WO2020044537A1 - Image comparison device, image comparison method, and program - Google Patents

Image comparison device, image comparison method, and program Download PDF

Info

Publication number
WO2020044537A1
WO2020044537A1 PCT/JP2018/032358 JP2018032358W WO2020044537A1 WO 2020044537 A1 WO2020044537 A1 WO 2020044537A1 JP 2018032358 W JP2018032358 W JP 2018032358W WO 2020044537 A1 WO2020044537 A1 WO 2020044537A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
comparison
unit
format definition
images
Prior art date
Application number
PCT/JP2018/032358
Other languages
French (fr)
Japanese (ja)
Inventor
智洋 林
郷 道場
幸代 川幡
央 佐々木
起一郎 渡邊
智也 萩原
Original Assignee
株式会社Pfu
株式会社富士通コンピュータテクノロジーズ
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 株式会社Pfu, 株式会社富士通コンピュータテクノロジーズ filed Critical 株式会社Pfu
Priority to PCT/JP2018/032358 priority Critical patent/WO2020044537A1/en
Priority to JP2020539985A priority patent/JPWO2020044537A1/en
Publication of WO2020044537A1 publication Critical patent/WO2020044537A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/10—Image acquisition

Definitions

  • the present invention relates to an image matching device, an image matching method, and a program.
  • Patent Literature 1 in an information sharing system that discloses personal information to a plurality of users and supports information sharing, a storage unit that stores personal public information, and a public storage unit that is stored by the storage unit.
  • Information information providing means for providing a notification condition for notifying an information provider who provided the public information of a use state of the user to the public information in response to a user request, and the information based on the notification condition.
  • Notification means for notifying the information provider of the detected use state of the user when the use state of the user with respect to the public information provided by the provision means is detected, wherein the notification condition is the information provider
  • an information sharing support system that can be changed only.
  • Patent Document 2 discloses a storage unit that stores coordinates of an area in a document and identification information corresponding to the area, and a plurality of areas for character recognition in a received document from a newly received document.
  • Creating means for creating a document wherein the area created by the creating means includes an area extracted by a block selection process on a document, and an arbitrary area designated by a user.
  • a comparison unit that compares the coordinates of the created region with the coordinates of the region stored by the storage unit; a comparison result obtained by the comparison unit; and identification information corresponding to the region stored by the storage unit.
  • Determining means for determining identification information corresponding to the coordinates of the area created by the creating means on the basis of the identification information determined by the determining means.
  • Transmitting means for transmitting text information based on character recognition for the area created by the creating means; and a script for inputting text information to an application based on the identification information sent by the sending means.
  • Executing means for executing the specified script, wherein the storing means, based on a result of transmission by the transmitting means, coordinates of the area created by the creating means, and corresponds to the area.
  • Patent Document 3 discloses a storage unit that stores a form image including an item name and data corresponding to the item name, a search unit that searches for a predetermined item name from the form image, An input unit for receiving information for selecting data on the image of the image, an associating unit for associating the selected data with the searched item name, and a character recognizing unit for character recognizing the associated data.
  • a form reader is disclosed.
  • An object of the present invention is to provide an image matching system that supports image matching.
  • An image matching device is a comparison image storage unit that stores a plurality of comparison images generated by performing processing on image data of the same document in areas different from each other, in association with each other, A comparison unit configured to compare each of the comparison images stored in the storage unit and the newly input image to determine a degree of coincidence;
  • the comparison image storage unit stores format definition information defining a format of a document in association with the comparison image, and, from among the format definition information stored in the comparison image storage unit, A format definition selection unit that selects format definition information to be applied based on the determination result by the degree determination unit, and information from a newly input image based on the format definition information selected by the format definition selection unit. And an extraction unit for extracting.
  • the image processing apparatus further includes a comparison image generation unit that generates a comparison image, and a comparison image registration unit that additionally registers a plurality of comparison images generated by the comparison image generation unit in the comparison image storage unit.
  • the comparison image generation unit performs processing on the same input image so that a plurality of generated comparison images have different data sizes.
  • the comparison image generation unit generates a plurality of comparison images by deleting images in different areas from the same input image.
  • the comparison image storage unit includes a comparison image that is the same as the input image, a comparison image in which an arbitrary area of the input image is deleted, a comparison image in which a ruled line frame is deleted from the input image, At least two of a comparison image in which only the outside of the ruled line frame is extracted from the image and a comparison image in which only the ruled line included in the input image is extracted are stored.
  • the coincidence determining unit excludes, from at least one of the comparison images, the image area changed by the exclusion area change unit from comparison targets, and compares the input image with the comparison image to determine a coincidence. I do.
  • An image matching method includes a step of generating a plurality of comparison images generated by performing processing on image data of the same document in areas different from each other, and generating the plurality of comparison images.
  • the method includes the steps of registering in a database in association with each other, and comparing each of the comparison images registered in the database with a newly input image to determine the degree of coincidence.
  • a program according to the present invention includes a step of generating a plurality of comparative images generated by performing processing on image data of the same document on areas different from each other, and associating the generated plurality of comparative images with each other. And causing the computer to execute a step of comparing each of the comparison images registered in the database with the newly input image to determine the degree of coincidence.
  • FIG. 3 is a diagram illustrating learning data in the image matching system 1;
  • FIG. 2 is a diagram illustrating an outline of OCR recognition in the image matching system 1.
  • FIG. 1 is a diagram illustrating a system configuration of an image matching system 1.
  • FIG. 2 is a diagram illustrating a hardware configuration of the image matching device 5.
  • FIG. 2 is a diagram illustrating a functional configuration of an image matching device 5; It is a figure which illustrates an OCR recognition result confirmation screen.
  • FIG. 9 is a diagram illustrating a layout correction screen. It is a figure showing an example of a comparative image patterned.
  • 5 is a flowchart illustrating a learning data generation process (S10) in the image matching system 1. It is a flowchart explaining the image collation processing (S30) in the image collation system 1.
  • FIG. 9 is a diagram for describing an outline of OCR recognition in a comparative example.
  • FIG. 11 is a diagram illustrating an outline of the OCR recognition process in the comparative example.
  • a user creates a format definition for OCR recognition for each type of document, and performs OCR recognition based on the created format definition.
  • the corrected information was not reflected in the format definition. For this reason, there is a problem in that the format definition is forgotten to be corrected after the OCR recognition, or the range of the recognition failure portion occurs every time the OCR recognition is performed, so that the recognition range must be reset.
  • FIG. 1 is a diagram exemplifying learning data managed by the image matching device 5 of the present invention.
  • FIG. 2 is a diagram illustrating an outline of OCR recognition in the image matching system 1.
  • the image matching apparatus 5 of the present invention specifies a document that does not completely match the document to be subjected to the OCR recognition, but matches the plurality of pieces of image data, and thereby, based on the format definition suitable for the document to be subjected to the OCR recognition. It performs character recognition in order to increase the matching rate.
  • the image matching system of the present invention when the user corrects the OCR recognition range, that is, when the character recognition layout is corrected, the image matching system of the present invention generates learning data based on the corrected contents. Therefore, it is not necessary for the user to reset the format definition.
  • FIG. 3 is a diagram illustrating an overall configuration of the image matching system 1.
  • the image matching system 1 includes a plurality of scanners 3a, 3b, 3c and an image matching device 5, and is connected to each other via a network 7.
  • the scanner 3a, the scanner 3b, and the scanner 3c are collectively referred to as a scanner 3.
  • the scanner 3 transmits image data (hereinafter, referred to as an input image) acquired by the optical reading device to the image matching device 5.
  • the image matching device 5 is a computer terminal and performs character recognition of image data received from the scanner 3.
  • the image matching device 5 specifies a format definition suitable for the input image to be used for character recognition, and performs character recognition of the input image by applying the specified format definition. More specifically, a format definition suitable for the input image is specified based on the comparison image generated by the image matching device 5.
  • FIG. 4 is a diagram illustrating a hardware configuration of the image matching device 5.
  • the image matching device 5 includes a CPU 200, a memory 202, an HDD 204, a network interface 206 (network IF 206), a display device 208, and an input device 210, and these components are connected via a bus 212. Connected to each other.
  • the CPU 200 is, for example, a central processing unit.
  • the memory 202 is, for example, a volatile memory and functions as a main storage device.
  • the HDD 204 is, for example, a hard disk drive, and stores a computer program and other data files as a nonvolatile recording device.
  • the network IF 206 is an interface for performing wired or wireless communication.
  • the display device 208 is, for example, a liquid crystal display.
  • the input device 210 is, for example, a keyboard and a mouse.
  • FIG. 5 is a diagram illustrating a functional configuration of the image matching device 5.
  • an image collation program 50 is installed in the image collation device 5, and the image collation program 50 is stored in a recording medium such as a CD-ROM, for example.
  • the learning data database 600 (learning data DB 600) is configured while being installed in the image matching device 5.
  • the learning data DB 600 manages layout data for each document as illustrated in FIG.
  • the layout data includes a format definition for character recognition of an input image, a comparison image associated with the format definition, and feature point data associated with the format definition.
  • the comparison image and the feature point data are elements that determine a format definition used for character recognition of the input image.
  • Part or all of the image matching program 50 may be realized by hardware such as an ASIC, or may be realized by partially borrowing the function of an OS (Operating System). Further, the entire program may be installed on a single computer terminal, or may be installed on a virtual machine on a cloud.
  • OS Operating System
  • the image matching program 50 includes an image acquisition unit 500, a comparison image storage unit 502, a coincidence determination unit 504, a format definition selection unit 506, an extraction unit 508, a layout correction unit 510, a comparison image generation unit 512, and a fixed format definition generation unit 514. , A feature point data extraction unit 516, and a comparison image registration unit 518.
  • the image acquisition unit 500 acquires image data of a document scanned by the scanner 3 and sets it as an input image.
  • the comparison image storage unit 502 stores a plurality of comparison images generated by performing processing on image data of the same document in different regions. More specifically, the comparison image storage unit 502 stores at least two of the five types of comparison images patterned for one document. Further, the comparison image storage unit 502 stores format definition information (hereinafter, referred to as a format definition) that defines the format of the document in association with the comparison image.
  • the format definition is information for specifying a document type for OCR recognition and information for specifying an OCR recognition range by using one of image data obtained by capturing a plurality of semi-standardized documents of the same type.
  • the format definition is information for specifying the OCR recognition range based on the keyword “customer name” and the position from the keyword (a condition consisting of upper, lower, left, and right).
  • the format definition is defined by the user.
  • the coincidence determining unit 504 determines the coincidence by comparing each of the comparison images stored in the comparative image storage unit 502 with the newly input image. When the degree of coincidence between the comparison image and the input image exceeds the reference, the coincidence determination unit 504 determines that the two coincide.
  • the matching degree determination unit 504 extracts learning data candidates used for character recognition of the input image based on the feature point data, and from among the candidates extracted based on the matching degree between the comparison image and the input image. Learning data having a matching degree exceeding a reference is determined.
