JP2007140703A - Method for reading insurance policy, system thereof, and insurance policy recognition system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To highly accurately recognize insurance policies with different reading positions of target items depending on insurance policies, and which have tint block, ruled lines, and seals, etc. <P>SOLUTION: An insurance policy recognition method has the following steps: a first step for producing an image of the insurance policy by imaging the insurance policy, a second step for detecting and removing the tint block from the image of the insurance policy, a third step for detecting and removing the ruled lines, a fourth step for detecting and removing the seal from the image of the insurance policy, and a step for recognizing character in the image of the insurance policy after completing the above steps. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a health insurance card reading method, a system thereof, and a health insurance card recognition device for reading description items of a health insurance card.

Conventionally, in order to input the description contents of a health insurance card into, for example, a computer, an operator visually checks the health insurance card, and, from the description items of the health insurance card, the necessary items are keyed from the keyboard, and the contents are stored in a computer storage device. (For example, refer to Patent Document 1).
Japanese Patent Laid-Open No. 11-143842

In order to eliminate the keying work of insurance card data, when trying to read the contents of the insurance card mechanically, the insurance card includes the following forms. There was a problem that it could not be read.
(1) There is a background pattern. (2) There are ruled lines. (3) There is an imprint. (4) The reading position of the target item differs depending on the insurance card.

  The insurance card reading method according to the present invention includes a first step of imaging a health insurance card to generate a health insurance card image, a second step of detecting and removing a tint block from the insurance card image, and the insurance card image. At least one step among the third step of detecting and removing ruled lines from and the fourth step of detecting and removing the imprint from the insurance card image, and the second to fourth steps And a step of recognizing characters of the insurance card image after at least one or more of the steps.

  The insurance card reading system according to the present invention detects and removes any one or more of a tint block, ruled line, and imprint from the insurance card image, and an imaging unit that images the insurance card to generate the insurance card image. And a recognition unit for recognizing the characters of the insurance card image.

  According to the present invention, even when the insurance card includes a background pattern, ruled line, and seal, the characters of the insurance card are recognized by removing them, so that the written contents of the insurance card can be recognized with high accuracy. it can.

Embodiment 1. FIG.
FIG. 1 is a system configuration diagram according to the first embodiment of the present invention, which includes an insurance card reading system 1 and a medical computer 2. In this insurance card reading system 1, an image of the insurance card is read by a scanner (FIG. 2), character recognition and confirmation correction are performed, and the reading result is sent to the medical computer 2. The medical computer 2 is a system for constructing a patient database in a medical institution. The output of the insurance card reading system 1 is not necessarily a medical computer. Any work that requires entry of the insurance card description items is acceptable.

  FIG. 2 is a configuration diagram of the insurance policy reading system 1 of FIG. The insurance card reading system 1 includes a scanner control unit 10, a recognition unit 11, and a confirmation / correction unit 12, and a scanner 13 is connected to the scanner control unit 10. The scanner control unit 10 controls the scanner 13 for capturing an insurance card. The recognition unit 11 recognizes characters in the captured image (insurance card image) 15 read by the scanner 13. The confirmation / correction unit 12 confirms the image recognition result 16 recognized by the recognition unit 11 by the operator and corrects it as necessary, and outputs the confirmation / correction result 17 to the medical computer 2. The scanner control unit 10 and the scanner 13 correspond to the imaging unit of the present invention. The recognition unit 11 corresponds to the insurance card recognition device of the present invention.

  FIG. 3 is a configuration diagram of the recognition unit 11 of FIG. The recognition unit 11 includes a control unit 11a, a primary image buffer 20, a secondary image buffer 21, an inclination correction unit 22, a background pattern removal unit 23, a ruled line removal unit 24, an imprint removal unit 25, a character detection unit 26, a character recognition unit 27, and The item search unit 28 is configured. The control unit 11a controls the entire recognition unit 11. The primary image buffer 20 stores the insurance card image captured by the scanner 13. The inclination correction unit 22 corrects the inclination of the insurance card image. First, a copy of the primary image buffer 20 is created and stored in the secondary image buffer 21, and then images after background pattern removal, ruled line removal, and imprint removal are respectively stored. The background pattern removal unit 23 detects and removes a background pattern from the image stored in the secondary image buffer 21. The ruled line removing unit 24 detects and removes ruled lines from the image stored in the secondary image buffer 21. The imprint removal unit 25 detects and removes the imprint from the image stored in the secondary image buffer 21. The character detection unit 26 detects a character position from the image stored in the secondary image buffer 21. The character recognition unit 27 performs character recognition using the image of the detected character position. The item search unit 28 collects the character recognition results and creates a recognition result for each item.

  FIG. 4 is a flowchart showing a process when reading the description of the insurance card by the insurance card reading system 1 of FIG. First, an insurance card is set in the scanner 13 (S11). Reading starts when the operator presses, for example, a reading start button (not shown) of the scanner controller 10 (S12). Imaging by the scanner 13 is started, and the insurance card image is stored in the primary image buffer 20 of the recognition unit 11 via the scanner control unit 10 (S13). The image stored in the primary image buffer 20 is copied to the secondary image buffer 21, and the background image removal, ruled line removal, and imprint removal are performed on the secondary image buffer 21, and character detection and character recognition are performed. A recognition result is created (S14). The operator confirms the recognition result in item units, and corrects necessary parts (S15).

