JP3673553B2 - Filing equipment - Google Patents

Description

[0001]
[Industrial application fields]
The present invention relates to a filing apparatus that recognizes and stores characters in a document image and a control method therefor.
[0002]
[Prior art]
In a document filing apparatus that captures a printed document with a scanner and stores it as image data, a document filing apparatus configured to recognize a text area in the image data and use the obtained character code as data for collation at the time of search. Exists. The configuration and operation will be described below.
[0003]
FIG. 6 is a flowchart showing the flow of processing during registration. In step S601, a document to be registered is read by a scanner and converted into binary image data. In step S02, region separation is performed on the obtained input image, and only the text region where the character image exists is extracted. Extracting only the text area is possible by extracting the connected components of black pixels in binary image data and combining only those estimated to be characters, which is an existing technique. In step S603, character recognition processing is performed on the text area to obtain text data. This text data is used for matching with a search keyword at the time of search. In step S604, the entire image data is compressed. In this embodiment, for example, MMR, which is optimal for binary images, is used. The image data and text data compressed in step S605 are integrated, and a header describing the date, registrant name, data size, etc. is added and stored in the external storage device.
[0004]
FIG. 7 is a schematic diagram of region separation. In the figure, the input image 701 is subjected to region separation to obtain an image 702 having only a text region. Character data is recognized for this text area to obtain text data. On the other hand, the compressed original image data is paired with the text data, and a header is added to hold the data as one piece of data 703.
[0005]
FIG. 8 shows the data format stored. The first data is stored in the area indicated by reference numeral 801, and the second and subsequent data are stored in order from 802.
[0006]
Next, the flow of processing at the time of search will be described using the flowchart of FIG.
[0007]
In step S901, a search keyword is input. In step S902, 1 is set to a counter i for storing the number of search data. Next, in step S903, the text data part of the i-th data stored in the data is collated with the input search keyword. This collation performs a so-called full-text search, and checks whether or not the search keyword is included in the text data portion as it is.
[0008]
In step S904, it is determined whether a search keyword is included. If it is determined that the search keyword is included in this determination, the process proceeds to step S905, where all image data of the found data is decompressed and displayed on the display, and the process proceeds to step S906. Note that detecting that the search keyword is included in the text data portion is referred to as “hit”.
[0009]
On the other hand, if it is determined in step S904 that the search keyword is not included, the process proceeds to step S906.
[0010]
In step S906, it is determined whether the end of the search target has been reached. If it is determined that the final data has not yet been reached, the counter i is incremented by one and the process returns to step S903.
[0011]
As described above, it is possible to search for an image of data containing the search keyword as a character image and display it on the display.
[0012]
[Problems to be solved by the invention]
However, in the above conventional example, since the search keyword is directly matched with the text data, the text data includes an erroneous character due to the misrecognition of the character recognition, and the erroneously recognized character corresponds to the search keyword. If it exists in the column, there is a problem that data that should be hit cannot be searched.
[0013]
[Means for Solving the Invention]
The present invention has been made in view of such problems, and in consideration of the fact that character recognition is not complete, an object of the present invention is to provide a filing apparatus and a control method therefor that can improve the search rate based on a search keyword. Is.
[0014]
In order to solve this problem, for example, the filing apparatus of the present invention has the following configuration. That is,
A filing device that stores document images and searches for document images based on a search keyword using data obtained by recognizing characters in the document images,
Storage means for storing candidate character groups obtained as a result of recognition of characters in the input document image in association with similarities;
First calculation means for calculating a degree of coincidence between a given search keyword and a first combination in which the similarity of the recognition result is a combination of the first candidate characters;
If the degree of coincidence calculated by the first calculating means is greater than or equal to a predetermined first threshold , the second and subsequent candidates in the search keyword and a candidate character group determined to be greater than or equal to the predetermined first threshold Second calculating means for calculating the degree of coincidence with the second combination including the character of
Determining means for determining that the search keyword has been hit when the degree of coincidence calculated by the second calculating means is equal to or greater than a predetermined second threshold value;
[0015]
The filing apparatus control method of the present invention includes the following steps. That is,
A method for controlling a filing device that stores document images and uses data obtained by recognizing characters in the document images to search for document images based on a search keyword,
An accumulation step of accumulating candidate character groups obtained as a result of recognition of characters in the input document image in association with similarities;
A first calculation step of calculating a degree of coincidence between a given search keyword and a first combination in which the similarity of the recognition result is a combination of the first candidate characters;
If the degree of coincidence calculated in the first calculation step is greater than or equal to a predetermined first threshold, the second and subsequent candidates of the search keyword and a candidate character group determined to be greater than or equal to the predetermined first threshold A second calculation step of calculating the degree of coincidence with the second combination including the character of
A determination step of determining that the search keyword is hit when the degree of coincidence calculated in the second calculation step is equal to or greater than a predetermined second threshold value .