  • the format definition selection unit 506 selects a format definition to be applied from the format definitions stored in the comparison image storage unit 502 based on the determination result by the matching degree determination unit 504. Specifically, the format definition selection unit 506 selects the format definition of the learning data determined by the matching degree determination unit 504 as the format definition used for character recognition of the input image.
  • the extracting unit 508 extracts information from a newly input image based on the format definition selected by the format definition selecting unit 506. Specifically, the extraction unit 508 performs character recognition on the input image based on the format definition, and displays the recognition result on an OCR recognition result confirmation screen as illustrated in FIG. On the OCR recognition result confirmation screen, each item name (date, telephone number, name, etc.) of the document and the value of the item are displayed. The user confirms the result of character recognition on the OCR recognition result confirmation screen, and corrects any error.
  • the layout correction unit 510 changes the character recognition range of the input image or the meaning (value of date, telephone number, name, etc.) of the item described in the character recognition range. Specifically, as illustrated in FIG. 7, an image of the input image is displayed on the layout correction screen, and when the character recognition range is reset by the user, the layout correction unit 510 receives the change, Change the character recognition range.
  • the comparative image generation unit 512 When the degree of coincidence determined by the degree-of-coincidence determination unit 504 is equal to or less than the reference for any of the comparative images, the comparative image generation unit 512 performs processing on the input image in regions different from each other, Generate a plurality of comparison images. Specifically, the comparison image generation unit 512 performs processing on the same input image so that a plurality of generated comparison images have different data sizes. Further, the comparison image generation unit 512 generates a plurality of comparison images by deleting images in different areas from the same input image.
  • the standard format definition generation unit 514 stores the format definition in which the character recognition range is changed by the layout correction unit 510 or the format definition in which the meaning of the item of the document is changed in the learning data DB 600 in association with the comparison image.
  • the feature point data extraction unit 516 extracts feature points of the comparison image corrected by the layout correction unit 510, and stores the feature points in the learning data DB 600 in association with the comparison image.
  • the comparison image registration unit 518 additionally registers a plurality of comparison images generated by the comparison image generation unit 512 in the comparison image storage unit 502. More specifically, the generated comparison images are stored in the learning data DB 600 in association with the format definition generated by the standard format definition generation unit 514 and the feature point data extracted by the feature point data extraction unit 516.
  • FIG. 8 is a diagram illustrating an example of a patterned comparative image.
  • the learning data DB 600 has five levels of comparison images for one document.
  • the five-stage comparison image includes the same comparison image (original image data) as the input image, a comparison image in which an arbitrary region of the input image is deleted (pattern 1), and a comparison image in which the ruled line frame is deleted from the input image. (Pattern 2), a comparative image extracted only from outside the ruled line frame from the input image, and a comparative image (pattern 4) extracted only from the ruled line included in the input image.
  • the comparison image of pattern 1 is image data in which an area not to be collated is created at random from the original image data. Specifically, in the image data, the area that is not to be collated is placed at a random position (the x-coordinate and the y-coordinate range from (0, 0) to the maximum pixel of the document image data) at a random size (the size of the document). There are a plurality of rectangles (the size is in the range of 5% to 20% of one side (pixel) per side) in the image data (the number is random in the range of 1 to 10).
  • FIG. 9 is a flowchart illustrating the learning data generation process (S10).
  • the image acquiring unit 500 acquires image data of a document scanned by the scanner 3 and sets the image data as an input image.
  • the coincidence determining unit 504 compares the input image with the comparative image, and searches for a comparative image whose coincidence exceeds the reference. If there is no comparison image with the matching degree exceeding the reference, the process proceeds to S110. If there is a comparison image with the matching degree exceeding the reference, the process proceeds to the image matching process (S30).
  • the format definition selection unit 506 acquires a format definition associated with the semi-standard document.
  • step 115 the extraction unit 508 performs character recognition of the input image based on the format definition selected by the format definition selection unit 506.
  • step 120 the extraction unit 508 displays the character recognition result on the OCR recognition result confirmation screen, and the user confirms the result.
  • step 125 if there is a character string that has not been recognized, the process proceeds to S145. If all character strings have been recognized, the process proceeds to S130.
  • step 130 the comparison image generation unit 512 generates comparison images having five levels of different information amounts based on the image data of the semi-standard document used for the character recognition by the extraction unit 508.
  • step 135 the standard format definition generation unit 514 generates a standard document format definition based on the semi-standard document format definition used for character recognition.
  • step 140 the feature point data extraction unit 516 extracts feature points of the image data of the semi-standard document used by the extraction unit 508 for character recognition.
  • the comparative image registration unit 518 stores the generated format definition, the comparative image generated in S130, and the feature point data in association with each other in the learning data DB 600.
  • the layout correction unit 510 resets the range in which the character string is to be recognized based on a user operation performed on the layout correction screen.
  • the extraction unit 508 performs character recognition in the range reset by the layout correction unit 510.
  • step 155 S155
  • the extraction unit 508 receives and reflects the correction of the character recognition result by the user.
  • the comparison image generation unit 512 generates comparison images having five levels of different information amounts based on the image data of the semi-standard document used for the character recognition by the extraction unit 508.
  • step 170 the standard format definition generation unit 514 generates a standard document format definition based on the format definition of the semi-standard document used for character recognition and the correction information by the layout correction unit 510.
  • step 175 the feature point data extraction unit 516 extracts feature points of the reset layout that has been reset.
  • the comparison image registration unit 518 stores the generated format definition, the comparison image generated in S165, and the feature point data in association with each other in the learning data DB 600.
  • step 180 the comparison image storage unit 502 manages the learning data stored in the learning data DB 600. Conventionally, it is necessary to correct the format definition of the character recognition range after OCR recognition.
  • the image collating device 5 sets the character recognition range by the user again or changes the meaning of the document item when the user changes the character recognition range. Since the learning data is generated based on the reset information, there is no need for the user to reset the format definition. There is no forgetting. In other words, it is not necessary to maintain the format definition necessary for recognizing a large number of OCRs.
  • FIG. 10 is a flowchart illustrating the image matching process (S30).
  • step 300 the image acquisition unit 500 acquires image data of a document scanned by the scanner 3 and sets the image data as an input image.
  • step 305 if there is no learning data, the process proceeds to a learning data generation process (S10), and if there is learning data, the process proceeds to S310.
  • step 310 the matching degree determination unit 504 compares the input image with the feature point data stored in the learning data DB 600, and extracts learning data candidates whose matching degree exceeds the reference.
  • step 315 the matching degree determination unit 504 compares the input image with the five-stage comparison image of the extracted candidate learning data.
  • the matching degree determination unit 504 compares the comparison image with the input image in descending order of the information amount. Specifically, the matching degree determination unit 504 determines the input image in the order of the first-stage comparison image, the second-stage comparison image, the third-stage comparison image, the fourth-stage comparison image, and the fifth-stage comparison image. Compare with By comparing the comparison image with the input image in the order of a large amount of information, more accurate collation can be performed.
  • step 320 when the matching degree determination unit 504 determines that there is a comparison image whose matching degree with the input image exceeds the reference, the image matching process (S30) proceeds to S325, and the matching degree If there is no comparison image exceeding the standard, the image comparison process (S30) proceeds to S110 of the learning data generation process (S10).
  • step 325 S325)
  • the format definition selection unit 506 acquires a format definition associated with the comparative image whose degree of coincidence with the comparative image exceeds the reference.
  • step 330 S330
  • the extraction unit 508 performs character recognition of the input image based on the format definition selected by the format definition selection unit 506.
  • step 335 S335), the user checks the recognition result on the OCR recognition result check screen.
  • step 340 if there is any unrecognized character string, the image matching process (S30) shifts to S130 of the learning data generation process (S10). finish.
  • the image collation system 1 of the present embodiment since a plurality of patterns of comparison images are generated for one document, the input image is slightly different from the original image data.
  • the user can specify the format definition by matching any one of the plurality of comparison images without correcting the character recognition range each time. That is, the work efficiency of the character recognition processing, the collation performance, and the collation rate of the character recognition are increased.
  • a comparative image of a plurality of patterns is generated, an area that is not collated is created at random, so that the area that is not collated differs for each document, and the pattern of the comparative image is not fixed.
  • the user recognizes the correction operation of the comparison image by the user and generates and manages new learning data based on the correction information, so that maintenance of the format definition is unnecessary. Become. Further, even if the scanner characteristics are changed due to the change of the model of the scanner 3 and the former format definition cannot be used, the image collating device 5 generates a new format definition by learning. There is no need to create a user-defined format definition.
  • the input image is compared with the five-stage learning data created by the comparison image generation unit 512, but the comparison image of the pattern 1 associated with one document may be changed.
  • the image matching device 5 includes an exclusion area changing unit 520 in addition to the functional configuration illustrated in FIG.
  • the comparison image generation unit 512 randomly creates a non-matching area of the pattern 1 for each document, while the exclusion area change unit 520 changes the already created matching area of the pattern 1.
  • the exclusion area change unit 520 changes at least one of the number of image areas, the size of the image area, and the position of the image area for the image area excluded from the comparison target.
  • a comparison image of one pattern 1 is generated for one document by the comparison image generation unit 512 and is managed
  • the comparison image of the pattern 1 is compared with the input image, no comparison is performed. Since the area is fixed, a document having a high collation rate and a document having a low collation rate appear. And low documents can be reduced.
  • the image scanned by the scanner 3 is transmitted to the image matching device 5 and the image matching device 5 compares the input image with the comparison image.
  • the matching program 50 may be installed, the scanner 3 scans the image, and compares the input image with the comparative image.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Character Input (AREA)
  • Image Analysis (AREA)

Abstract

This image comparison device comprises: a comparison image storage unit that mutually associates and stores a plurality of comparison images, each of which is generated by performing a processing process on a different region represented by image data of the same document; and a matching degree determination unit that compares each comparison image stored in the comparison image storage unit with a newly input image to determine a degree of matching therebetween. The comparison image storage unit associates and stores format definition information, which defines document formats, with the comparison images, and comprises: a format definition selection unit that selects format definition information to be applied, from the format definition information stored in the comparison image storage unit, on the basis of the result of the determination by the matching degree determination unit; and an extraction unit that extracts information from the newly input image on the basis of the format definition information selected by the format definition selection unit.

