JP3486168B2 - Search system, filing system, recording medium - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a search system for searching a character string included in image information and a filing system equipped with the search system.

[0002]

2. Description of the Related Art Conventionally, a document or the like printed on paper is read by a scanner,
A filing system is known in which image information (image information) is stored, and the stored image information can be displayed or printed as needed.

Some of such filing systems allow retrieval of image information by a keyword so that the image information corresponding to a desired document can be easily extracted from the stored image information. In such a system, when an image including a document printed on paper is loaded into a computer and stored, a search keyword indicating the content of the read document is input from a keyboard or the like, and the input search keyword is used as image information. It is stored in association with. Then, when the user searches for image information of a required document, the user inputs a keyword indicating the content of the document to be searched, and the system compares the keyword with the stored search keyword and The image information with the search keyword matching is output. Therefore, when the search keyword input when storing the image information of the document is insufficient as the information indicating the document, the keyword does not match the search keyword and the desired keyword is not found. The image information of the document cannot be extracted. If you do not enter an accurate search keyword that indicates the contents of the document, the search keyword will not match the keyword entered by the user, and the image information will not be used as it is stored. To do. In addition, it is very troublesome to input such a search keyword for each image information of each document.

On the other hand, there is known an OCR (optical character reading device) for reading a document as image information, converting the image information into text information, and storing the text information. It is also converted to and saved. With such a method, it is not necessary to register a keyword in advance, and a document can be searched with a free keyword. However, OCR generally has a recognition rate of about 95% to 99%, and stores only the converted text information. Therefore, in order to retain the content of the document, the converted text information is the content of the document. There is a problem that the user has to check the text information after conversion so as to be consistent with and correct it. Such an effort becomes enormous as the number of documents to be imported increases.

Therefore, according to the present invention, there is no need for inputting,
An object of the present invention is to provide an image information retrieval system that can retrieve a desired document and a filing system including the retrieval system.

[0006]

According to the search system of claim 1, which is made to solve the above-mentioned problems, recognition processing is performed at the time of keyword search, and the input search keyword is included. Image information or information for identifying the image information is output.
Therefore, it is not necessary to input the search keyword for each document in advance, and it is not necessary to convert the document into text information by OCR in advance. That is, the characters contained in the stored image information can be directly searched by keyword. That is, a desired document can be taken out as image information without the trouble of previously inputting a search keyword corresponding to the image information and correcting the text converted by OCR.

The recognition process for the image information being recognized is stopped when it is determined that the search keyword is included. That is, the subsequent recognition process for the image information being recognized is not performed. Therefore, at that point, it is possible to move to other processing such as reading another image information and performing a search processing. The information specifying the image information includes, for example, information indicating the location where the image information is stored, a file name, character information up to the time when the recognition is completed, and the like.

Further, if there is a portion that cannot be converted from image information into text information, as described in claim 2,
It is advisable to ignore that part and proceed with the recognition process. In this way, the search can be continued even if there is image information that has an unrecognizable portion. In addition,
The image information having a portion that cannot be recognized may be output to a display device together with the fact that there is a portion that cannot be recognized.

The image information may be composed of one image file, or may be composed of a plurality of image files. Further, the one image file may be, for example, one sheet of paper (document) captured by a scanner, or a plurality of sheets of paper may be captured as one image file.

As for a plurality of related image information, it is preferable that the image information is stored as a set as in claim 3 and the set is read out for retrieval. That is, for example, when performing a search, a set of image information of the first document group, a set of image information of the second document group, a set of image information of the third document group, ..., Of the nth document group When a set of image information is stored in the image information storage means, the set of image information of each document group is read and retrieved. That is, from the set of image information of the first document group, for example, the image information of the first document, the image information of the second document, ..., The image information of the mth document are searched, and the keyword is The set of image information of the second document group is searched at the stage where it is found. That is,
It is possible to search whether or not a keyword is included in a set of image information. Therefore, for example, if there is a keyword in the image information of the first document of the first document group, then the second image information of the first document group, the third image information of the first document group, ... It is possible to prevent the search from being repeated. That is, the unrecognized image information included in the set of image information determined to include the keyword is not searched, and the search can be performed at high speed. Further, in this way, for example, when parallel processing is performed as in claim 4, the image information of each document group can be retrieved at the same time, and a plurality of image information in the image information of one document group can be retrieved. There is no waste of searching at the same time.

