JP6250307B2 - Image information processing apparatus and image information processing method - Google Patents

Description

  The present invention relates to an image information processing apparatus and an image information processing method capable of registering image images and searching for registered image images, for example, registering a form as an image image, and flexibly dealing with future specification changes. The present invention relates to a searchable image information processing apparatus and an image processing method.

  Along with the diversification of operations in recent years, various types of documents, slips, forms, and other forms are being handled in various industries, and necessary information in the information described in the processing target documents is visually checked. After confirming and inputting keys, it is converted into character information, and the input form is stored in a state where it can be easily taken out when it becomes necessary to verify the input result after that (the entire text of the input information is confirmed).

  Since then, with the improvement of recognition technology, a system has been introduced in which forms are read by an OCR device or the like, and necessary information is recognized and converted into digital information. Then, whether or not the digitalized information is correctly recognized is confirmed separately by visually checking the original form, and if there is a recognition error, it is corrected here.

  When the digitization is completed, the original form is stored in a warehouse or the like until the storage period expires. Even in the above processing, the input form must be stored in a state where it can be easily taken out when it becomes necessary to verify the input result thereafter (confirming the full text of the input information).

  Since then, the capacity of the storage device has been further increased, and the processing speed can be stored as an image of the front and back surfaces of the original forms. For this reason, the original forms are imaged as they are, taken in as they are, and a partial area of this image information is recognized and converted into character data, for example, and the read image information is also displayed whether or not the recognition processing has been performed correctly. By confirming the display information, it is possible to eliminate the need for referring to the actual form that has been converted into image information.

  In this type of device, when registering the read image information in a predetermined file, a predetermined search keyword is set so that the image information can be read out, and an image can be searched with the set keyword. It is desirable to configure. However, conventionally, there has been no document in which forms are converted into image data and registered as image information, and the registered image is referred to by a plurality of processing systems.

  A number of systems for searching text data (document data) are proposed instead of searching only image data by a plurality of systems. For example, in Patent Document 1, it is preferable that a keyword corresponding to the purpose of a system for registering a document is picked up and limited from a dictionary, but this is a technique that uses data in a dictionary as it is. .

Japanese Patent Laid-Open No. 2003-186884

  However, the keyword must be based on the initial purpose, and if the purpose is different, the keyword will be different. If it is necessary to perform processing with a different purpose, the keyword is determined again and reset. Had to do.

  Even if the processing target is a form, it is necessary to re-select the keyword item from the beginning if the processing form is different. If the processing target is different, it is necessary to redesign the processing system each time there were.

  For example, if you are going to use the payment image search system for tax payments to search for direct debit requests, the keys for searching payment images are the bank storage date, storage store, storage city, company name , And the amount of money. On the other hand, in searching for an account transfer request image, the store number, subject, account number (hereinafter abbreviated as account number), registration date, consignment company code, and the like are the keys. As described above, since the key items are completely different, it is unavoidable to have two completely different search systems.

  As described above, in the conventional processing system, the business menu to be processed and the search program for each business are a pair, and it is necessary to change the system and the search program when adding business / additional forms. The same applies when registering image data.

  In order to prevent this, it is necessary to identify in advance all the keys of work that may be added in the future, but this is impossible in reality.

  In addition, once you have processed the payment slip, it is unlikely that you will need to repeatedly refer to the original form in the future, but it is highly likely that you will repeatedly refer to the bank loan application documents and permit approval documents. In many cases, you may want to check the information in different areas. For this reason, it is desirable that the entire text can be a search target as a search key for image information.

  The present invention has been made for the purpose of solving the above-described problems, and an image information processing apparatus and an image information processing method that do not require a search program to be changed even when there are changes in the forms to be processed and business operations. Is to provide.

  For example, the following configuration is provided as a means for achieving the object. That is, an image information processing apparatus for storing various document images related to various jobs, the image storage unit storing the document image in a readable manner, and the document image registered in the image storage unit In the digital information corresponding to the character information, a character information index means that uses a character string for each word of the character information as an index corresponding to the document image, and a unique document type ID for each type of the document image, Document type index means for associating with an index by the index specifying means, and index storage means for storing the index and the document type ID as read keywords for the document image registered in the image storage means.

  And, for example, an image for receiving the index and the document type ID stored in the index storage means and reading out the document image stored in the image storage means specified by all of the received index and document identification ID And a reading means.