Description

画像照合装置、画像照合方法、及びプログラムImage matching device, image matching method, and program
　本発明は、画像照合装置、画像照合方法、及びプログラムに関する。 The present invention relates to an image matching device, an image matching method, and a program.
　例えば、特許文献１には、個人の情報を複数のユーザに公開して、情報の共有を支援する情報共有システムにおいて、個人の公開情報を蓄積する蓄積手段と、この蓄積手段で蓄積された公開情報と、前記公開情報を提供した情報提供者にその公開情報に対するユーザの利用状態を通知するための通知条件をユーザの要求に応じて提供する情報提供手段と、前記通知条件に基づき、前記情報提供手段で提供された公開情報に対するユーザの利用状態を検知したとき、その検知したユーザの利用状態を前記情報提供者に通知する通知手段と、を具備し、前記通知条件は、前記情報提供者のみ変更可能であることを特徴とする情報共有支援システムが開示されている。 For example, in Patent Literature 1, in an information sharing system that discloses personal information to a plurality of users and supports information sharing, a storage unit that stores personal public information, and a public storage unit that is stored by the storage unit. Information, information providing means for providing a notification condition for notifying an information provider who provided the public information of a use state of the user to the public information in response to a user request, and the information based on the notification condition. Notification means for notifying the information provider of the detected use state of the user when the use state of the user with respect to the public information provided by the provision means is detected, wherein the notification condition is the information provider There is disclosed an information sharing support system that can be changed only.
　また、特許文献２には、文書内の領域の座標と、前記領域に対応する識別情報とを保存する保存手段と、新たに受信した文書から、受信した文書内における文字認識用の複数の領域を作成する作成手段であって、前記作成手段によって作成される領域が、文書に対するブロックセレクション処理によって抽出される領域と、ユーザが指定する任意の領域とを含む、作成手段と、前記作成手段によって作成された領域の座標と、前記保存手段によって保存された領域の座標とを比較する比較手段と、前記比較手段による比較の結果と、前記保存手段によって保存された領域に対応する識別情報とに基づいて、前記作成手段によって作成された領域の座標に対応する識別情報を決定する決定手段と、前記決定手段によって決定された前記識別情報と、前記作成手段によって作成された領域に対する文字認識に基づくテキスト情報とを送信する送信手段と、前記送信手段によって送信された前記識別情報に基づいて、テキスト情報をアプリケーションに入力するためのスクリプトを特定し、特定されたスクリプトを実行する実行手段と、を有し、前記保存手段は、前記送信手段による送信の結果に基づいて、前記作成手段によって作成された領域の座標と、前記領域に対応する識別情報とを保存することを特徴とするシステムが開示されている。 Further, Patent Document 2 discloses a storage unit that stores coordinates of an area in a document and identification information corresponding to the area, and a plurality of areas for character recognition in a received document from a newly received document. Creating means for creating a document, wherein the area created by the creating means includes an area extracted by a block selection process on a document, and an arbitrary area designated by a user. A comparison unit that compares the coordinates of the created region with the coordinates of the region stored by the storage unit; a comparison result obtained by the comparison unit; and identification information corresponding to the region stored by the storage unit. Determining means for determining identification information corresponding to the coordinates of the area created by the creating means on the basis of the identification information determined by the determining means. Transmitting means for transmitting text information based on character recognition for the area created by the creating means; and a script for inputting text information to an application based on the identification information sent by the sending means. Executing means for executing the specified script, wherein the storing means, based on a result of transmission by the transmitting means, coordinates of the area created by the creating means, and corresponds to the area There is disclosed a system characterized by storing identification information.
　また、特許文献３には、項目名と、この項目名に対応するデータとを含む帳票の画像を記憶する記憶部と、前記帳票の画像から所定の項目名を探索する探索部と、前記帳票の画像上のデータを選択する情報を受け取る入力部と、前記選択されたデータと前記探索された項目名を関連付ける関連付け部と、前記関連付けられたデータを文字認識する文字認識部と、を具備する帳票読取装置が開示されている。 Patent Document 3 discloses a storage unit that stores a form image including an item name and data corresponding to the item name, a search unit that searches for a predetermined item name from the form image, An input unit for receiving information for selecting data on the image of the image, an associating unit for associating the selected data with the searched item name, and a character recognizing unit for character recognizing the associated data. A form reader is disclosed.
特開２００９－１２２７２３号公報JP 2009-122723 A 特開２０１７－８４１９８号公報JP 2017-84198 A 特開２０１８－３７０３６号公報JP 2018-37036 A
　画像の照合を支援する画像照合システムを提供することを目的とする。 を An object of the present invention is to provide an image matching system that supports image matching.
　本発明に係る画像照合装置は、同一の書類の画像データに対して、互いに異なる領域に加工処理を施して生成された複数の比較画像を互いに関連付けて格納する比較画像格納部と、前記比較画像格納部に格納されている比較画像それぞれと、新たに入力された入力画像とを比較して、一致度を判定する一致度判定部とを有する。 An image matching device according to the present invention is a comparison image storage unit that stores a plurality of comparison images generated by performing processing on image data of the same document in areas different from each other, in association with each other, A comparison unit configured to compare each of the comparison images stored in the storage unit and the newly input image to determine a degree of coincidence;
　好適には、前記比較画像格納部は、書類の書式を定義する書式定義情報を、前記比較画像に関連付けて格納し、前記比較画像格納部に格納されている書式定義情報の中から、前記一致度判定部による判定結果に基づいて、適用する書式定義情報を選択する書式定義選択部と、前記書式定義選択部により選択された書式定義情報に基づいて、新たに入力された入力画像から情報を抽出する抽出部とをさらに有する。 Preferably, the comparison image storage unit stores format definition information defining a format of a document in association with the comparison image, and, from among the format definition information stored in the comparison image storage unit, A format definition selection unit that selects format definition information to be applied based on the determination result by the degree determination unit, and information from a newly input image based on the format definition information selected by the format definition selection unit. And an extraction unit for extracting.
　好適には、前記一致度判定部により判定された一致度が、いずれの比較画像についても基準以下であった場合に、前記入力画像に対して、互いに異なる領域に加工処理を施して、複数の比較画像を生成する比較画像生成部と、前記比較画像生成部により生成された複数の比較画像を、前記比較画像格納部に追加登録する比較画像登録部とをさらに有する。 Preferably, when the degree of coincidence determined by the degree of coincidence determination unit is equal to or less than a reference for any of the comparison images, the input image is subjected to processing in mutually different regions, The image processing apparatus further includes a comparison image generation unit that generates a comparison image, and a comparison image registration unit that additionally registers a plurality of comparison images generated by the comparison image generation unit in the comparison image storage unit.
　好適には、前記比較画像生成部は、同一の入力画像に対して、生成される複数の比較画像が互いに異なるデータサイズとなるような加工処理を施す。 Preferably, the comparison image generation unit performs processing on the same input image so that a plurality of generated comparison images have different data sizes.
　好適には、前記比較画像生成部は、同一の入力画像に対して、互いに異なる領域の画像を削除して、複数の比較画像を生成する。 Preferably, the comparison image generation unit generates a plurality of comparison images by deleting images in different areas from the same input image.
　好適には、前記比較画像格納部は、前記入力画像と同一の比較画像、前記入力画像の任意の領域が削除された比較画像、前記入力画像から罫線枠内が削除された比較画像、前記入力画像から罫線枠外のみが抽出された比較画像、及び、前記入力画像に含まれる罫線のみを抽出した比較画像、のうち、少なくとも２つを格納している。 Preferably, the comparison image storage unit includes a comparison image that is the same as the input image, a comparison image in which an arbitrary area of the input image is deleted, a comparison image in which a ruled line frame is deleted from the input image, At least two of a comparison image in which only the outside of the ruled line frame is extracted from the image and a comparison image in which only the ruled line included in the input image is extracted are stored.
　好適には、比較対象から除外される画像領域について、画像領域の数、画像領域の大きさ、及び、画像領域の位置のうち、少なくとも一つを変更する除外領域変更部をさらに有し、前記一致度判定部は、前記比較画像の少なくとも一つについて、前記除外領域変更部により変更された画像領域を比較対象から除外して、前記入力画像と前記比較画像とを比較して一致度を判定する。 Preferably, for the image region to be excluded from the comparison target, the number of image regions, the size of the image region, and the position of the image region, further comprising an exclusion region change unit that changes at least one, The coincidence determining unit excludes, from at least one of the comparison images, the image area changed by the exclusion area change unit from comparison targets, and compares the input image with the comparison image to determine a coincidence. I do.
　本発明に係る画像照合方法は、同一の書類の画像データに対して、互いに異なる領域に加工処理を施して生成された複数の比較画像を生成するステップと、前記生成された複数の比較画像を互いに関連付けてデータベースに登録するステップと、前記データベースに登録された比較画像それぞれと、新たに入力された入力画像とを比較して、一致度を判定するステップとを有する。 An image matching method according to the present invention includes a step of generating a plurality of comparison images generated by performing processing on image data of the same document in areas different from each other, and generating the plurality of comparison images. The method includes the steps of registering in a database in association with each other, and comparing each of the comparison images registered in the database with a newly input image to determine the degree of coincidence.
　本発明に係るプログラムは、同一の書類の画像データに対して、互いに異なる領域に加工処理を施して生成された複数の比較画像を生成するステップと、前記生成された複数の比較画像を互いに関連付けてデータベースに登録するステップと、前記データベースに登録された比較画像それぞれと、新たに入力された入力画像とを比較して、一致度を判定するステップとをコンピュータに実行させる。 A program according to the present invention includes a step of generating a plurality of comparative images generated by performing processing on image data of the same document on areas different from each other, and associating the generated plurality of comparative images with each other. And causing the computer to execute a step of comparing each of the comparison images registered in the database with the newly input image to determine the degree of coincidence.
　画像の照合を支援することができる。 照 合 We can support image collation.
画像照合システム１における学習データを例示する図である。FIG. 3 is a diagram illustrating learning data in the image matching system 1; 画像照合システム１におけるＯＣＲ認識の概要を説明する図である。FIG. 2 is a diagram illustrating an outline of OCR recognition in the image matching system 1. 画像照合システム１のシステム構成を例示する図である。FIG. 1 is a diagram illustrating a system configuration of an image matching system 1. 画像照合装置５のハードウェア構成を例示する図である。FIG. 2 is a diagram illustrating a hardware configuration of the image matching device 5. 画像照合装置５の機能構成を例示する図である。FIG. 2 is a diagram illustrating a functional configuration of an image matching device 5; ＯＣＲ認識結果確認画面を例示する図である。It is a figure which illustrates an OCR recognition result confirmation screen. レイアウト補正画面を例示する図である。FIG. 9 is a diagram illustrating a layout correction screen. パターン化された比較画像の例を示す図である。It is a figure showing an example of a comparative image patterned. 画像照合システム１における、学習データ生成処理（Ｓ１０）を説明するフローチャートである。5 is a flowchart illustrating a learning data generation process (S10) in the image matching system 1. 画像照合システム１における、画像照合処理（Ｓ３０）を説明するフローチャートである。It is a flowchart explaining the image collation processing (S30) in the image collation system 1. 比較例におけるＯＣＲ認識の概要を説明する図である。FIG. 9 is a diagram for describing an outline of OCR recognition in a comparative example.
