 is specified. Further, the position of the form ID on the image is detected (step S22). As described above, as the position of the form ID 84, the center coordinate of the number closest to the center of the form ID 84 is set as the ID reference 85. The center coordinate of the number is obtained as the center of the circumscribed rectangle of the number.

  FIG. 12 is a subroutine of definition information acquisition processing. Each definition information corresponding to the form ID is read (step S30). The definition information corresponding to the form ID 84 specified in step S20 is read from the definition information prepared for each form ID and stored in the HDD 14 in advance.

  FIG. 13 is a subroutine for ruled line search processing. Based on the definition information 16 acquired in step S30, a reference ruled line is detected from the image (step S40). The inclination θ on the image is calculated from the inclination of the reference ruled line 86 (step S42). The inclination θ of the reference ruled line 86 with respect to the horizontal axis or vertical axis on the image is calculated to calculate the inclination θ. Next, the distance between the form ID 84 on the image and the reference ruled line 86 is calculated (step S44). Further, the distance between the form ID 84 and the reference ruled line 86 in the definition information 16 is compared (step S46). The scaling ratio is calculated by comparing the distance on the image with the distance in the definition information 16 (step S48). In calculating the enlargement / reduction ratio, as shown in FIG. 7, the horizontal and vertical enlargement / reduction ratios may be calculated using two horizontal and vertical reference ruled lines 86. Note that step S40 corresponds to the function of the ruled line detection unit 132, step S42 corresponds to the function of the skew feeding information calculation unit 134, and step S44, step S46, and step S46 correspond to the function of the enlargement / reduction information calculation unit 136. .

  FIG. 14 is a subroutine of item coordinate correction processing. The item coordinates are converted from the coordinates in the definition information based on the end portion 80a of the form 80 to the ID coordinates based on the position of the form ID 84 (step S50). Next, the ID coordinate of the item is enlarged / reduced (step S52). Using the enlargement / reduction ratio calculated in step S48, the ID coordinate of the item is enlarged / reduced.

Furthermore, the ID coordinate of the item is skewed (step S54). The ID coordinate of the item is corrected using the slope θ calculated in step S42.
Then, based on the ID coordinates of each item 82 of the form 80 on the converted and corrected image, an image portion corresponding to each item is extracted, and characters are recognized from the extracted image (step S18).

  FIG. 15 is a diagram for explaining an example of correction without using the reference ruled line 86. In the description so far, the example in which the skew and enlargement / reduction ratio of the image of the form 80 is calculated and corrected using the reference ruled line 86 has been described, but it is also calculated using printed characters instead of the reference ruled line 86. You can also.

  FIG. 15A is an example of an image of the form 80 that has been photographed. In the definition information 16, reference characters 88 a (ABCD) and reference characters 88 b (XYZ), which are reference characters among characters printed in advance on the form 80, according to the type of the form 80, include a reference ruled line 86. Or instead of the reference ruled line 86. The definition information 16 also includes coordinate information of the reference character 88a and the reference character 88b.

  The reference character 88a is for calculating the enlargement / reduction ratio. The reference character 88a and the reference character 88b are used in combination to calculate the inclination. As the reference character 88a, a character located at a certain distance from the form ID 84 is selected. In addition, as the reference character 88b, a character located at a certain distance in the horizontal direction from the reference character 88a is selected.

  Similar to the ID reference 85 of the form ID 84, the center coordinates of the reference character 88a are calculated on the image as the position of the reference character 88a. When the reference character 88a is a plurality of characters, a character selected in advance is used as the center coordinates. The same applies to the reference character 88b.

  Then, the scaling ratio is calculated by comparing the distance between the definition information 16 of the form ID 84 and the reference character 88a and the distance between the form ID 84 and the reference character 88a on the image. In FIG. 15A, since the reference character 88a is located in the downward direction of the form ID 84, the enlargement / reduction ratio in the vertical direction of the form 80 can be calculated. Then, as shown in FIG. 15B, the inclination θ is calculated by calculating the inclination between the line connecting the reference character 88a and the reference character 88b and the horizontal axis 92 on the image indicated by the one-dot chain line. Can do.

  FIG. 16 is an image diagram for explaining the influence of the cutting of the form 80 on the item reading reference. This is to explain that the item reading reference is converted from the edge 80a to the form ID 84 so that it is not affected by cutting. A form 80b whose cutting position is shifted inward is indicated by a broken line. FIG. 16A shows a case where the item 82 is determined based on the end portion 80a of the form 80 as described above. If the cutting position is shifted as in the form 80b, it is clear that the reading position with respect to the item position 82 is shifted by that amount.

  FIG. 16B shows a case where the item 82 is determined based on the form ID 84 as described in the present embodiment. Although only the extraction of the form ID 84 is affected by the cutting accuracy, the reading of specific items 82 is not affected by the cutting accuracy of the form 80, so there is no possibility that the reading position of the item 82 is shifted. The extraction of the form ID 84 can be handled by setting the extraction range of the form ID 84 wide. Thus, by using the form ID printed at the same time as each item as a reference, it is possible to avoid the influence of the deviation of the cutting position or the deviation of the printing position with respect to the paper.

  As described above, since the position (coordinates) of the item 82 is calculated and determined based on the position of the form ID 84, the position of the item 82 can be correctly determined without being affected by the outline accuracy and floating of the form 80. Can do. This improves the accuracy of character recognition.