FIG. 5 is a flowchart showing details of the recognition process (S14) of FIG.
The inclination correction unit 22 of the recognition unit 11 detects the inclination of the image read into the primary image buffer 20, rewrites the insurance card image after the inclination correction into the primary image buffer 20, and the insurance card image in the primary image buffer 20. Is written in the secondary image buffer 21 (S20). FIG. 6A shows an example before tilt correction, and FIG. 6B shows an example after tilt correction. In this tilt correction, for example, the tilt angle of the image is obtained, and image processing is performed according to the tilt angle to obtain a corrected image.

  The background pattern removal unit 23 extracts a background pattern component from the insurance card image stored in the secondary image buffer 21, removes the background pattern, and writes it in the secondary image buffer 21 (S21). In the background pattern removal processing, whether or not it is a background pattern is determined based on the change in shading and area size of each pixel. Various methods have already been proposed for the background pattern removal process. For example, after the background pattern is detected by performing the following process (see, for example, Japanese Patent Laid-Open No. 2001-5912), the background pattern removal process is performed. Do.

(A) Red, green, and blue primary color component signals representing each pixel of the target image are obtained for each pixel.
(B) Based on the primary color component signals of red, green and blue, the pixel is identified as a chromatic color pixel or an achromatic pixel to obtain a color identification signal indicating the identification result, and the red, green and blue primary color component signals The signal having the highest luminance is obtained as a multi-value luminance signal.
(C) The color identification signal and the luminance signal are input, and the chromatic color histogram and the achromatic color histogram are obtained by counting the number of pixels that enter each gradation level of the luminance signal for each of the chromatic color pixel and the achromatic color pixel.
(D) The first and second binarized slice values for analyzing the unevenness of the chromatic color histogram and the achromatic color histogram and distinguishing the tint block and the character with respect to the pixel determined to be the chromatic color pixel are determined as the achromatic color pixel. A second binarized slice value for distinguishing the background pattern and the character with respect to each pixel is obtained.
(E) A color identification signal and a luminance signal are input, and for a pixel in which the color identification signal indicates a chromatic pixel, the luminance signal is binarized to white or black with a first binarized slice value, and color identification is performed. For a pixel whose signal indicates an achromatic pixel, the luminance signal is binarized to white or black with the second binarized slice value, and the binarization result is accumulated as a binarized image.
(F) A tint block is detected or recognized based on the accumulated binarized image.

  FIG. 7A shows an example in which small characters are used for the background pattern, and FIG. 7B shows an example in which a color is used for the background pattern. FIGS. 8A and 8B are views showing images after the background pattern is removed by the above processing.

  The ruled line removal unit 24 reads the image from which the background pattern has been removed as described above from the secondary image buffer 21, detects the ruled line from the image, removes it, and writes it in the secondary image buffer 21 (S22). This ruled line removal processing is performed, for example, by regarding an orthogonal straight line as a ruled line and removing it. FIGS. 9A and 9B show an example in which ruled lines are removed from the images of FIGS. 8A and 8B.

  The imprint removal unit 25 reads the image from which the ruled lines have been removed as described above from the secondary image buffer 21, detects the imprint from the image, removes it, and writes it in the secondary image buffer 21 (S23). For this imprint detection, based on either information on the color of the imprint or information on the shape of the imprint, the imprint image of the imprint by the seal is searched. Various methods have already been proposed. Yes. For example, in Japanese Patent Application Laid-Open No. 2002-123794, an imprint image is searched based on the color information of an image, that is, the difference in color information between the character color of an insurance card and the color of an imprint, and Japanese Patent Application Laid-Open No. 2002-269547. In the gazette, the imprint image is searched based on the information on the shape of the imprint search area and the shape of the imprint, and in Japanese Patent Laid-Open No. 2002-304631, based on the shape of the imprint existing in the imprint search area on the form. Although it is disclosed that a search for an imprint image is performed, any method may be used in the present embodiment. FIGS. 10A and 10B show an example in which the seal stamp is removed from FIGS. 9A and 9B.

  The character detection unit 26 reads the image from which the seal is removed as described above from the secondary image buffer 21, and detects the character position of the image (S24). The character recognition unit 27 extracts an image of the detected character position, performs character recognition, and writes it in the secondary image buffer 21 (S25). For example, collation data stored in a memory (not shown) is collated with the image at the character position to recognize a character. FIGS. 11A and 11B show examples of the results of character detection and character recognition performed on the insurance card images of FIGS. 10A and 10B.