　［背景］
　本発明がなされた背景を説明する。
　顧客との取引で発生する書類は紙であり、多種多様な書類となっている。これらの書類を画像データから準定型の書類としてＯＣＲ（Ｏｐｔｉｃａｌ　Ｃｈａｒａｃｔｅｒ　Ｒｅｃｏｇｎｉｔｉｏｎ）認識するにあたり、書類の種類毎に準定型のＯＣＲ認識用の書式定義を行う必要がある。書類としてのＯＣＲ認識時、ＯＣＲ認識用の書式定義が照合できなかったり、ＯＣＲ認識されない部分があったりなど、ＯＣＲ認識の不具合が発生することがある。ＯＣＲ認識の不具合が発生した場合、ＯＣＲ認識結果の修正や、ＯＣＲ認識する範囲定義の修正などを行なわなければならず、作業効率が悪い。また、紙からの電子データ化などの事務作業における効率化のニーズも高まっている。
　準定型書類とは、請求書などの書類において、請求する会社によりフォーマットが微妙に異なる書類をいう。 [background]
The background on which the present invention is made will be described.
Documents generated in transactions with customers are paper, which is a wide variety of documents. In recognizing these documents as OCR (Optical Character Recognition) from image data as semi-standard documents, it is necessary to define a format for semi-standard OCR recognition for each type of document. At the time of OCR recognition as a document, OCR recognition problems may occur such that the OCR recognition format definition cannot be collated, or there is a portion that is not OCR recognized. When a problem occurs in the OCR recognition, correction of the OCR recognition result, correction of the definition of the OCR recognition range, and the like must be performed, resulting in poor work efficiency. There is also a growing need for more efficient office work, such as the conversion of electronic data from paper.
The term "semi-standard document" refers to a document such as an invoice, the format of which is slightly different depending on the requesting company.
　図１１は、比較例におけるＯＣＲ認識処理の概要を説明する図である。
　図１１に例示するように、比較例のＯＣＲ認識処理では、書類の種類毎にＯＣＲ認識のための書式定義をユーザが作成し、作成された書式定義に基づいてＯＣＲ認識を行っている。また、準定型の書類として認識し、運用時にＯＣＲ認識に不具合があったとき、ＯＣＲ認識する範囲を修正しても、修正した情報は書式定義に反映されなかった。そのため、ＯＣＲ認識後に書式定義を修正し忘れたり、ＯＣＲ認識を行うたびに、認識不良部分の範囲が発生するため、認識する範囲を再設定しなければならないという問題があった。なお、ＯＣＲ認識する範囲の抽出率を向上するためには、認識する書類毎に書式定義が必要となり、書式定義が膨大な量となった。そのため、書式定義の照合時に、合致する書式定義が見つからなかったり、書式定義の照合に時間がかかったりした。修正したＯＣＲ認識用の書式定義の管理も煩雑になるという問題もあった。 FIG. 11 is a diagram illustrating an outline of the OCR recognition process in the comparative example.
As illustrated in FIG. 11, in the OCR recognition process of the comparative example, a user creates a format definition for OCR recognition for each type of document, and performs OCR recognition based on the created format definition. In addition, when a document was recognized as a semi-standard form and there was a defect in OCR recognition during operation, even if the range of OCR recognition was corrected, the corrected information was not reflected in the format definition. For this reason, there is a problem in that the format definition is forgotten to be corrected after the OCR recognition, or the range of the recognition failure portion occurs every time the OCR recognition is performed, so that the recognition range must be reset. In order to improve the extraction rate of the OCR recognition range, a format definition is required for each document to be recognized, and the format definition becomes enormous. Therefore, at the time of collating the format definitions, a matching format definition was not found, or it took time to collate the format definitions. There is also a problem that management of the corrected OCR recognition format definition becomes complicated.
　図１は、本発明の画像照合装置５が管理する学習データを例示する図である。
　図２は、画像照合システム１におけるＯＣＲ認識の概要を説明する図である。
　上記課題に対して、本発明は、図１に例示するように、一つの書類に対して情報量の異なる複数の画像データ（比較画像）を保持し、複数の画像データは、一つの書式定義に関連付けられている。本発明の画像照合装置５は、ＯＣＲ認識の対象書類と完全に一致しなくとも、これらの複数の画像データに一致する書類を特定することにより、ＯＣＲ認識の対象書類に適した書式定義に基づいた文字認識を行い、照合率を上げるものである。
　また本発明の画像照合システムは、図２に例示するように、ユーザがＯＣＲ認識する範囲を修正した場合、すなわち、文字認識するレイアウトを補正した場合、修正した内容に基づいて、学習データを生成するため、ユーザによる書式定義の再設定が不要である。 FIG. 1 is a diagram exemplifying learning data managed by the image matching device 5 of the present invention.
FIG. 2 is a diagram illustrating an outline of OCR recognition in the image matching system 1.
In order to solve the above problem, the present invention holds a plurality of pieces of image data (comparison images) having different amounts of information for one document as illustrated in FIG. Associated with The image matching apparatus 5 of the present invention specifies a document that does not completely match the document to be subjected to the OCR recognition, but matches the plurality of pieces of image data, and thereby, based on the format definition suitable for the document to be subjected to the OCR recognition. It performs character recognition in order to increase the matching rate.
Further, as illustrated in FIG. 2, when the user corrects the OCR recognition range, that is, when the character recognition layout is corrected, the image matching system of the present invention generates learning data based on the corrected contents. Therefore, it is not necessary for the user to reset the format definition.
　本発明の実施形態を、図面を参照して説明する。
　図３は、画像照合システム１の全体構成を例示する図である。
　図３に例示するように、画像照合システム１は、複数のスキャナ３a、スキャナ３ｂ、スキャナ３ｃ及び画像照合装置５を含み、ネットワーク７を介して互いに接続している。
　スキャナ３ａ、スキャナ３ｂ、スキャナ３ｃを合わせてスキャナ３と称する。スキャナ３は、光学式の読取装置で取得した画像データ（以下、入力画像という）を画像照合装置５に送信する。
　画像照合装置５は、コンピュータ端末であり、スキャナ３から受信した画像データの文字認識を行う。具体的には、画像照合装置５は、文字認識するために使用する、入力画像に適した書式定義を特定し、特定した書式定義を適用して入力画像の文字認識を行う。より具体的には、画像照合装置５が生成した比較画像に基づいて入力画像に適した書式定義を特定する。 An embodiment of the present invention will be described with reference to the drawings.
FIG. 3 is a diagram illustrating an overall configuration of the image matching system 1.
As illustrated in FIG. 3, the image matching system 1 includes a plurality of scanners 3a, 3b, 3c and an image matching device 5, and is connected to each other via a network 7.
The scanner 3a, the scanner 3b, and the scanner 3c are collectively referred to as a scanner 3. The scanner 3 transmits image data (hereinafter, referred to as an input image) acquired by the optical reading device to the image matching device 5.
The image matching device 5 is a computer terminal and performs character recognition of image data received from the scanner 3. Specifically, the image matching device 5 specifies a format definition suitable for the input image to be used for character recognition, and performs character recognition of the input image by applying the specified format definition. More specifically, a format definition suitable for the input image is specified based on the comparison image generated by the image matching device 5.
　図４は、画像照合装置５のハードウェア構成を例示する図である。
　図４に例示するように、画像照合装置５は、ＣＰＵ２００、メモリ２０２、ＨＤＤ２０４、ネットワークインタフェース２０６（ネットワークＩＦ２０６）、表示装置２０８、及び入力装置２１０を有し、これらの構成はバス２１２を介して互いに接続している。
　ＣＰＵ２００は、例えば、中央演算装置である。
　メモリ２０２は、例えば、揮発性メモリであり、主記憶装置として機能する。
　ＨＤＤ２０４は、例えば、ハードディスクドライブ装置であり、不揮発性の記録装置としてコンピュータプログラムやその他のデータファイルを格納する。
　ネットワークＩＦ２０６は、有線又は無線で通信するためのインタフェースである。
　表示装置２０８は、例えば、液晶ディスプレイである。
　入力装置２１０は、例えば、キーボード及びマウスである。 FIG. 4 is a diagram illustrating a hardware configuration of the image matching device 5.
As illustrated in FIG. 4, the image matching device 5 includes a CPU 200, a memory 202, an HDD 204, a network interface 206 (network IF 206), a display device 208, and an input device 210, and these components are connected via a bus 212. Connected to each other.
The CPU 200 is, for example, a central processing unit.
The memory 202 is, for example, a volatile memory and functions as a main storage device.
The HDD 204 is, for example, a hard disk drive, and stores a computer program and other data files as a nonvolatile recording device.
The network IF 206 is an interface for performing wired or wireless communication.
The display device 208 is, for example, a liquid crystal display.
The input device 210 is, for example, a keyboard and a mouse.
　図５は、画像照合装置５の機能構成を例示する図である。
　図５に例示するように、画像照合装置５には、画像照合プログラム５０がインストールされ、画像照合プログラム５０は、例えば、ＣＤ－ＲＯＭ等の記録媒体に格納されており、この記録媒体を介して、画像照合装置５にインストールされると共に学習データデータベース６００（学習データＤＢ６００）が構成される。
　学習データＤＢ６００とは、図２に例示するように、書類毎のレイアウトデータを管理する。レイアウトデータとは、入力画像の文字認識をするための書式定義、書式定義に関連付けられる比較画像、及び書式定義に関連付けられる特長点データを含む。比較画像及び特長点データは、入力画像に対する文字認識のために使用する書式定義を決定する要素である。
　なお、画像照合プログラム５０の一部又は全部は、ＡＳＩＣなどのハードウェアにより実現されてもよく、また、ＯＳ（Ｏｐｅｒａｔｉｎｇ Ｓｙｓｔｅｍ）の機能を一部借用して実現されてもよい。また、このプログラム全体が一台のコンピュータ端末にインストールされてもよいし、クラウド上の仮想マシンにインストールされてもよい。 FIG. 5 is a diagram illustrating a functional configuration of the image matching device 5.
As illustrated in FIG. 5, an image collation program 50 is installed in the image collation device 5, and the image collation program 50 is stored in a recording medium such as a CD-ROM, for example. The learning data database 600 (learning data DB 600) is configured while being installed in the image matching device 5.
The learning data DB 600 manages layout data for each document as illustrated in FIG. The layout data includes a format definition for character recognition of an input image, a comparison image associated with the format definition, and feature point data associated with the format definition. The comparison image and the feature point data are elements that determine a format definition used for character recognition of the input image.
Part or all of the image matching program 50 may be realized by hardware such as an ASIC, or may be realized by partially borrowing the function of an OS (Operating System). Further, the entire program may be installed on a single computer terminal, or may be installed on a virtual machine on a cloud.
　画像照合プログラム５０は、画像取得部５００、比較画像格納部５０２、一致度判定部５０４、書式定義選択部５０６、抽出部５０８、レイアウト補正部５１０、比較画像生成部５１２、定型書式定義生成部５１４、特長点データ抽出部５１６、及び比較画像登録部５１８を有する。 The image matching program 50 includes an image acquisition unit 500, a comparison image storage unit 502, a coincidence determination unit 504, a format definition selection unit 506, an extraction unit 508, a layout correction unit 510, a comparison image generation unit 512, and a fixed format definition generation unit 514. , A feature point data extraction unit 516, and a comparison image registration unit 518.
　画像照合プログラム５０において、画像取得部５００は、スキャナ３によりスキャンされた書類の画像データを取得し、入力画像とする。
　比較画像格納部５０２は、同一の書類の画像データに対して、互いに異なる領域に加工処理を施して生成された複数の比較画像を互いに関連付けて格納する。具体的には、比較画像格納部５０２は、一つの書類に対してパターン化された比較画像５種類のうち、少なくとも２つを格納する。また、比較画像格納部５０２は、書類の書式を定義する書式定義情報（以下、書式定義とする。）を、比較画像に関連付けて格納する。書式定義とは、同じ種類の準定型書類を複数取り込んだ画像データから１つを使用し、ＯＣＲ認識するために書類の種類を特定する情報、及びＯＣＲ認識する範囲を特定する情報である。例えば、書式定義とは、ＯＣＲ認識する範囲を、キーワードである「お客様名」とキーワードからの位置(上、下、左、右からなる条件)に基づいて特定する情報である。書式定義はユーザにより定義される。 In the image collation program 50, the image acquisition unit 500 acquires image data of a document scanned by the scanner 3 and sets it as an input image.
The comparison image storage unit 502 stores a plurality of comparison images generated by performing processing on image data of the same document in different regions. More specifically, the comparison image storage unit 502 stores at least two of the five types of comparison images patterned for one document. Further, the comparison image storage unit 502 stores format definition information (hereinafter, referred to as a format definition) that defines the format of the document in association with the comparison image. The format definition is information for specifying a document type for OCR recognition and information for specifying an OCR recognition range by using one of image data obtained by capturing a plurality of semi-standardized documents of the same type. For example, the format definition is information for specifying the OCR recognition range based on the keyword “customer name” and the position from the keyword (a condition consisting of upper, lower, left, and right). The format definition is defined by the user.