[0011] Claim 4 indicates that the processing of different means may be performed in parallel (parallel), and the processing of the same means may be performed in parallel (parallel). For example, the recognition target image reading means reads a plurality of image information in parallel, the character recognizing means recognizes the plurality of image information in parallel, and the search means makes a character as a recognition result of each image information. And the keyword are compared in parallel to determine whether or not they match, and a search result is obtained. For example, the character recognition unit divides one image information of the recognition target into a plurality of regions and parallelizes them. It is good to recognize. In this way, since recognition and search can be performed in parallel at the same time, search results can be obtained faster.

The first system group including the image information storage means, the recognition target image reading means, and the character recognition means, and the second system including the search keyword input means and the search means.
May be provided, and the first system group and the second system may be connected to each other by a network to perform distributed processing (parallel processing). Further, the first system group may be provided with the function of the second system. In this way, the load is distributed and the search can be performed at a higher speed.

Further, as described in claim 5,
It is advisable to store the recognition result once recognized and perform a search using the stored recognition result. In this way, since it is not necessary to perform character recognition again for image information that has once undergone character recognition, the faster the search, the faster the search can be performed.

Incidentally, storing in association with the image information also includes the case of storing the information including the position where the recognition is completed. In this case, when the next keyword search is performed from the same image information, the search means is
First, the stored recognition result is searched up to the position where the recognition is completed, and the recognition unit recognizes the part after the position and the recognition result is searched. These processes may be performed in parallel. That is, the character recognizing unit may perform the recognition, and the searching unit may search the character stored in the recognition result storage unit and the character recognized by the character recognizing unit in parallel.

Now, in such a system, claim 6
As shown in, the search may be performed at any time by replacing the character recognition means with good performance. For example, when the character recognition means is composed of hardware, the hardware can be replaced, and when the character recognition means is realized by software, the software can be replaced. In this way, for example, it is possible to change the character recognition means to have a high recognition rate depending on the image information of the recognition target. For example, when retrieving a legal document, the character recognition means having a high recognition rate of the legal document is switched to, or in the case of a handwritten character, the character recognition means having a high recognition rate of the handwritten character is switched to. Further, the character recognition means may be able to update or change, for example, via a network. In particular, since the present system recognizes characters at the time of searching, if this is done, the optimum character recognition means can be used at the time of searching to perform the search. Since the recognition rate of the character recognizing means can be expected to improve year by year, the recognition accuracy is higher than that in the case of converting the text information into text information in advance at the time of capturing image information as in the conventional OCR, and the accuracy of the search can be improved. .

The search system as described above can also search for image information input from a network or a recording medium, for example. Further, according to the seventh aspect, it is possible to provide a filing system in which a document written on a paper is taken in as image information and the result of searching the image information is displayed. The search result may be sent via the network without being displayed. It may also be printed by a printer or the like. Also, the search keyword input means is
You can use a keyboard, etc. provided for such a system,
It may be input via a network.

As described in claim 8, claim 1
The function of realizing each means described in 7 by a computer is
For example, it can be provided as a program that is activated by a computer. In the case of such a program, for example, it is recorded in a computer-readable recording medium such as a floppy (registered trademark) disk, a magneto-optical disk, a CD-ROM, a hard disk, and loaded into a computer system and activated as necessary. Can be used by Further, since each unit may constitute a search system or a filing system as a whole, each unit may be realized by a plurality of computers. For example, each unit may be realized by a different computer.