  Also, for example, the document image is a form image, and when adding a document type to be processed, the document type index unit assigns a unique document type ID to the newly added document type, and the character information index unit includes: This can be handled by additionally registering a character string corresponding to the character information displayed in the added document to be processed as an index in the index storage means. Further, for example, the character information index means is characterized in that the entire text displayed in the processing target document is an index target.

  Alternatively, an image information processing method in an image information processing apparatus including document image storage means for storing various document images relating to various business operations and index storage means for storing index information for the document images stored in the document image storage means And recognizing the character information displayed in the document image stored in the document image storage means, extracting all character strings for each word as an index corresponding to the document image, A unique document type ID is assigned to each type, is associated with the extracted index, and is stored in the index storage unit as index information for the document image.

  And, for example, the index and the document type ID stored in the index storage means are received, and the document image stored in the image storage means specified by all of the received index and document identification ID is read. It is characterized by.

  Also, for example, the document image is a form image, and when a processing target document type is added, a unique document type ID is given to the newly added document type, and the character information displayed on the added processing target document It is possible to cope with this by additionally registering a character string corresponding to the above as an index in the index storage means.

  According to the present invention, it is possible to provide an image information processing apparatus and an image information processing method capable of searching for a desired target document without the need to newly create a search program even if the target document to be processed is added or changed.

1 is a block diagram illustrating an overall configuration of an image information processing apparatus according to an embodiment of the present invention. It is a figure which shows the example of the processing form which concerns on the example of this Embodiment.

It is a figure for demonstrating the example of generation | occurrence | production of the index of the form shown in FIG. It is a figure for demonstrating the example of the process target form of the image information processing apparatus which concerns on the example of this embodiment, the example of a form display character data file, and the production | generation example of an image storage file.

It is a figure for demonstrating the example of generation | occurrence | production of the path | pass to the index generation file, index file, image file, etc. of the image information processing apparatus concerning this Embodiment. It is a figure for demonstrating the relationship between the file for index production | generation of an image information processing apparatus which concerns on this Example, and an image reference index file.

It is a figure for demonstrating the image search process of the image information processing apparatus which concerns on the example of this embodiment.

  Embodiments of the present invention will be described in detail below with reference to the accompanying drawings. In the present embodiment, an image information processing apparatus and an image information processing method capable of performing a substantially full-text search corresponding to imaged information are provided.

  In this embodiment, it is not necessary to design a search system for each search job by enabling the setting of a free search key that does not have the concept of an individual job, and search for skewering between different jobs. It has a feature in that it can be done.

  In the following description of the present embodiment, an example in which an apparatus and a processing method optimal for a system for searching for forms to be processed by a financial institution will be described. Furthermore, an information processing system that does not cause a decrease in performance even after long-term accumulation of images is provided. However, it is not limited to processing forms of financial institutions, and there is no limitation on the type of image to be registered and searched.

  First, a schematic configuration of the image information processing apparatus in the present embodiment will be described with reference to FIG. FIG. 1 is a block diagram showing the overall configuration of an image information processing apparatus according to an embodiment of the present invention.

  In FIG. 1, reference numeral 10 denotes an operation terminal for searching registered image information. 110 to 130 are systems for processing financial institution forms, 110 is an account transfer system for performing an account transfer process, 120 is a deposit / withdrawal slip processing system for processing a deposit / withdrawal slip, and 130 is a process for financing a corporation. It is a corporate loan processing system. Each system does not have to be provided individually, but may have one configuration common to all processes.

  Each system has at least 150 configurations. That is, an image reader 151 that reads an image of a document to be processed, a data recognition unit 152 that recognizes necessary character and numeral information from an image read by the image reader 151, and converts the information into digital data, and data that is recognized by the data recognition unit 152 A keyword extraction unit 153 that extracts a keyword for searching the image information, outputs the read image information and the extracted keyword to the document image storage / retrieval apparatus 200, and calibrates the recognition result of the data recognition unit 152. An input / output unit 155 and the like are included.

  In these processing systems 150, the image reader 151 reads the front / back display information of the processing target form as image information. Then, the character / numeric data displayed in the predetermined area in the read image data is character-recognized by the data recognition unit 152, for example, both the recognized data and the corresponding read image are displayed in parallel, and the recognition result Perform verification calibration.

