JP2007004393A - Document search device and document search method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a document search device and its method for more surely searching out a read document. <P>SOLUTION: This document search device is provided with extraction parts 4, 6, and 8, which individually extract a UUID, a characteristic amount and a search word from read image, and search parts 10, 12, and 14 searching and extracting respective corresponding databases 20, 22, and 24 by using the extracted UUID, characteristic amount, and the search word as keys. When the read image is formed by scanning a document, an extraction processing control part 2 carries out combined search processing based on the extracted UUID, characteristic amount, and the search word. A search result evaluation part 16 displays document information specified by the search result on a screen of a display part 18. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an apparatus and method for searching for a document that matches input image data from a database in which information about the document is stored.

  In recent years, the handling of information such as strengthening information security, compliance or protection of personal information in companies has been regarded as important. For example, the company needs to disclose what information the business has been performed based on, for example, an audit. For this purpose, it is necessary to log the access status of information so that it is possible to identify when and who processed what information from where.

  By the way, in order to strengthen security management of confidential information that has been conventionally handled by paper media, a UUID (Universal Unique Identifier) is assigned to a medium such as printing paper, and the UUID is printed on the medium with a barcode or an IC tag. There is technology to embed or embed. If this UUID is used, it is possible to identify when the confidential document at hand has created the confidential document by searching the management information regarding the medium. Therefore, even if the confidential document at hand is a document that has been copied illegally or without permission, it is possible to find the creation source of the confidential document.

  As described above, the document creation source can be easily specified by using the UUID. However, in addition to the technique using the UUID, the feature amount is obtained from the image data of the document as data used when specifying the document. There is a technique for calculating and specifying original image data based on the feature amount (for example, Patent Documents 1 to 4).

JP 2004-139210 A Japanese Patent Laid-Open No. 9-270902 JP 2003-281176 A Japanese Patent Laid-Open No. 10-49659

  However, conventionally, there are cases where the UUID cannot be specified due to being erased or the paper becoming dirty. In such a case, the document creation source cannot be specified. Also, the input of UUID is troublesome. In addition, only the technology that uses the feature amount has a difficulty in specifying the document creation source.

  The present invention has been made to solve the above-described problems, and an object of the present invention is to provide a document search apparatus and method that can more reliably find a read document.

  In order to achieve the above object, a document search apparatus according to the present invention is a document search apparatus for searching for a document that matches input image data from a database in which information on documents is stored. Image data input means for inputting the image data, document feature extraction processing means for extracting a plurality of types of document feature information capable of specifying the document from the image data of the input document, and extraction from the document feature extraction processing means Obtaining means for acquiring selection candidates for the document by searching the database based on each document feature information, and evaluation means for identifying the document by evaluating the selection candidates obtained by the obtaining means; Output means for outputting an evaluation result by the evaluation means.

  The document feature extraction processing unit includes an identification information extraction unit that extracts identification information uniquely assigned to the document as document feature information from the image data, and the acquisition unit assigns the document information uniquely to each document. And an identification information search unit for searching the database in which the registered identification information is registered.

  The document feature extraction processing unit includes a feature amount extraction unit that calculates a feature amount of the document from image data as document identification information, and the acquisition unit stores the database in which the feature amount of each document is registered. It has the feature-value search part to search.

  The document feature extraction processing unit includes a search term extraction unit that extracts a search term used for database search from image data of the document as document identification information, and the acquisition unit includes the database in which each document is registered. It has a document search part for searching.

  Further, the document feature extraction processing means includes an extraction process control unit that controls the order of each process for extracting a plurality of types of document feature information.

  The document search method according to the present invention is implemented in a document search apparatus that searches for a document that matches input image data from a database in which information about the document is stored, and image data input for inputting image data of the document A document feature extraction step for extracting a plurality of types of document feature information capable of specifying the document from image data of the input document, and the database based on each document feature information extracted from the document feature extraction step. An acquisition step of acquiring a selection candidate of the document by searching, an evaluation step of specifying the document by evaluating the selection candidate acquired by the acquisition means, and an output step of outputting an evaluation result by the evaluation step It is characterized by including.

  According to the present invention, it is possible to find out more reliably by combining means for extracting and searching a plurality of types of document feature information.

  Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings.