[0018]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments to which the present invention is applied will be described below with reference to the drawings. Needless to say, the embodiment of the present invention is not limited to the following examples, and various forms can be adopted as long as they are within the technical scope of the present invention.

FIG. 1 shows a filing system 1 of this embodiment.
Shows the configuration of. PC 1 for filing system 1
1, a display 12 connected to the personal computer 11, a keyboard 13, a mouse 14, a scanner 15, a printer 1
A computer system including 6 and a personal computer 11
Is a communication interface connectable to the network 50, a hard disk for recording data, an RO
A display 12, a keyboard 13, a mouse 14, a scanner 15, a printer 16 including M, RAM, I / O, etc.
Etc.

The scanner 15 is equipped with an ADF (automatic document feeder) and can continuously read the set documents and output them as image data to the personal computer 11. When the document to be read is set in the scanner 15 and the scan start button of the scanner 15 is pressed, the scanner 15 transmits the fact that the reading instruction has been input to the personal computer 11. Then, the personal computer 11 outputs an image reading instruction to the scanner 15. Then, the scanner 15 scans the set documents one after another with a sensor to generate image data of the document and transmits the image data to the personal computer 11. The personal computer 11 stores the image data transmitted from the scanner 15 in the hard disk as an image file in a predetermined format (compressed format such as JPEG or PNG). One image file is stored for each document.

An image file of a document containing a search keyword can be searched from the image files recorded in this way. This search process is a process of searching the hard disk for an image file containing the characters of the keyword input via the keyboard 13, the network 50, or the like. The flow of this keyword search processing will be described with reference to the flowchart of FIG.

In S110 of FIG. 2, the image file is read from the hard disk, developed as an image, and read in the memory. In step S120, the characters in the image on the memory are recognized by a known recognition algorithm, and the recognition result characters are written in the cache.

In S130, it is searched whether or not the character string matching the search keyword is in the cache. In S140, it is determined whether or not there is a character string that matches the keyword as a result of the search. If there is a character string that matches the keyword (S140: YES), the process proceeds to S150, and the character string that matches the keyword. If there is no (S140:
(NO), the process proceeds to S145.

In S145, it is determined whether or not all the character recognition and search processing in the image file being processed are completed, and if completed (S145: YES), S160.
If it is not completed yet (S145: N
O), the process moves to S120 to recognize the next character.

In step S150, the image determined to contain the character string matching the keyword in step S140 is displayed on the display 12. In S160, it is determined whether a user's end instruction is input and whether or not all of the image files to be searched stored in the hard disk have been read in S110. If there is a user's end instruction, If all the data has been read (hereinafter, referred to as the case where the processing is completed) (S160: YES), the search processing ends. On the other hand, if the user has not input the end instruction and the processing of the image file has not been completed, the read target is set to the next image file, and the process returns to S110 and the same search processing is performed.

It should be noted that the characters are recognized one by one from the beginning of the document (image) (the beginning is determined by a known method), the characters of the recognition result are written in the cache, the following characters are recognized, and the characters are continuously recognized. The character string is formed in the cache by repeatedly writing the character in the cache. Then, it is determined whether or not there is a character string that matches the keyword with respect to the character string. If they match, the recognition is terminated at that point, and the corresponding image file is read and displayed. If there is an unrecognizable character, that part is ignored and recognition and search are continued. In this way, the recognition process is stopped when there is a character string that matches the keyword. For example, when there is a character string that matches the keyword at the position 10% from the beginning of the image, the remaining 90% Does not perform recognition processing. Therefore, the search can be performed faster than the entire image is recognized.

In this way, it is possible to perform a keyword search while reading an image file and performing character recognition at the time of search. Therefore, it is possible to search for and extract an image file of a desired document by using an arbitrary character string as a keyword, without the trouble of inputting a keyword for search and correction after OCR as in the conventional case. Therefore, the image file captured and stored is effectively used.