  For example, in an image file 230 that stores various document images related to various tasks in a readable manner, and digital information corresponding to character information displayed in the document image registered in the image file 230, for each word of the character information A character string is used as an index corresponding to the document image, a unique document type ID is assigned to each type of the document image, the index is associated with the index, and the index and the document type ID are registered in the image file 230. The search engine 250 receives the index and the document type ID, and is identified by all of the received index and document identification ID with reference to the index file 300. Characterized in that it comprises a image reading means for reading a document image stored in a file 230.

  After confirming that the recognition result is correct, the keyword extracted by the keyword extraction method described later and the read image information are sent to the document image storage / retrieval apparatus 200 through the input / output unit 155.

  Reference numeral 200 denotes a document image storage / retrieval device that stores document images read by each processing system, and can search and retrieve the stored document images as necessary. A search engine 250 that searches for image files registered in the index file, and an indexer 260 that reads a desired index (search keyword) from the index file 300 and supplies the index to the search engine 250.

  Reference numeral 300 denotes an index file 300 that registers and holds keyword information extracted by the keyword extraction unit 153.

  The image of the handling document in the financial institution which is the processing target of the present embodiment having the above configuration is difficult to delete in practice. It is legally required to store for 7 years or 10 years, and the number of sheets to be stored becomes very large.

  For example, deposit / withdrawal slips are about 20,000 per day even for small regional banks, and more than 5 million per year. Megabanks are expected to have more than 50 million copies per year. For this reason, it is not necessary to check each form by making these documents into an image so that they can be searched. You can simply store them in a warehouse, etc., and even after discarding forms. If saved as image information, little space is required.

  However, only a relatively simple index can be used to maintain the performance of processing for searching for a desired form image in an environment for processing a large amount of form images as described above. In particular, a search for an unindexed data item cannot expect a response.

  In this embodiment, from the above, the form reading image information obtained in each individual business process is used, and the original form is stored and saved in a warehouse or the like, and all of the code data is saved. The data item name is added to the full-text search index as a key, and the data item name and the entity are searched by AND (AND) at the time of the search. As a result, a general-purpose image search system that does not depend on applications can be realized.

The outline of the search method according to the present embodiment will be described below using a specific example.
For example, when a bigram index is created with a character string “Today is good weather”, the ID of the target document is first assigned, for example, number 1. Subsequently, the following index information is generated.

"Today 1"
"Day is 1"
"Harah 1"
"Good one"
"Iten 1"
"Weather 1"
“Care is 1”
"Is 1"
"Su. 1"
". 1"

Similarly, the ID of the document “Today is heavy rain” is set to No. 2, and the next search information (bigram information) is generated.
"Today 2"
"Day is 2"
"Haha 2"
"Heavy rain 2"
"In the rain 2"
"It is 2"
"Su. 2"
". 2"

In the present embodiment, the index information obtained by adding the document IDs No. 1 and No. 2 is assigned as follows.
"Today 1, 2"
“Day is 1, 2”
"Harah 1"
"Good one"
"Iten 1"
"Weather 1"
“Care is 1”
"Is 1, 2"
"Su. 1, 2"
". 1, 2"
"Haha 2"
"Heavy rain 2"
"In the rain 2"

  This is a transposed index by bigram in the present embodiment. In order to search for the character string “weather” using the index information created in this way, the document whose document ID is No. 1 corresponds. In addition, when the character string “today” is searched, the sentences with the document IDs 1 and 2 correspond.

  When searching for a document including two character strings “today” and “heavy rain”, the search result of “today” and the search result of “heavy rain” are combined with a logical product (AND) condition. The result of No. 2 is obtained. The above search method is the basis of the search method of this embodiment.

  Next, a specific example applied to an actual form image will be described next. In this embodiment, a common image reference base is used for each business. More specifically, when managing the image of the tax payment document image and the account transfer request document shown in FIG. 2, the processing for searching the image for display contents is as follows.

  In order to index and include key attributes (names) in the index, for example, in order to use a common image reference platform for reference to tax payment payment image and fund transfer request, it is performed as follows.

In a system for registering a payment slip (tax public money) and performing an image search process, the following items displayed are used as search keywords.
For example, as the code data, the bank storage date, the storage store, the storage municipality / company name, the amount of money, and the like shown on the payment slip are obtained.