　一致度判定部５０４は、比較画像格納部５０２に格納されている比較画像それぞれと、新たに入力された入力画像とを比較して、一致度を判定する。一致度判定部５０４は、比較画像と入力画像との一致度が基準を超える場合に両者が一致していると判定する。また、一致度判定部５０４は、特長点データに基づいて、入力画像に対する文字認識に使用する学習データの候補を抽出し、比較画像と入力画像との一致度に基づいて抽出した候補の中から基準を超える一致度を有する学習データを決定する。 The coincidence determining unit 504 determines the coincidence by comparing each of the comparison images stored in the comparative image storage unit 502 with the newly input image. When the degree of coincidence between the comparison image and the input image exceeds the reference, the coincidence determination unit 504 determines that the two coincide. In addition, the matching degree determination unit 504 extracts learning data candidates used for character recognition of the input image based on the feature point data, and from among the candidates extracted based on the matching degree between the comparison image and the input image. Learning data having a matching degree exceeding a reference is determined.
　書式定義選択部５０６は、比較画像格納部５０２に格納されている書式定義の中から、一致度判定部５０４による判定結果に基づいて、適用する書式定義を選択する。具体的には、書式定義選択部５０６は、一致度判定部５０４により決定された学習データの書式定義を、入力画像の文字認識に使用する書式定義として選択する。 The format definition selection unit 506 selects a format definition to be applied from the format definitions stored in the comparison image storage unit 502 based on the determination result by the matching degree determination unit 504. Specifically, the format definition selection unit 506 selects the format definition of the learning data determined by the matching degree determination unit 504 as the format definition used for character recognition of the input image.
　抽出部５０８は、書式定義選択部５０６により選択された書式定義に基づいて、新たに入力された入力画像から、情報を抽出する。具体的には、抽出部５０８は、入力画像を書式定義に基づいて文字認識し、図６に例示するように、認識結果をＯＣＲ認識結果確認画面に表示する。ＯＣＲ認識結果確認画面では、書類の各項目名（日付、電話番号、名前等）と項目の値が表示される。ユーザは、ＯＣＲ認識結果確認画面で文字認識の結果を確認し、誤りがある場合は、修正する。 The extracting unit 508 extracts information from a newly input image based on the format definition selected by the format definition selecting unit 506. Specifically, the extraction unit 508 performs character recognition on the input image based on the format definition, and displays the recognition result on an OCR recognition result confirmation screen as illustrated in FIG. On the OCR recognition result confirmation screen, each item name (date, telephone number, name, etc.) of the document and the value of the item are displayed. The user confirms the result of character recognition on the OCR recognition result confirmation screen, and corrects any error.
　レイアウト補正部５１０は、入力画像の文字認識する範囲、または文字認識する範囲に記載されている項目の意味（日付、電話番号、名前等の値）を変更する。具体的には、図７に例示するように、レイアウト補正画面には、入力画像のイメージが表示され、ユーザにより文字認識範囲が再設定された場合に、レイアウト補正部５１０は、変更を受け付け、文字認識する範囲を変更する。 The layout correction unit 510 changes the character recognition range of the input image or the meaning (value of date, telephone number, name, etc.) of the item described in the character recognition range. Specifically, as illustrated in FIG. 7, an image of the input image is displayed on the layout correction screen, and when the character recognition range is reset by the user, the layout correction unit 510 receives the change, Change the character recognition range.
　比較画像生成部５１２は、一致度判定部５０４により判定された一致度が、いずれの比較画像についても基準以下であった場合に、入力画像に対して、互いに異なる領域に加工処理を施して、複数の比較画像を生成する。具体的には、比較画像生成部５１２は、同一の入力画像に対して、生成される複数の比較画像が互いに異なるデータサイズとなるような加工処理を施す。また、比較画像生成部５１２は、同一の入力画像に対して、互いに異なる領域の画像を削除して、複数の比較画像を生成する。 When the degree of coincidence determined by the degree-of-coincidence determination unit 504 is equal to or less than the reference for any of the comparative images, the comparative image generation unit 512 performs processing on the input image in regions different from each other, Generate a plurality of comparison images. Specifically, the comparison image generation unit 512 performs processing on the same input image so that a plurality of generated comparison images have different data sizes. Further, the comparison image generation unit 512 generates a plurality of comparison images by deleting images in different areas from the same input image.
　定型書式定義生成部５１４は、レイアウト補正部５１０により文字認識する範囲を変更された場合に、または、書類の項目の意味が変更された書式定義を比較画像に関連付けて学習データＤＢ６００に保存する。
　特長点データ抽出部５１６は、レイアウト補正部５１０により補正された比較画像の特長点を抽出し、比較画像に関連付けて学習データＤＢ６００に保存する。
　比較画像登録部５１８は、比較画像生成部５１２により生成された複数の比較画像を、比較画像格納部５０２に追加登録する。具体的には、生成された複数の比較画像を定型書式定義生成部５１４により生成された書式定義、及び特長点データ抽出部５１６により抽出された特長点データに関連付けて学習データＤＢ６００に保存する。 The standard format definition generation unit 514 stores the format definition in which the character recognition range is changed by the layout correction unit 510 or the format definition in which the meaning of the item of the document is changed in the learning data DB 600 in association with the comparison image.
The feature point data extraction unit 516 extracts feature points of the comparison image corrected by the layout correction unit 510, and stores the feature points in the learning data DB 600 in association with the comparison image.
The comparison image registration unit 518 additionally registers a plurality of comparison images generated by the comparison image generation unit 512 in the comparison image storage unit 502. More specifically, the generated comparison images are stored in the learning data DB 600 in association with the format definition generated by the standard format definition generation unit 514 and the feature point data extracted by the feature point data extraction unit 516.
　次に比較画像について説明する。
　図８は、パターン化された比較画像の例を示す図である。
　本例では、図８に例示するように、学習データＤＢ６００は、一つの書類に対して５段階の比較画像を有する。５段階の比較画像とは、入力画像と同一の比較画像（オリジナル画像データ）、入力画像の任意の領域が削除された比較画像（パターン１）、入力画像から罫線枠内が削除された比較画像（パターン２）、入力画像から罫線枠外のみ抽出された比較画像、及び、入力画像に含まれる罫線のみを抽出した比較画像（パターン４）である。
　書類毎に５段階の比較画像が用意されているため、軽微な変更がなされた書類が入力画像である場合でも、５段階のいずれかと一致すると判定されれば、書式定義を特定することができ、入力画像に対する文字認識が可能となり、照合率が向上する。
　また、パターン１の比較画像は、オリジナルの画像データから照合しない領域をランダムに作成した画像データである。具体的には、画像データ中に、照合しない領域は、ランダムな位置(ｘ座標及びｙ座標は(０，０)から書類の画像データの最大ピクセルの範囲)に、ランダムな大きさ(書類の画像データにおける１辺当たり(ピクセル)の５％～２０％の範囲の大きさ)の矩形で、複数個(個数は１～１０の範囲でランダム)存在する。 Next, a comparative image will be described.
FIG. 8 is a diagram illustrating an example of a patterned comparative image.
In this example, as illustrated in FIG. 8, the learning data DB 600 has five levels of comparison images for one document. The five-stage comparison image includes the same comparison image (original image data) as the input image, a comparison image in which an arbitrary region of the input image is deleted (pattern 1), and a comparison image in which the ruled line frame is deleted from the input image. (Pattern 2), a comparative image extracted only from outside the ruled line frame from the input image, and a comparative image (pattern 4) extracted only from the ruled line included in the input image.
Since five levels of comparison images are prepared for each document, even if the document that has been slightly changed is an input image, if it is determined that the document matches one of the five levels, the format definition can be specified. In addition, the character recognition for the input image becomes possible, and the matching rate is improved.
Further, the comparison image of pattern 1 is image data in which an area not to be collated is created at random from the original image data. Specifically, in the image data, the area that is not to be collated is placed at a random position (the x-coordinate and the y-coordinate range from (0, 0) to the maximum pixel of the document image data) at a random size (the size of the document). There are a plurality of rectangles (the size is in the range of 5% to 20% of one side (pixel) per side) in the image data (the number is random in the range of 1 to 10).
　図９は、学習データ生成処理（Ｓ１０）を説明するフローチャートである。
　図９に例示するように、ステップ１００（Ｓ１００）において、画像取得部５００は、スキャナ３によりスキャンされた書類の画像データを取得し、入力画像とする。
　ステップ１０５（Ｓ１０５）において、一致度判定部５０４は、入力画像と比較画像とを比較し、一致度が基準を超える比較画像を検索する。一致度が基準を超える比較画像がない場合は、Ｓ１１０へ移行し、一致度が基準を超える比較画像が存在する場合は、画像照合処理（Ｓ３０）に移行する。
　ステップ１１０（Ｓ１１０）において、書式定義選択部５０６は、準定型書類に関連付けられる書式定義を取得する。
　ステップ１１５（Ｓ１１５）において、抽出部５０８は、入力画像の文字認識を書式定義選択部５０６により選択された書式定義に基づいて行う。
　ステップ１２０（Ｓ１２０）において、抽出部５０８は、文字認識結果を、ＯＣＲ認識結果確認画面に表示し、ユーザは結果を確認する。 FIG. 9 is a flowchart illustrating the learning data generation process (S10).
As illustrated in FIG. 9, in step 100 (S100), the image acquiring unit 500 acquires image data of a document scanned by the scanner 3 and sets the image data as an input image.
In step 105 (S105), the coincidence determining unit 504 compares the input image with the comparative image, and searches for a comparative image whose coincidence exceeds the reference. If there is no comparison image with the matching degree exceeding the reference, the process proceeds to S110. If there is a comparison image with the matching degree exceeding the reference, the process proceeds to the image matching process (S30).
In step 110 (S110), the format definition selection unit 506 acquires a format definition associated with the semi-standard document.
In step 115 (S115), the extraction unit 508 performs character recognition of the input image based on the format definition selected by the format definition selection unit 506.
In step 120 (S120), the extraction unit 508 displays the character recognition result on the OCR recognition result confirmation screen, and the user confirms the result.
　ステップ１２５（Ｓ１２５）において、認識されていない文字列がある場合は、Ｓ１４５へ移行し、すべて認識されている場合は、Ｓ１３０に移行する。
　ステップ１３０（Ｓ１３０）において、比較画像生成部５１２は、抽出部５０８により文字認識に使用された準定型書類の画像データに基づいて、５段階の情報量の異なる比較画像を生成する。
　ステップ１３５（Ｓ１３５）において、定型書式定義生成部５１４は、文字認識に使用した準定型書類の書式定義に基づいて、定型書類の書式定義を生成する。
　ステップ１４０（Ｓ１４０）において、特長点データ抽出部５１６は、抽出部５０８により文字認識に使用された準定型書類の画像データの特長点を抽出する。比較画像登録部５１８は、生成された書式定義とＳ１３０において生成した比較画像と特長点データとを関連づけて学習データＤＢ６００に格納する。
　ステップ１４５（Ｓ１４５）において、レイアウト補正部５１０は、レイアウト補正画面に対してなされたユーザの操作に基づいて、文字列を認識させたい範囲を再設定する。
　ステップ１５０（Ｓ１５０）において、抽出部５０８は、レイアウト補正部５１０により再設定された範囲において文字認識を行う。
　ステップ１５５（Ｓ１５５）において、文字認識の結果に誤りがある場合は、Ｓ１６０へ移行し、誤りがない場合は、Ｓ１６５へ移行する。
　ステップ１６０（Ｓ１６０）において、抽出部５０８は、ユーザによる文字認識結果の修正を受け付け、反映する。
　ステップ１６５（Ｓ１６５）において、比較画像生成部５１２は、抽出部５０８により文字認識に使用された準定型書類の画像データに基づいて、５段階の情報量の異なる比較画像を生成する。 In step 125 (S125), if there is a character string that has not been recognized, the process proceeds to S145. If all character strings have been recognized, the process proceeds to S130.
In step 130 (S130), the comparison image generation unit 512 generates comparison images having five levels of different information amounts based on the image data of the semi-standard document used for the character recognition by the extraction unit 508.
In step 135 (S135), the standard format definition generation unit 514 generates a standard document format definition based on the semi-standard document format definition used for character recognition.