In the embodiment, the keyboard 13 corresponds to search keyword input means, and the hard disk of the personal computer 11 corresponds to image information storage means. Also,
The image file corresponds to the image information, and S1 in FIG.
The process of 10 corresponds to the process of the recognition target image reading unit, the process of S120 corresponds to the process of the character recognition unit, and the process of S130 corresponds to the process of the searching unit. The scanner 15 corresponds to a reading unit, and the display 12 corresponds to a display unit. [Other embodiment] (1) In the above embodiment, the image file is scanned by the scanner 1
Although it is decided that the input is made from the input terminal 5, the input may be made via the network 50, a CDROM or a DV.
You may make it input from a recording medium such as D. Also,
The search result image is displayed on the display 12, but may be output to the printer 16. (2) In the above embodiment, the search keyword is the keyboard 1
However, the input may be made via the network 50. For example, the personal computer 11 functions as a search server, a search keyword is input from a client personal computer connected to the network 50, and the search result is transmitted to the client personal computer. By doing so, the image file of the desired document can be retrieved from the personal computer connected to the network 50. (3) Character recognition processing in the image (S120 in FIG. 2)
Then, the search process of the keyword of the character string in the cache (S130 in FIG. 2) may be performed in parallel. For example, a recognition processing thread that performs recognition processing and a keyword search thread that performs keyword search are executed in parallel, and if the keyword search processing thread determines that there is a keyword match, an end message is sent to the recognition processing thread to perform recognition. When the processing thread receives the end message, the processing thread ends the recognition processing. (4) In addition to the above (3), it is preferable to execute the entire processing of FIG. 2 in parallel (for example, in multithread). That is, the keyword search process is simultaneously performed on a plurality of image files. By doing so, the image files that match the keyword are output in order. Moreover, if these threads are executed by a plurality of different CPUs, the processing speed is increased. Further, as shown in FIG. 3, the plurality of servers 20 read the image file stored in the personal computer 11 via the network (corresponding to S110) and execute the recognition process of S120, and the cache of the personal computer 11 is executed. Alternatively, the character of the recognition result may be sent via the network 50, and the personal computer 11 may perform the processing of S130 to S160 to perform the search. Generally, the recognition process in S120 takes a long processing time, and thus the processing can be distributed to the plurality of servers 20 in this way, whereby a higher speed search can be realized. (5) In the above embodiment, one file is used for each document, but a set of related documents (for example, one book, one report, etc.) is collectively stored in one image. Alternatively, the search may be performed. In this way, the search is performed from the beginning for each set of documents. In particular, as in the cases of (3) and (4) above, when searching in parallel, the search processing is performed in parallel for each set of documents.

On the other hand, one image file may be generated for each document, and a set ID may be given to each set of documents and recorded in each image file.
The set ID is information indicating a related document including a plurality of documents such as a book or a report,
For example, when a document set in the scanner at one time is scanned, the same set ID may be automatically recorded in the image file generated by the scan,
The scanned image data may be displayed on the display 12, and the user may select a set of images using the mouse 14 or the like and record the same set ID as a set of image files. In this way, a plurality of image files in which the set ID is recorded are recorded in the hard disk. When the search processing is performed in parallel, the image files having different set IDs are searched at the same time. If you do this,
Multiple sets of documents can be searched concurrently. (6) In the above embodiment, the recognition process (the process of S120) is performed every time when the image file stored in the hard disk is searched. However, for the image file that has been recognized once, the recognition result character is changed. The image file may be stored in a hard disk as a recognition result storage unit in association with the image file, and the character stored in the hard disk may be searched for when the next search is performed. That is, as shown in FIG. 4, a process of storing the recognized text in S147 is added to the keyword search process of FIG. 2, and the recognized character (text) is stored in the hard disk. When a search is performed for the same image file as the image file stored in association with each other in this way, it is determined in S101 that the text corresponding to the image file to be searched has been stored (S101:
YES), and the stored text is searched by keyword in S103. Then, in S105, it is determined whether or not the keyword matches, and if they match (S105: YES),
The process proceeds to S150 and the corresponding image is displayed. If they do not match (S105: NO), the process proceeds to S160. In this way, the image file once recognized does not need to be recognized again, and the keyword search can be performed at high speed.