Embodiment 1 FIG.
FIG. 1 is a functional block configuration diagram showing an embodiment of a document search apparatus according to the present invention. The document search apparatus according to the present embodiment includes an image reading unit 1, an extraction process control unit 2, a UUID extraction unit 4, a feature amount extraction unit 6, a search word extraction unit 8, a UUID search unit 10, a similar image search unit 12, and a document A search unit 14, a search result evaluation unit 16, and a display unit 18 are provided. The image reading unit 1 reads a document on which a document is written. The UUID extraction unit 4 extracts a UUID from the read image using an OCR (optical character reader). The feature amount extraction unit 6 calculates the feature amount of the image from the read image. The extraction process control unit 2 performs process control such as determining the execution order of each extraction process in each extraction unit 6. The search word extraction unit 8 extracts text characters from the read image using OCR, and extracts search words from the text characters. In the present embodiment, a UUID extraction unit 4, a feature amount extraction unit 6, and a search word extraction unit 8 are provided as document feature extraction processing means for extracting document feature information that can identify the document from the read image. , 6 and 8, different types of information such as UUID, feature amount, and search word are extracted as document feature information. Then, UUID search unit 10, similar image search unit 12, and document search unit 14 are provided corresponding to each of extraction units 4, 6, and 8 as acquisition means for acquiring selection candidates for the document by database search. That is, since the UUID assigned to each document is registered in the UUID database 20, the UUID search unit 10 searches the UUID database 20 using the UUID extracted from the read image as a key to find the corresponding document. Identify. In addition, since the feature amount extracted from the read image is registered in the similar image database 22, the similar image search unit 12 searches the similar image database 22 based on the feature amount extracted from the read image. Identify relevant documents. Further, since the document itself is registered in the document database 24, the document search unit 14 specifies the corresponding document by searching the document database 24 using the search word extracted from the read image as a key. Note that, as a result of the search processing in each of the search units 10 to 14, only one corresponding document is not necessarily selected, and a plurality of documents may be specified. The expression of acquiring selection candidates is more appropriate. The search result evaluation unit 16 specifies the document by evaluating the search result obtained by each search unit 12. The display unit 18 extracts information about the document specified according to the evaluation result by the search result evaluation unit 16 from the document information database 26 and displays it on the screen.

  In FIG. 1, although illustrated as a single document search apparatus, in reality, it is assumed to be realized by the network system shown in FIG. That is, FIG. 2 shows a configuration in which the image forming apparatus 30 and the database server 32 are connected by a LAN (local area network) 34, but the databases 20 to 24 shown in FIG. The other configurations are assumed to be mounted on the image forming apparatus 30, respectively. However, the present invention is not limited to the configuration in which the databases 20 to 24 are collectively managed by one database server. In FIG. 2, only one image forming apparatus 30 is shown for convenience, but actually, a plurality of image forming apparatuses 30 having the functions described with reference to FIG. Performs a database search in response to a request from one of the image forming apparatuses 30 and returns the search result to the image forming apparatus 30 that is the request source.

  In addition, the document search apparatus according to the present embodiment is formed inside an image forming apparatus 30 such as a multifunction peripheral except for the databases 20 to 24. Actually, it is realized by a computer mounted on the image forming apparatus 30. The image reading unit 1 operates in cooperation with the scanner, the UUID extraction unit 4 and the search word extraction unit 8 operate in conjunction with the OCR, and the display unit 18 operates in cooperation with the operation panel. Each processing function of each component 2 to 16 is realized by a cooperative operation between a computer and a device such as a scanner mounted on the image forming apparatus 30 and a software program executed by the computer.

  Next, the document search process in the present embodiment will be described with reference to the flowchart shown in FIG.

  For example, a document that has been leaked for some reason is now available and you want to pursue information about this document, for example, who created this document when. Here, for convenience, it is assumed that the document is composed of only one sheet, and a UUID represented by a character string code is printed at a predetermined position (for example, 20 to 30 mm from the lower end of the sheet). It is assumed that The user scans this document using the scanning function of the image forming apparatus. The image reading unit 1 forms a read image by reading the document by this scanning (step 101). Note that the image reading unit 1 includes the case where the read image scanned and generated by another apparatus is acquired via the network without forming the read image itself.