In step 140 (S140), the feature point data extraction unit 516 extracts feature points of the image data of the semi-standard document used by the extraction unit 508 for character recognition. The comparative image registration unit 518 stores the generated format definition, the comparative image generated in S130, and the feature point data in association with each other in the learning data DB 600.
In step 145 (S145), the layout correction unit 510 resets the range in which the character string is to be recognized based on a user operation performed on the layout correction screen.
In step 150 (S150), the extraction unit 508 performs character recognition in the range reset by the layout correction unit 510.
In step 155 (S155), if there is an error in the result of character recognition, the process proceeds to S160, and if there is no error, the process proceeds to S165.
In step 160 (S160), the extraction unit 508 receives and reflects the correction of the character recognition result by the user.
In step 165 (S165), the comparison image generation unit 512 generates comparison images having five levels of different information amounts based on the image data of the semi-standard document used for the character recognition by the extraction unit 508.
　ステップ１７０（Ｓ１７０）において、定型書式定義生成部５１４は、文字認識に使用した準定型書類の書式定義、及びレイアウト補正部５１０による補正情報に基づいて定型書類の書式定義を生成する。
　ステップ１７５（Ｓ１７５）において、特長点データ抽出部５１６は再設定された補正レイアウトの特長点を抽出する。比較画像登録部５１８は、生成された書式定義とＳ１６５において生成した比較画像と特長点データとを関連づけて学習データＤＢ６００に格納する。
　ステップ１８０（Ｓ１８０）において、比較画像格納部５０２は、学習データＤＢ６００に格納される学習データを管理する。
　従来では、ＯＣＲ認識後に文字認識の範囲の書式定義の修正が必要であったが、画像照合装置５は、ユーザによる文字認識範囲の再設定、または書類の項目の意味が変更された場合に、再設定された情報に基づいて学習データを生成するため、ユーザが書式定義の再設定をする必要はなく、従来のようなユーザによる書式定義の修正の手間が省け、さらに、書式定義の修正のし忘れが生じることもない。つまり、膨大な数のＯＣＲ認識に必要な書式定義のメンテナンスが不要となる。 In step 170 (S170), the standard format definition generation unit 514 generates a standard document format definition based on the format definition of the semi-standard document used for character recognition and the correction information by the layout correction unit 510.
In step 175 (S175), the feature point data extraction unit 516 extracts feature points of the reset layout that has been reset. The comparison image registration unit 518 stores the generated format definition, the comparison image generated in S165, and the feature point data in association with each other in the learning data DB 600.
In step 180 (S180), the comparison image storage unit 502 manages the learning data stored in the learning data DB 600.
Conventionally, it is necessary to correct the format definition of the character recognition range after OCR recognition. However, the image collating device 5 sets the character recognition range by the user again or changes the meaning of the document item when the user changes the character recognition range. Since the learning data is generated based on the reset information, there is no need for the user to reset the format definition. There is no forgetting. In other words, it is not necessary to maintain the format definition necessary for recognizing a large number of OCRs.
　図１０は、画像照合処理（Ｓ３０）を説明するフローチャートである。
　図１０に例示するように、ステップ３００（Ｓ３００）において、画像取得部５００は、スキャナ３によりスキャンされた書類の画像データを取得し、入力画像とする。
　ステップ３０５（Ｓ３０５）において、学習データがない場合は、学習データ生成処理（Ｓ１０）へ移行し、学習データが存在する場合は、Ｓ３１０へ移行する。
　ステップ３１０（Ｓ３１０）において、一致度判定部５０４は、入力画像と学習データＤＢ６００に保持される特長点データとを比較し、一致度が基準を超える習データの候補を抽出する。
　ステップ３１５（Ｓ３１５）において、一致度判定部５０４は、抽出された候補となる学習データの５段階の比較画像と入力画像とを比較する。一致度判定部５０４は、比較画像の情報量の多い順に入力画像と比較する。具体的には、一致度判定部５０４は、第１段階の比較画像、第２段階の比較画像、第３段階の比較画像、第４段階の比較画像、第５段階の比較画像の順に入力画像と比較する。情報量の多い比較画像の順に入力画像と比較することでより正確性の高い照合が可能になる。 FIG. 10 is a flowchart illustrating the image matching process (S30).
As illustrated in FIG. 10, in step 300 (S300), the image acquisition unit 500 acquires image data of a document scanned by the scanner 3 and sets the image data as an input image.
In step 305 (S305), if there is no learning data, the process proceeds to a learning data generation process (S10), and if there is learning data, the process proceeds to S310.
In step 310 (S310), the matching degree determination unit 504 compares the input image with the feature point data stored in the learning data DB 600, and extracts learning data candidates whose matching degree exceeds the reference.
In step 315 (S315), the matching degree determination unit 504 compares the input image with the five-stage comparison image of the extracted candidate learning data. The matching degree determination unit 504 compares the comparison image with the input image in descending order of the information amount. Specifically, the matching degree determination unit 504 determines the input image in the order of the first-stage comparison image, the second-stage comparison image, the third-stage comparison image, the fourth-stage comparison image, and the fifth-stage comparison image. Compare with By comparing the comparison image with the input image in the order of a large amount of information, more accurate collation can be performed.
　ステップ３２０（Ｓ３２０）において、一致度判定部５０４により、入力画像との一致度が基準を超える比較画像が存在すると判定された場合に、画像照合処理（Ｓ３０）は、Ｓ３２５へ移行し、一致度が基準を超える比較画像がない場合に、画像照合処理（Ｓ３０）は、学習データ生成処理（Ｓ１０）のＳ１１０へ移行する。
　ステップ３２５（Ｓ３２５）において、書式定義選択部５０６は、比較画像との一致度が基準を超える比較画像に関連付けられる書式定義を取得する。
　ステップ３３０（Ｓ３３０）において、抽出部５０８は、書式定義選択部５０６により選択された書式定義に基づいて入力画像の文字認識を行う。
　ステップ３３５（Ｓ３３５）において、ユーザは、ＯＣＲ認識結果確認画面において認識結果を確認する。
　ステップ３４０（Ｓ３４０）において、認識されていない文字列が存在する場合に、画像照合処理（Ｓ３０）は、学習データ生成処理（Ｓ１０）のＳ１３０へ移行し、すべて認識されている場合は、処理を終了する。 In step 320 (S320), when the matching degree determination unit 504 determines that there is a comparison image whose matching degree with the input image exceeds the reference, the image matching process (S30) proceeds to S325, and the matching degree If there is no comparison image exceeding the standard, the image comparison process (S30) proceeds to S110 of the learning data generation process (S10).
In step 325 (S325), the format definition selection unit 506 acquires a format definition associated with the comparative image whose degree of coincidence with the comparative image exceeds the reference.
In step 330 (S330), the extraction unit 508 performs character recognition of the input image based on the format definition selected by the format definition selection unit 506.
In step 335 (S335), the user checks the recognition result on the OCR recognition result check screen.
In step 340 (S340), if there is any unrecognized character string, the image matching process (S30) shifts to S130 of the learning data generation process (S10). finish.
　以上説明したように、本実施形態の画像照合システム１によれば、一つの書類に対して複数のパターンの比較画像が生成されるため、オリジナルの画像データと軽微な違いがある入力画像であっても、ユーザがその都度文字の認識範囲を補正することなく、複数パターンの比較画像のいずれかに一致することで書式定義を特定できる。すなわち、文字認識処理の作業効率、照合性能、及び文字認識の照合率が高くなる。
　また、複数のパターンの比較画像を生成する場合に、ランダムに照合しない領域を作成するため、書類毎に照合しない領域が異なり、比較画像のパターンが定型化しない。
　そして、入力画像に適する学習データが存在しない場合でも、ユーザによる比較画像の補正操作を認識し、補正情報に基づいて、新たに学習データを生成し、管理するため、書式定義のメンテナンスが不要となる。
　さらに、スキャナ３の機種が変更されたことにより、スキャナの特性が変わり、これまでの書式定義が使用できない場合でも、画像照合装置５によれば、学習により新たな書式定義を生成するため、新規にユーザによる書式定義を作成する必要はない。 As described above, according to the image collation system 1 of the present embodiment, since a plurality of patterns of comparison images are generated for one document, the input image is slightly different from the original image data. However, the user can specify the format definition by matching any one of the plurality of comparison images without correcting the character recognition range each time. That is, the work efficiency of the character recognition processing, the collation performance, and the collation rate of the character recognition are increased.
In addition, when a comparative image of a plurality of patterns is generated, an area that is not collated is created at random, so that the area that is not collated differs for each document, and the pattern of the comparative image is not fixed.
Then, even when there is no learning data suitable for the input image, the user recognizes the correction operation of the comparison image by the user and generates and manages new learning data based on the correction information, so that maintenance of the format definition is unnecessary. Become.
Further, even if the scanner characteristics are changed due to the change of the model of the scanner 3 and the former format definition cannot be used, the image collating device 5 generates a new format definition by learning. There is no need to create a user-defined format definition.
　上記実施形態では、比較画像生成部５１２により作成された５段階の学習データと入力画像とを比較していたが、一つの書類に関連付けられるパターン１の比較画像を変更してもよい。
　具体的には、変形例における画像照合装置５は、図５に例示する機能構成に加え、除外領域変更部５２０を有する。比較画像生成部５１２は、書類毎にパターン１の照合しない領域をランダムに作成するが、除外領域変更部５２０は、既に作成されたパターン1の照合領域を変更する。具体的には、除外領域変更部５２０は、比較対象から除外される画像領域について、画像領域の数、画像領域の大きさ、及び画像領域の位置のうち、少なくとも一つを変更する。例えば、比較画像生成部５１２により、一つの書類に対して一つのパターン１の比較画像が生成されており、管理されていた場合、パターン１の比較画像と入力画像を照合する際に、照合しない領域が固定されているため、照合率の高い書類と低い書類とが出てくるが、除外領域変更部５２０により、既に存在するパターン１の照合しない領域を変更することにより、照合率の高い書類と低い書類とのバラつきを軽減することが可能となる。 In the above embodiment, the input image is compared with the five-stage learning data created by the comparison image generation unit 512, but the comparison image of the pattern 1 associated with one document may be changed.
Specifically, the image matching device 5 according to the modified example includes an exclusion area changing unit 520 in addition to the functional configuration illustrated in FIG. The comparison image generation unit 512 randomly creates a non-matching area of the pattern 1 for each document, while the exclusion area change unit 520 changes the already created matching area of the pattern 1. Specifically, the exclusion area change unit 520 changes at least one of the number of image areas, the size of the image area, and the position of the image area for the image area excluded from the comparison target. For example, when a comparison image of one pattern 1 is generated for one document by the comparison image generation unit 512 and is managed, when the comparison image of the pattern 1 is compared with the input image, no comparison is performed. Since the area is fixed, a document having a high collation rate and a document having a low collation rate appear. And low documents can be reduced.
　本実施形態では、スキャナ３がスキャンした画像を画像照合装置５に送信して画像照合装置５が入力画像と比較画像とを比較しているが、これに限定されず、例えば、スキャナ３に画像照合プログラム５０がインストールされ、スキャナ３が画像をスキャンし、入力画像と比較画像とを比較してもよい。 In the present embodiment, the image scanned by the scanner 3 is transmitted to the image matching device 5 and the image matching device 5 compares the input image with the comparison image. However, the present invention is not limited to this. The matching program 50 may be installed, the scanner 3 scans the image, and compares the input image with the comparative image.
　１…画像照合システム
　３…スキャナ
　５…画像照合装置
　５０…画像照合プログラム DESCRIPTION OF SYMBOLS 1 ... Image collation system 3 ... Scanner 5 ... Image collation device 50 ... Image collation program