  The item search unit 28 reads out the character-recognized image from the secondary image buffer 21, collects the recognition results of the neighborhood from the character recognition results, edits them into the recognition results in units of read items, and stores them in the secondary image buffer 21. Write (S26). The position of the character printed on the insurance card differs depending on the institution (insurer) issuing the insurance card, but since the reading item is searched and edited, the fluctuation of the character position can be absorbed. 12 (a) and 12 (b) are examples in which the recognition result for each item is edited by collecting characters for each item from the results of FIGS. 11 (a) and 11 (b). The images shown in FIGS. 12A and 12B are output to the medical computer 2 after being confirmed and corrected by the operator in the confirmation / correction unit 12 and stored in the storage device of the medical computer 2.

  As described above, in the first embodiment, an insurance card is imaged by the scanner 13 to generate an insurance card image, and after the background pattern, ruled line, and imprint are sequentially detected and removed from the insurance card image by the recognition unit 11, Since the characters of the certificate image are recognized, and the items are searched and edited for the characters, the items can be appropriately removed even when the insurance card has a background pattern, ruled line, or seal. Even when the reading position of the item differs depending on the insurance card, since the reading item is searched and the item is edited, the fluctuation of the reading position can be absorbed. For this reason, it is possible to read the contents of the insurance card with high accuracy. Further, the read result is displayed on the confirmation / correction unit to prompt confirmation / correction by the operator, and is confirmed by the operator at the final stage, so that it becomes accurate data when input to the medical computer 2. Yes. In addition, since the removal process of the ground pattern, ruled line, and imprint of the insurance card is performed in the order of the ground pattern removal, ruled line removal, and imprint removal, these unnecessary information can be surely removed. The accuracy of character recognition has been improved.

  In the above description, an example in which the background pattern, ruled line, and seal stamp are sequentially detected and removed from the insurance card image has been described. However, depending on the nature of the insurance card, only one or more of these processes may be performed. Also good.

It is a system configuration figure concerning Embodiment 1 of the present invention. It is a block diagram of a health insurance card reading system. It is a block diagram of a recognition part. It is a flowchart which shows the process at the time of reading the description item of an insurance card by an insurance card reading system. It is the flowchart which showed the detail of the recognition process. It is the figure which showed the example before inclination correction, and the example after inclination correction. It is the figure which showed the example in which the small character was used for the background pattern, and the example which used the color for the background pattern. It is the figure which showed the image after a tint block was removed. It is the figure which showed the example which removed the ruled line from the image. It is the figure which showed the example which removed the seal impression. It is the figure which showed the example of the result of performing character detection and character recognition. It is the figure which showed the example which edited the recognition result of the item unit.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Insurance card reading system, 2 Medical computer, 10 Scanner control part, 11 Recognition part, 12 Confirmation correction part, 13 Scanner, 15 Captured image, 16 Image recognition result, 17 Confirmation correction result, 20 Primary image buffer, 21 Secondary image Buffer, 22 Tilt correction unit, 23 Background pattern removal unit, 24 Ruled line removal unit, 25 Imprint removal unit, 26 Character detection unit, 27 Character recognition unit, 28 Item search unit

Claims (10)

A first step of imaging a health insurance card to generate a health insurance card image;
A second step of detecting and removing a tint block from the insurance card image, a third step of detecting and removing a ruled line from the insurance card image, and a fourth step of detecting and removing an imprint from the insurance card image At least one of the steps, and
A method for reading a health insurance card, comprising the step of recognizing characters of the health insurance card image after at least one of the second to fourth steps.   The insurance card reading method according to claim 1, wherein the second step, the third step, and the fourth step are performed in time series.   3. The insurance card reading method according to claim 1, further comprising a step of searching for a reading item and editing the item after recognizing characters of the insurance card image.   4. The insurance card reading method according to claim 3, further comprising the step of displaying the item-edited insurance card image on a display means to prompt confirmation / correction. An imaging means for imaging a health insurance card and generating a health insurance card image;
An insurance card reading system, comprising: a recognition unit that recognizes characters of the insurance card image after detecting and removing any one or more of a tint block, a ruled line, and an imprint from the insurance card image. The recognition unit
A tint block removing unit that detects and removes a tint block from the insurance card image, a ruled line removing unit that detects and removes a ruled line from the insurance card image, and an imprint removing unit that detects and removes a stamp from the insurance card image Any one or more means;
A insurance card reading system comprising: a character recognition unit that recognizes characters of the insurance card image after the processing by any one or more of the means.   The insurance card reading system according to claim 6, wherein the recognition unit causes the processing by the tint block removing unit, the ruled line removing unit, and the imprint removing unit to be performed in time series.   The insurance card reading system according to claim 6 or 7, wherein the recognition unit includes an item search unit that searches for a reading item and edits the item after recognizing characters of the insurance card image.   9. The insurance card reading system according to claim 8, further comprising a confirmation correction unit that displays an image of the item-edited insurance card to prompt confirmation and correction. Any of the tint block removing means for detecting and removing the tint block from the insurance card image, the ruled line removing unit for detecting and removing the ruled line from the insurance card image, and the imprint removing unit for detecting and removing the seal stamp from the insurance card image Or one or more means,
A character recognition unit for recognizing characters of the insurance card image after the processing by any one or more means;
An insurance card recognition apparatus comprising: an item search unit that searches for a reading item and edits an item after recognizing characters of the insurance card image.

Images