If a keyword is found during the recognition, that is, if S140: YES in FIG. 5, the text up to the recognized position and the information about the position where the recognition of the image file is completed are stored. You may do it. In this case, the next time the same image file is searched, the stored text is searched up to that position, and recognition processing is performed from the subsequent positions. Good. By doing so, it is not necessary to perform the recognition again for the part that has been recognized once, and the search can be performed at a higher speed. (7) In the above embodiment, the keyword search process may be further stopped when the user inputs a stop instruction. By doing this, for example, if the user mistakenly inputs the search keyword or if it is enough to find some image files containing the search keyword, the search process will be completed until the end. Can be prevented and the system becomes easy to use. (8) In the above embodiment, the program for performing the recognition process in S120 of FIG. 2 may be downloaded from a center (not shown) via the network 50 connected to the Internet. The center always tries to improve the performance of the program that performs this recognition processing. In this way, it is possible to search using the latest recognition processing program at the time of search.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of a filing system according to an embodiment.

FIG. 2 is a flowchart showing a keyword search process of the embodiment.

FIG. 3 is a diagram showing a configuration of a filing system according to another embodiment.

FIG. 4 is a flowchart showing a keyword search process of another embodiment.

[Explanation of symbols]

1 ... Filing system 11 ... PC 12 ... Display 13 ... Keyboard 14 ... Mouse 15 ... Scanner 16 ... Printer 20 ... server 50 ... Network

─────────────────────────────────────────────────── ─── Continuation of the front page (58) Fields surveyed (Int.Cl. ⁷ , DB name) G06F 17/30 G06T 1/00

Claims (9)

(57) [Claims] 1. A search keyword input means for inputting a search keyword, an image information storage means for storing a plurality of image information, and image information to be recognized among a plurality of image information stored in the image information storage means. A recognition target image reading means, a character recognition means for sequentially recognizing characters included in the image information read by the recognition target image reading means, and a recognition result of the character recognition means, which is input by the search keyword input means. When the search keyword is included, the sequential recognition of the character recognition means is stopped, and the search means for outputting the image information read out by the recognition target image reading means or the information for specifying the image information. A search system comprising: 2. The search system according to claim 1, wherein if there is a portion that cannot be converted from the image information into the character, the character recognition unit ignores the portion and continues recognition. Search system. 3. The retrieval system according to claim 1, wherein the image information storage means stores a plurality of related image information as a set, and the recognition target image reading means stores the recognition target image information. A retrieval system, wherein the image information is read as a set of target image information in units of sets. 4. The search system according to claim 1, wherein at least any two or more of the recognition target image reading means, the character recognition means, and the search means perform processing in parallel. At least one of executing the processing, or executing at least one of the recognition-target image reading means, the character recognition means, and the search means in parallel as a processing means. A search system characterized by that. 5. The search system according to claim 1, further comprising a recognition result storage unit that stores a recognition result of the character recognition unit in association with image information of a recognition target, the search unit. Is a search system, wherein the recognition result stored in the recognition result storage means is preferentially searched. 6. The search system according to claim 1, wherein the character recognition means is replaceable. 7. A filing system including the search system according to claim 1, wherein a paper is read by a sensor to generate image information, and the generated image information is stored in the image information storage means. And a display unit for displaying the image information output by the search unit or the information for specifying the image information. 8. A computer-readable recording medium storing a program for causing a computer to function as each means for definitive the retrieved <br/> stearyl beam according to any one of claims 1-6. 9. The filing system according to claim 7.
To make a computer function as each means in
Computer-readable record of the program
Medium.