  Next, the UUID extraction unit 4 extracts the UUID from the read image (step 102). Since the UUID is attached to a predetermined area of the image (20 to 30 mm from the lower end of the image), it can be extracted by reading the code in the predetermined area. In this embodiment, since a UUID described in a character string (character mode) is assumed, characters are recognized by OCR. However, for example, when a barcode is used, the UUID is read by a barcode reader. become. In any case, it is sufficient to use UUID extraction means that matches the UUID addition method.

  If the UUID can be normally extracted from the read image, the UUID search unit 10 searches the UUID database 20 using the UUID extracted from the read image as a key (steps 103 and 104). Since the UUID is uniquely assigned to each document, normally, only one document is extracted and specified by this search. Accordingly, the search result evaluation unit 16 in this case does not need to evaluate the search result in particular, and the display unit 18 searches and retrieves the document information database 26 using the UUID specified by the search by the UUID search unit 10 as a key. Information about the document is displayed on the screen (step 112).

  The document information database 26 according to the present embodiment includes information on image data, creation date, creator, and device of the document in association with each UUID. As the image data, the image itself read by the image reading means such as the image reading unit 1 is assumed to be data, but it may be a thumbnail because the amount of data becomes enormous. The creation date and time is time information for specifying the date and time when the document was created. The creator is information for identifying a person who created the document, is a user ID or a user name, and is obtained from user information specified when logging in to a device that creates the document. The device is information for specifying the device on which the document is created, and an IP address is used in the present embodiment. When print data is transmitted to the image forming apparatus via the network and the print processing is performed as in the print function, the IP address of the print data transmission source is also recorded. In the case of the FAX transmission function, the destination IP address is also recorded.

  Therefore, the display unit 18 displays the above-described information related to the document corresponding to the UUID specified by the search by the UUID search unit 10 on the screen. Thus, the user can specify when and by which device the leaked document was printed by whom.

  If the UUID can be normally extracted from the read image, the process is basically performed as described above. However, in the exceptional case, even if it is determined that the UUID can be normally extracted from the read image, an OCR decoding error is detected. For example, it can be assumed that there is no UUID in the UUID database 20 that matches the UUID extracted from the read image. In this case, a plurality of selection candidates are extracted by performing a search using several digits as wild cards, or by performing a search while replacing numbers that are prone to decoding errors, such as 8 and 9, 0, and 6, for example. May be. Alternatively, means for displaying the automatically extracted UUID on the screen and correcting it by the user may be provided. When a plurality of selection candidates are extracted, the search result evaluation unit 16 displays information on each document retrieved by searching the document information database 26 using the extracted UUID as a key (step 112). The user searches for a document that matches the leaked document with reference to the displayed information. In this case, the document search apparatus cannot identify only one document, but can extract selection candidates, and therefore can support the identification of a leaked document creation source.

  On the other hand, when the UUID cannot be normally extracted from the read image intentionally or due to dirt on the paper, the search word extraction unit 8 extracts the text from the read image using OCR (steps 103 and 105). Then, morphological analysis is performed to cut out words from the text and extract valid words (step 106). The term “effective word” as used herein is a generic term for a word having a predetermined number of characters or more, a word having a predetermined appearance frequency or more, and a word having a word existence probability larger than a predetermined value. The word existence probability is “the word recognition rate of N letters is the Nth power of the character recognition rate.” And “If this word appears M times in the document, 1-(( 1- (character recognition rate to the Nth power)) to the Mth power) is defined as the probability of being in the document.

  Here, when the number of extracted effective words can be extracted more than the predetermined n, the document search unit 14 uses the effective words extracted from the read image as search words (keywords) or by a combination of effective words. A search expression is generated (steps 107 and 108). For example, the search expression is ANDed up to the highest i-th effective word when the effective words are arranged in order of appearance frequency, or the number of characters constituting the effective word is in descending order. It is generated according to a rule such as ANDing up to the j-th effective word when arranged. A document is extracted by searching the document database 24 in which all documents are registered by using the generated search word or search expression thus obtained (step 109). If there are no search results, the search terms may be reduced automatically or by user selection and re-created.

  The search result evaluation unit 16 refers to the search result, determines that a document having a higher degree of coincidence with the search word has a higher probability of being a leaked document, and extracts up to the top kth documents having a higher hit rate as selection candidates To do. The display unit 18 displays information on the document retrieved by searching the document information database 26 using the UUID added to the extracted document as a key (step 112). The user can find a document that matches the leaked document from a plurality of candidates by referring to the displayed information and comparing it with the leaked document.