Claims (9)

  1. 　同一の書類の画像データに対して、互いに異なる領域に加工処理を施して生成された複数の比較画像を互いに関連付けて格納する比較画像格納部と、
    　前記比較画像格納部に格納されている比較画像それぞれと、新たに入力された入力画像とを比較して、一致度を判定する一致度判定部と
    　を有する画像照合装置。     A comparison image storage unit that stores a plurality of comparison images generated by performing processing on different areas for image data of the same document in association with each other;
    An image collating apparatus comprising: a comparison image stored in the comparison image storage unit; and a coincidence determining unit that determines a coincidence by comparing a newly input image with the input image.
  2. 　前記比較画像格納部は、書類の書式を定義する書式定義情報を、前記比較画像に関連付けて格納し、
    　前記比較画像格納部に格納されている書式定義情報の中から、前記一致度判定部による判定結果に基づいて、適用する書式定義情報を選択する書式定義選択部と、
    　前記書式定義選択部により選択された書式定義情報に基づいて、新たに入力された入力画像から情報を抽出する抽出部と
    　をさらに有する請求項１に記載の画像照合装置。     The comparison image storage unit stores format definition information defining a format of a document in association with the comparison image,
    From the format definition information stored in the comparison image storage unit, based on a determination result by the matching degree determination unit, a format definition selection unit that selects format definition information to be applied,
    The image collating apparatus according to claim 1, further comprising: an extracting unit configured to extract information from a newly input image based on the format definition information selected by the format definition selecting unit.
  3. 　前記一致度判定部により判定された一致度が、いずれの比較画像についても基準以下であった場合に、前記入力画像に対して、互いに異なる領域に加工処理を施して、複数の比較画像を生成する比較画像生成部と、
    　前記比較画像生成部により生成された複数の比較画像を、前記比較画像格納部に追加登録する比較画像登録部と
    　をさらに有する請求項１に記載の画像照合装置。     When the matching degree determined by the matching degree determination unit is equal to or less than a reference for any of the comparison images, a processing process is performed on the input image in areas different from each other to generate a plurality of comparison images. A comparison image generation unit,
    The image matching device according to claim 1, further comprising: a comparison image registration unit that additionally registers a plurality of comparison images generated by the comparison image generation unit in the comparison image storage unit.
  4. 　前記比較画像生成部は、同一の入力画像に対して、生成される複数の比較画像が互いに異なるデータサイズとなるような加工処理を施す
    　請求項３に記載の画像照合装置。     The image comparison device according to claim 3, wherein the comparison image generation unit performs processing on the same input image so that a plurality of generated comparison images have different data sizes.
  5. 　前記比較画像生成部は、同一の入力画像に対して、互いに異なる領域の画像を削除して、複数の比較画像を生成する
    　請求項３に記載の画像照合装置。     The image comparison device according to claim 3, wherein the comparison image generation unit generates a plurality of comparison images by deleting images in different regions from the same input image.
  6. 　前記比較画像格納部は、前記入力画像と同一の比較画像、前記入力画像の任意の領域が削除された比較画像、前記入力画像から罫線枠内が削除された比較画像、前記入力画像から罫線枠外のみが抽出された比較画像、及び、前記入力画像に含まれる罫線のみを抽出した比較画像、のうち、少なくとも２つを格納している
    　請求項１に記載の画像照合装置。     The comparison image storage unit includes: a comparison image that is the same as the input image; a comparison image in which an arbitrary region of the input image is deleted; a comparison image in which the inside of a ruled line frame is deleted from the input image; The image matching device according to claim 1, wherein at least two of the comparison image from which only the extracted image is extracted and the comparison image from which only the ruled line included in the input image is extracted are stored.
  7. 　比較対象から除外される画像領域について、画像領域の数、画像領域の大きさ、及び、画像領域の位置のうち、少なくとも一つを変更する除外領域変更部
    　をさらに有し、
    　前記一致度判定部は、前記比較画像の少なくとも一つについて、前記除外領域変更部により変更された画像領域を比較対象から除外して、前記入力画像と前記比較画像とを比較して一致度を判定する
    　請求項１に記載の画像照合装置。     The image area excluded from the comparison target, further includes an exclusion area change unit that changes at least one of the number of image areas, the size of the image area, and the position of the image area,
    The coincidence determining unit, for at least one of the comparative images, excludes the image area changed by the exclusion area changing unit from the comparison target, compares the input image with the comparative image, and determines the coincidence. The image collation device according to claim 1.
  8. 　同一の書類の画像データに対して、互いに異なる領域に加工処理を施して生成された複数の比較画像を生成するステップと、
    　前記生成された複数の比較画像を互いに関連付けてデータベースに登録するステップと、
    　前記データベースに登録された比較画像それぞれと、新たに入力された入力画像とを比較して、一致度を判定するステップと
    　を有する画像照合方法。     Generating a plurality of comparative images generated by performing processing on different regions with respect to image data of the same document;
    Registering the plurality of generated comparative images in a database in association with each other,
    Comparing each of the comparison images registered in the database with the newly input image to determine the degree of coincidence.
  9. 　同一の書類の画像データに対して、互いに異なる領域に加工処理を施して生成された複数の比較画像を生成するステップと、
    　前記生成された複数の比較画像を互いに関連付けてデータベースに登録するステップと、
    　前記データベースに登録された比較画像それぞれと、新たに入力された入力画像とを比較して、一致度を判定するステップと
    　をコンピュータに実行させるプログラム。     Generating a plurality of comparative images generated by performing processing on different regions with respect to image data of the same document;
    Registering the plurality of generated comparative images in a database in association with each other,
    Comparing each of the comparison images registered in the database with a newly input image to determine the degree of coincidence.
PCT/JP2018/032358 2018-08-31 2018-08-31 Image comparison device, image comparison method, and program WO2020044537A1 (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/JP2018/032358 WO2020044537A1 (en) 2018-08-31 2018-08-31 Image comparison device, image comparison method, and program
JP2020539985A JPWO2020044537A1 (en) 2018-08-31 2018-08-31 Image matching device, image matching method, and program