The index of the data shown in FIG. 2 when creating the system is information indicating, for example, a payment notice (hereinafter abbreviated as “done”). The document ID is the following information etc. that are displayed in a completed manner.
Bank storage date, for example “2013/01/15”,
Storage store. For example, “102”,
Storage city name `` ○○ city '',
Amount of money, for example "40,000 yen"
Each of the above is regarded as one word, and a full-text search index is generated. The document ID portion at this time associates (links) the image storage destination path.

In the example of registering an account transfer request form and performing an image search process, the index is, for example, information indicating that it is an account transfer request form obtained in the business process of registering the account transfer request form. Information,
The document ID includes the following information displayed in the account transfer request form.

Registration date such as “2013/01/15”,
Account number, for example, “4719220205264”,
Consignment company code, for example “0264”
Consigned company name “□□ Corporation”
A full-text search index is generated with each as one word. Note that the document ID portion at this time associates (links) the image storage destination path.

FIG. 3 shows the relationship between the full-text search index and the image in the above case.
The present embodiment has a simple configuration in which only the relationship between logical words and images is shown, and the structure of a physical index table includes an image index and an index as shown in FIG. The pointer to the image from which the pointer information to the corresponding image can be obtained is very simple.

  When the full text search is simply used, for example, when searching with a search keyword including a number, date, name and address, the following inconvenience occurs. For example, if the account number “4719220205264” is placed in the index and a search for “471922” is performed, all items including “471922” are hit. In business, there are very many things that use numeric numbers as keys.

  For example, even if it is limited to banking operations, there are CIF numbers, project numbers, proposal numbers, and reception numbers that are similar to account numbers. Similarly, the date includes a reception date, a processing date, a registration date, a response date, a receipt date, an execution date, a cancellation date, an issue date, and the like. Even when an address is used as a keyword, there are a principal address, a work address, a guarantor address, a property address, etc., and in order to perform an efficient search, it is necessary to perform a search separately.

  For this purpose, it is necessary to inform the system that the data has the same attribute using XML or the like, but informing the system that the data has the same attribute is essentially defining and structuring data items. It becomes difficult (weak) to respond to the addition of business.

  For this reason, in the present embodiment, the data item name is added to the full-text search index, and the data item name and the entity index are searched by logical product (AND) at the time of the search. As a result, when a process for another business that is not scheduled at the time of previous registration is added, the business can be added without changing the reference business program.

  For example, the payment notice shown in FIG. 2 is read by the image reader 151 of the processing system 150, recognized by the data recognition unit 152, and extracted by the keyword extraction unit 163. A specific example of the image file (image file 230) is shown in FIG.

  FIG. 4 shows the data item names and contents “XX city”, “40,000 yen (storage tax amount)”, “102 (storage bank store code)”, “2013/01” from the payment notice shown in FIG. / 05 (bank storage date) ”and“ image path ”which is the storage coordinates of the image file are generated and stored in the index file 300. FIG.

  Subsequently, the processing system 150 creates an index generation file in which a form name and a data item name are added to the generated payment notice processing result file. The left side of FIG. 5 shows an index generation file in which a form name and a data item name are added to the paid notice processing result file shown in FIG. FIG. 5 shows an example in which “123” is assigned as the form ID of the paid notice.

  The index generation file is sent to the document image storage / retrieval apparatus 200, and the logical word stored in the processing result image file 230 is used as an index. The index file 300 has a form ID 123 file and the path information to the image file for each image ( (Stored coordinate information).

  The above explanation has been made with respect to the payment notice. However, in the account transfer request document, “trusted company name: □□ corporation” “account number: 47 1922202020264” “account holder / Similarly, a processing result file and an index generation file are created for the “seal” (storage city name, storage store). Then, as shown in FIG. 6, an index file is generated by adding both indexes to the combined or paid notice index.

  That is, it is only necessary to add or synthesize the contents of the index generation file of the form to be processed in common with the existing form index. For example, as shown in FIG. 3, even if there is an addition or change in the processing target form, it is only necessary to add a newly added index, and it is only necessary to specify an index (search key) during the search. .

  That is, in the example of FIG. 3, compared to the case where “consignment company code 0264” is set as an index, an index is conventionally assigned to the item “consignment company”, and subsequently an index “0264” is assigned below. If this “consignment company code” is not an item in the account transfer request processing system, it is necessary to provide a new item “consignment company”. What is an account transfer request processing system? A completely different system design was required. In the present embodiment, such a system reconstruction is unnecessary.