  On the other hand, when the number of extracted effective words cannot be extracted more than the predetermined n (step 107), the process proceeds to document search processing based on the feature amount. This is because it is possible to determine that an accurate result cannot be obtained by document search using a search word because a search word exceeding the necessary number could not be extracted. At this time, the feature amount extraction unit 6 analyzes features such as color information, texture information, and shape information from the read image and extracts feature amounts (step 110). And the similar image search part 12 specifies the document estimated from the extracted feature-value by searching the similar image database 22 based on the extracted feature-value (step 112). Also in this case, the top k-th documents with high similarity are extracted as selection candidates. The display unit 18 displays on the screen information related to the document retrieved by searching the document information database 26 using the UUID added to the extracted similar image as a key (step 112). The user can find a document that matches the leaked document from a plurality of candidates by referring to the displayed information and comparing it with the leaked document.

  According to the present embodiment, information relating to a leaked document can be found more reliably by combining three different search methods as described above. In this embodiment, as is clear from the processing procedure shown in the flowchart, if the UUID can be extracted, a search by UUID is executed. If the UUID cannot be extracted, the number of effective words as a search word is checked, and there are search words exceeding a predetermined value. If it does not exist, the search term is extracted, and if it does not exist, the feature amount is extracted. That is, the search method is applied by prioritizing in the order of UUID, search word, and feature value.

Embodiment 2. FIG.
The document search process in this embodiment will be described with reference to the flowchart shown in FIG. In FIG. 4, the same processes as those described in Embodiment 1 with reference to FIG. 3 are assigned the same step numbers, and the description thereof is omitted as appropriate. Note that the apparatus configuration in the present embodiment may be the same as that in the first embodiment, and a description thereof will be omitted.

  When a read image is formed by document scanning (step 101), a UUID is extracted from the read image (step 102). When the UUID can be normally extracted from the read image, the UUID search unit 10 searches the UUID database 20 using the UUID as a key (steps 103 and 104), and displays the search result on the screen (steps 104 and 112).

  On the other hand, when the UUID cannot be normally extracted from the read image, the processing in this case is a processing procedure different from that in the first embodiment. At this time, the extraction processing control unit 2 scans when the document is read. The mode is confirmed, and if the scan mode is the photo mode, the feature amount is extracted, and similar images are searched and displayed (steps 201, 110 to 112). On the other hand, if the scan mode is not the photo mode, the search term is extracted, and the document is searched and displayed (steps 201, 105 to 109, 112).

  In other words, the above description is based on a “document” written in characters, but the paper to be read may actually have all or most of the print area of the photo. In this case, it is considered that the user selects and scans a photo mode in consideration of reading accuracy. In this case, it is considered that it is more appropriate to search using the feature amount than to extract the search word. Therefore, in this embodiment, the search method is selected by examining whether or not the photo mode is selected as the scan mode. That is, in the present embodiment, the search method is applied by prioritizing in the order of UUID, feature amount, and search word.

Embodiment 3 FIG.
The document search process in this embodiment will be described with reference to the flowchart shown in FIG. In FIG. 5, the same processes as those described in Embodiment 2 with reference to FIG. 4 are given the same step numbers, and the description thereof is omitted as appropriate. Note that the apparatus configuration in the present embodiment may be the same as that in the first embodiment, and a description thereof will be omitted.

  When the scanned image is formed by scanning the document (step 101), the extraction processing control unit 2 confirms the scan mode when the document is scanned, and if the scan mode is the photo mode, the feature amount is extracted and similar. Images are retrieved and displayed (steps 201, 110 to 112). On the other hand, if the scan mode is not the photo mode, the UUID is extracted from the read image (step 102). When the UUID can be normally extracted from the read image, the UUID search unit 10 searches the UUID database 20 using the UUID as a key, and displays the search result on the screen (steps 103, 104, and 112).

  That is, in the second embodiment, it has been described that it is suitable for the case of a photographic document by giving priority to the feature amount over the search word. In the present embodiment, the tendency is further strengthened. That is, in the present embodiment, in the case of a photographic document that cannot be printed with a UUID, the feature amount is to be extracted without confirming the UUID. In other words, in the present embodiment, the search method is applied by prioritizing in the order of feature quantity, UUID, and search word.