According to the embodiment described above, at least the following effects can be obtained.
1) In the present embodiment, since the form ID is set as the origin, even if the form is lifted at the time of shooting, it is not easily affected, and it is not necessary to pay attention to prevent the form from being lifted. Will improve.
2) In the method where the edge of the form is the origin (reference), the position from the edge of the form to the item is important, and high accuracy is required for printing and cutting. In this embodiment, the cutting accuracy of the form is reduced. This can reduce the cost of creating a form.
3) By using the form ID printed at a certain position as a reference regardless of the form type, the reference position can be easily detected.
4) Ruled lines and characters for correcting skew and enlargement / reduction of the form are selected in advance according to the type of the form and registered in the definition information 16 to correct the skew and enlargement / reduction of the image. Regardless, it is possible to correct skew and enlargement / reduction of the image. Even if the form floats or the outer shape is bent or distorted, it is difficult to be affected.
5) Since the reference ruled lines and characters for correction are registered in the definition information in advance, it is not necessary to detect the ruled lines and characters suitable for correction every scan, and correction processing can be performed in a short time.
6) In the method of the present embodiment, since the conventional form can be used as it is, the investment cost can be reduced.
7) Since the relative printing accuracy with respect to the form paper is relaxed, even a form created by a general printer for a personal computer can be used.
8) Since the importance of form edge detection is not reduced, the need for a black background is eliminated, and a stand and image processing for creating a black background are not required.

  Note that the present invention is not limited to the above-described embodiments as they are, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage. Moreover, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the embodiment. For example, all the constituent elements shown in the embodiments may be appropriately combined. Furthermore, constituent elements over different embodiments may be appropriately combined. Of course, various modifications and applications are possible without departing from the spirit of the invention.

1 Terminal device 10 CPU
12 DRAM
14 HDD
16 Definition Information 18 IF Unit 20 Image Processing Unit 50 Scanner 52 Camera Unit 54 Base 60 Operation Unit 70 Monitor 80 Form 80a Edge 82 Item 84 Form ID
85 ID reference 86 Reference ruled line 88 Reference character 90 Black background 100 Character recognition device 110 Form ID extraction unit 120 Definition information acquisition unit 130 Ruled line search unit 132 Ruled line detection unit 134 Skew information calculation unit 136 Enlargement / reduction information calculation unit 140 Item coordinate correction unit 142 Image coordinate conversion unit 144 Enlargement / reduction correction unit 146 Skew correction unit 150 Character recognition unit

Claims (6)

In a character recognition device that performs character recognition based on the image of a photographed form,
A form ID that identifies the type of form, and a form ID extraction unit that extracts the position of the form ID from the image;
A definition information acquisition unit for acquiring definition information corresponding to the form based on the extracted form ID from a storage unit in which the definition information of the form is stored;
A ruled line search unit that searches for a reference ruled line of the form from the image based on the ruled line information included in the acquired definition information;
The item coordinates included in the acquired definition information are converted into coordinates based on the form ID, and the item coordinates are determined using the position of the reference ruled line on the image. An item coordinate correction unit for correcting the image corresponding to the image,
A character recognition unit that performs character recognition based on the item coordinates corrected by the item coordinate correction unit ,
The character recognition device, wherein the form ID extraction unit extracts the position of the form ID of the image based on a character closer to the center of the form among the characters of the form ID . The item coordinate correction unit includes:
An enlargement / reduction information calculation unit that calculates the enlargement / reduction information of the image by comparing the distance between the form ID in the definition information and the reference ruled line and the distance between the form ID in the image and the reference ruled line;
The character recognition apparatus according to claim 1, further comprising an enlargement / reduction correction unit that performs enlargement / reduction correction on the item coordinates based on the calculated enlargement / reduction information. The item coordinate correction unit includes:
A skew information calculation unit that calculates inclination information of the reference ruled line on the image, and a skew correction unit that performs skew correction on the item coordinates based on the calculated skew information. The character recognition device according to claim 1. The character recognition device according to claim 1, wherein the ruled line search unit searches the reference ruled line on the image from the position of the form ID of the image. In a character recognition method that performs character recognition based on the image of a taken form,
Extracting the form ID identifying the form type and the position of the form ID from the image;
Obtaining definition information corresponding to the form based on the extracted form ID from a storage unit storing the form definition information;
Searching for a reference ruled line of the form from the image based on the ruled line information included in the acquired definition information;
Converting item coordinates included in the acquired definition information with reference to the outline of the form into coordinates based on the position of the form ID;
Correcting the item coordinates corresponding to the image using the position of the reference ruled line on the image;
Including character recognition based on the corrected item coordinates;
In the step of extracting from the image, the position of the form ID of the image is extracted with reference to a character closer to the center of the form in the form ID.
A character recognition method characterized by the above . In a program that causes a computer to execute a character recognition method that performs character recognition based on a captured form image,
Extracting from the image a form ID specifying the form type and the position of the form ID;
Obtaining definition information corresponding to the form based on the extracted form ID from a storage unit storing the form definition information;
Searching for a reference ruled line of the form from the image based on the ruled line information included in the acquired definition information;
Converting item coordinates included in the acquired definition information based on the outline of the form into coordinates based on the position of the form ID;
Correcting the item coordinates corresponding to the image using the position of the reference ruled line on the image;
Performing character recognition based on the corrected item coordinates;
In the step of extracting from the image, the position of the form ID of the image is extracted based on a character closer to the center of the form among the characters of the form ID.
A program characterized by that.