  In this embodiment, the purpose is to search for image information, and it is not necessary to extract fine digital data. Conventionally, a search method specific to such an image search system has been mostly noticed. It has not been done and is unique to the present invention.

  As described above, in this embodiment, in order to use a common image reference platform for each business, each logical word displayed in the form is treated as an index, and the relationship between the logical word and the image is associated. Therefore, even if other forms or items are added, it is only necessary to add and synthesize, and it is very easy to cope with changes in the specifications of the image information registration / search system.

Processing for searching for image information registered in the image file 230 as described above using the index file 300 will be described below with reference to FIG.
When searching (referring to) image information, a search process is performed by applying a logical sum (AND) condition to the form name (for example, paid notice or account transfer request form) and the data item name. For example, when a form ID is specified in the index file 300, an image file corresponding to the form ID is specified from the image file 230, and then an index for specifying the form is specified.

  For example, as shown in FIG. 7, when “stored” is specified as “search city” as the search key from the operation terminal 10, the search engine 250 and the indexer 260 access the index file 300 and form ID “123”. Get. Thereby, it is specified that the paid notice image file of the image file 230 is the target image file.

  If the number of images is not so large, the image files may be sequentially confirmed to specify the desired image. However, if the number of images is large, any or all of indexes such as “102 (storage bank store code)”, “2013/01/05 (bank storage date)”, “40,000 yen (storage tax amount)” It is possible to obtain a desired image by designating.

  As described above, according to this embodiment, even when a document type to be processed is added, a document type ID is simply assigned and an index is set from the character string displayed on the document to be processed. By simply extracting and registering in the index file 300, it is possible to easily handle the document image registered in the image file 230. In this case, there is no need to create a new search program. This is realized simply by registering a character string displayed in a document newly registered in the image file 230 in the index file 300 as an index for search.

  The above explanation mainly explained the account transfer request processing and payment notice processing in the account transfer system 110, but an image of generating a necessary index by imaging the deposit / withdrawal slip processing system 120. Even in the registration process to the file 230, the entire text of the character string displayed on the slip is registered as an index in the index file 300, a unique document ID is assigned to each slip type, and the index is similarly associated with the index. It will be registered in the file 300. The same applies to the corporate loan processing system 130.

Claims (7)

An image information processing apparatus for storing various document images related to various tasks,
Image storage means for storing the document image in a readable manner;
In the digital information corresponding to the character information displayed in the document image registered in the image storage means, the character information indexing means that uses a character string for each word of the character information as an index corresponding to the document image;
A document type index unit that assigns a unique document type ID to each type of the document image and associates the index with the character information index unit;
Index storage means for storing the index and the document type ID as read keywords for the document image registered in the image storage means ,
An image information processing apparatus capable of reading out the document image stored in the image storage unit by using the index stored in the index storage unit as a read keyword . Image reading means for receiving the index and the document type ID stored in the index storage means and reading out the document image stored in the image storage means specified by all of the received index and document identification ID 2. The image information processing apparatus according to claim 1, further comprising: The document image is a form image. When a document type to be processed is added, the document type index unit assigns a unique document type ID to the newly added document type, and the character information index unit is added. 3. The image information processing apparatus according to claim 1, wherein the image information processing apparatus can cope with the problem by additionally registering a character string corresponding to the character information displayed in the document to be processed as an index in the index storage unit. . 4. The image information processing apparatus according to claim 1, wherein the character information indexing unit sets the entire text displayed in the processing target document as an index target. An image information processing method in an image information processing apparatus comprising document image storage means for storing various document images relating to various business operations and index storage means for storing index information for the document images stored in the document image storage means. And
Recognizing the character information displayed in the document image stored in the document image storage means and extracting all character strings for each word as an index corresponding to the document image;
A unique document type ID is assigned to each type of the document image, associated with the extracted index, and stored in the index storage unit as index information for the document image,
Image information processing method comprising the next possible Rukoto reads the document image stored in said image memory means as a keyword reading the index stored in the index storage unit. Receiving the index and the document type ID stored in the index storage means, and reading the document image stored in the image storage means specified by all of the received index and document identification ID; The image information processing method according to claim 5. The document image is a form image, and when a processing target document type is added, a unique document type ID is assigned to the newly added document type, and corresponds to the character information displayed on the added processing target document. 7. The image information processing method according to claim 5, wherein the image information processing method can cope by additionally registering a character string as an index in the index storage unit.

Images