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/JP2018/032358 WO2020044537A1 (en) 2018-08-31 2018-08-31 Image comparison device, image comparison method, and program

Publications (1)

Publication Number Publication Date
WO2020044537A1 true WO2020044537A1 (en) 2020-03-05

Family

ID=69643204

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2018/032358 WO2020044537A1 (en) 2018-08-31 2018-08-31 Image comparison device, image comparison method, and program

Country Status (2)

Country Link
JP (1) JPWO2020044537A1 (en)
WO (1) WO2020044537A1 (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2021103581A (en) * 2020-06-12 2021-07-15 北京百度網訊科技有限公司 Method for extracting form in image, device for extracting form in image, electronic apparatus, computer readable storage medium, method for training form extraction module, and computer program

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH10222587A (en) * 1997-02-07 1998-08-21 Glory Ltd Method and device for automatically discriminating slip or the like
JPH11296676A (en) * 1998-04-08 1999-10-29 Oki Electric Ind Co Ltd Image data classification method and image data registration method
JP2002358521A (en) * 2001-05-31 2002-12-13 Oki Electric Ind Co Ltd Device, method and program for registering and identifying document format
JP2005242786A (en) * 2004-02-27 2005-09-08 Oki Electric Ind Co Ltd Form identification apparatus and form identification method
JP2006127451A (en) * 2004-09-30 2006-05-18 Oki Electric Ind Co Ltd Form processor
JP2009011874A (en) * 2007-06-29 2009-01-22 Hitachi Computer Peripherals Co Ltd Business form sorting method, and optical character reading system using the same
JP2010003155A (en) * 2008-06-20 2010-01-07 Fujitsu Frontech Ltd Form recognition apparatus, method, database generation apparatus, method, and program

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH10222587A (en) * 1997-02-07 1998-08-21 Glory Ltd Method and device for automatically discriminating slip or the like
JPH11296676A (en) * 1998-04-08 1999-10-29 Oki Electric Ind Co Ltd Image data classification method and image data registration method
JP2002358521A (en) * 2001-05-31 2002-12-13 Oki Electric Ind Co Ltd Device, method and program for registering and identifying document format
JP2005242786A (en) * 2004-02-27 2005-09-08 Oki Electric Ind Co Ltd Form identification apparatus and form identification method
JP2006127451A (en) * 2004-09-30 2006-05-18 Oki Electric Ind Co Ltd Form processor
JP2009011874A (en) * 2007-06-29 2009-01-22 Hitachi Computer Peripherals Co Ltd Business form sorting method, and optical character reading system using the same
JP2010003155A (en) * 2008-06-20 2010-01-07 Fujitsu Frontech Ltd Form recognition apparatus, method, database generation apparatus, method, and program

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2021103581A (en) * 2020-06-12 2021-07-15 北京百度網訊科技有限公司 Method for extracting form in image, device for extracting form in image, electronic apparatus, computer readable storage medium, method for training form extraction module, and computer program
JP7278321B2 (en) 2020-06-12 2023-05-19 ベイジン バイドゥ ネットコム サイエンス テクノロジー カンパニー リミテッド Method for extracting tables in images, apparatus for extracting tables in images, electronic devices, computer readable storage media, methods and computer programs for training table extraction modules

Also Published As

Publication number Publication date
JPWO2020044537A1 (en) 2021-03-18

Similar Documents

Publication Publication Date Title
US20210192202A1 (en) Recognizing text in image data
JP7013182B2 (en) Information processing equipment, information processing methods and programs
US8108764B2 (en) Document recognition using static and variable strings to create a document signature
JP6357621B1 (en) Accounting processing apparatus, accounting processing system, accounting processing method and program
JP2016048444A (en) Document identification program, document identification device, document identification system, and document identification method
CN111310426A (en) Form format recovery method and device based on OCR and storage medium
US11023720B1 (en) Document parsing using multistage machine learning
EP2884425B1 (en) Method and system of extracting structured data from a document
JP6435934B2 (en) Document image processing program, image processing apparatus and character recognition apparatus using the program
JP6579456B1 (en) Search target information narrowing system
WO2020044537A1 (en) Image comparison device, image comparison method, and program
JP5623574B2 (en) Form identification device and form identification method
JP2004013813A (en) Information management system and method
CN113591657B (en) OCR layout recognition method and device, electronic equipment and medium
JP2020087112A (en) Document processing apparatus and document processing method
US11972208B2 (en) Information processing device and information processing method
CN114529933A (en) Contract data difference comparison method, device, equipment and medium
JP5169648B2 (en) Original image search device and original image search program
JP2010237909A (en) Knowledge correction program, knowledge correcting device and knowledge correction method
JP2021064123A (en) Data input support system, data input support method, and program
JP2020047031A (en) Document retrieval device, document retrieval system and program
JP5272664B2 (en) Information processing apparatus, image search method, and program
JP2012146147A (en) Document processing device and document processing program
CN114120016B (en) Character string extraction method, device, equipment and storage medium
US20230273952A1 (en) Image processing apparatus, image processing method, and storage medium

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18932142

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2020539985

Country of ref document: JP

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18932142

Country of ref document: EP

Kind code of ref document: A1