  According to each of the above embodiments, each search process is selectively executed and the search result is displayed, so that it is possible to reliably find document information that matches the leaked document. In addition, since the user only needs to perform an operation of scanning a document, it is simple.

Embodiment 4 FIG.
The document search process in this embodiment will be described with reference to the flowchart shown in FIG. In FIG. 6, the same processes as those described in Embodiment 1 with reference to FIG. Note that the apparatus configuration in the present embodiment may be the same as that in the first embodiment, and a description thereof will be omitted.

  When a scanned image is formed by document scanning (step 101), the extraction processing control unit 2 performs all the extraction processing of all document feature information extracted by the present embodiment, that is, UUID, search word, and feature amount simultaneously. (Steps 102 to 104, 105 to 109, 110 to 111). Then, the search result evaluation unit 16 merges the search results obtained by the search units 10, 12, and 14 (step 401). In the merge, a predetermined evaluation criterion, for example, the number of extractions is set for each search result, and the search results are arranged in a predetermined order such as UUID, search word, and feature amount. The display unit 18 displays the merged result on the screen (step 112).

  According to the present embodiment, each search process is executed in combination, and the search results can be merged and displayed. Therefore, there is a high possibility that document information matching the leaked document can be found. In addition, since the user only needs to perform an operation of scanning a document, it is simple.

It is a functional block block diagram which showed one Embodiment of the document search device concerning this invention. 1 is an overall configuration diagram of a network system to which a document search device according to an embodiment is applied. 3 is a flowchart showing document search processing in the first embodiment. 10 is a flowchart illustrating document search processing in the second embodiment. 10 is a flowchart showing document search processing in the third embodiment. 10 is a flowchart showing document search processing in the fourth embodiment.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 Image reading part, 2 Extraction process control part, 4 UUID extraction part, 6 Feature-value extraction part, 8 Search term extraction part, 10 UUID search part, 12 Similar image search part, 14 Document search part, 16 Search result evaluation part, 18 display unit, 20 UUID database, 22 similar image database, 24 document database, 26 document information database, 30 image forming apparatus, 32 database server, 34 LAN.

Claims (6)

In a document search apparatus for searching for a document that matches input image data from a database in which information about the document is stored,
Image data input means for inputting image data of a document;
Document feature extraction processing means for extracting a plurality of types of document feature information capable of specifying the document from image data of the input document;
Obtaining means for obtaining selection candidates for the document by searching the database based on each document feature information extracted from the document feature extraction processing means;
Evaluation means for identifying the document by evaluating the selection candidates acquired by the acquisition means;
Output means for outputting an evaluation result by the evaluation means;
A document search apparatus characterized by comprising: The document search apparatus according to claim 1, wherein
The document feature extraction processing unit includes an identification information extraction unit that extracts identification information uniquely assigned to the document as document feature information from image data,
The document search apparatus, wherein the acquisition unit includes an identification information search unit that searches the database in which identification information uniquely assigned to each document is registered. The document search apparatus according to claim 1 or 2,
The document feature extraction processing unit includes a feature amount extraction unit that calculates a feature amount of the document from image data as document identification information.
The document search apparatus, wherein the acquisition unit includes a feature amount search unit that searches the database in which feature amounts of each document are registered. The document search apparatus according to claim 1 or 2,
The document feature extraction processing unit includes a search word extraction unit that extracts a search word used for database search from image data of the document as document identification information,
The document search apparatus, wherein the acquisition unit includes a document search unit that searches the database in which each document is registered. The document search apparatus according to claim 1, wherein
The document feature extraction processing means includes an extraction process control unit that controls the order of each process for extracting a plurality of types of document feature information. It is implemented in a document search device that searches a document that matches input image data from a database in which information about the document is stored,
An image data input step for inputting image data of the document;
A document feature extraction step of extracting a plurality of types of document feature information capable of specifying the document from the image data of the input document;
An acquisition step of acquiring selection candidates for the document by searching the database based on each document feature information extracted from the document feature extraction step;
An evaluation step of identifying the document by evaluating the selection candidates acquired by the acquisition means;
An output step of outputting an evaluation result by the evaluation step;
A document retrieval method comprising:

Images