[0016]
Furthermore, a display means for displaying the matching degree, a list of corresponding data, an instruction means for instructing desired data in the displayed list, and a display means for displaying the instructed data are provided. Is desirable. As a result, the user can determine the certainty of the retrieved data, and can display the data based on the determination result.
[0018]
The second calculating means determines a range to be the second combination in the second and subsequent candidate character groups based on the similarity of the first candidate character obtained as the recognition result. It is desirable to do. As a result, the number of second combinations to be collated can be reduced, and the search becomes faster.
[0019]
【Example】
Hereinafter, embodiments according to the present invention will be described in detail with reference to the accompanying drawings.
[0020]
<Description of the first embodiment>
FIG. 1 is a block diagram of a filing apparatus according to the first embodiment. In FIG. 1, reference numeral 101 denotes a scanner that irradiates light on an image original, reads the reflected light and converts it into an electrical signal, and 102 converts an electronic signal obtained by the scanner 101 into a binary digital electrical signal. A scanner interface circuit for transmitting to the element; 103, a pointing device (for example, a mouse) for inputting desired coordinates on a display window; 104, a signal received from the pointing device 103, and the other device An interface circuit for transmission to the constituent elements, 105 is a CPU for executing control of the entire apparatus, character segmentation processing and recognition processing, 106 is a control program executed by the CPU 105, various processing programs, font data, etc. ROM 107 includes character image expansion and character recognition processing. Working area for a RAM used as such. Reference numeral 108 denotes a display for displaying an input image and a recognition result, and 109 denotes a display interface circuit. The display 108 displays an image of the VRAM area stored in a predetermined address area of the RAM 107. Reference numeral 110 denotes an external storage device such as a hard disk in which registered data is stored. In addition to storing data, a dictionary for character recognition is stored. Reference numeral 111 denotes the interface. Reference numeral 112 denotes a bus for connecting each device component.
[0021]
The processing for registering the input document image and the data structure of the stored data are substantially the same as those described above with reference to FIGS. However, when filing an input document image, the character codes up to the fourth candidate are registered instead of registering only the first candidate character obtained by the character recognition process as the text data.
[0022]
Next, the flow of processing at the time of search will be described with reference to the flowcharts of FIGS. 2 and 3, FIG. 4, and FIG.
[0023]
For example, it is assumed that a character string “internal processing integrated type” is input as a search keyword.
[0024]
First, in step S201, a keyword is input, and “1” is assigned as an initial value to a counter i indicating the number of search data.
[0025]
Next, the process proceeds to step S203, and matching with the search keyword is performed for all candidates in the text data portion of the stored i-th data.
[0026]
In step S204, it is determined whether there is a character string corresponding to the search keyword based on the processing result in step S203. If it is determined that they match or almost match, the image portion of the data to be outlined is expanded and displayed together with the character string in step S205. If the user gives an instruction to retrieve the next data, the process proceeds to step S206. If it is determined that the data is terminated, the retrieval process is terminated. If not, the variable i is incremented in step S207. The next data is searched.
[0027]
The processing procedure of step S203 in the above processing is shown in FIG. Hereinafter, the contents of the process will be described.
[0028]
Note that the variables n, c, j, and k in the following description are secured in the RAM 107. The variable n is a counter (pointer) indicating the character position of the text data, the variable c is a counter indicating the number of characters matching the keyword, the variable j is a counter indicating one character position in the character string of the keyword, and the variable k is a candidate character. It is a counter representing the order.
[0029]
In steps S301 to S304, “1” is given to each variable as an initial value.
[0030]
In step S305, the matching of the kth candidate X (n, k) with the nth character of the text data of the data of interest (ith data) and the jth character Y (j) of the search keyword is performed. Judge whether it can be taken. That is, it is determined whether X (n, k) and Y (j) are equal.
[0031]
If it is determined that they are not equal, the process proceeds to step S306 to determine whether or not all candidates for the nth character in the text data have been collated. If it is determined that there is an unmatched candidate, in step S307, the variable k is incremented by "1" to refer to the next candidate for the target character position, and the process of step S305 is performed. As a result, as shown in FIG. 5, collation is sequentially performed on each combination of recognition candidates.
[0032]
On the other hand, if a character equal to the j-th character in the keyword character string is found, the process proceeds to step S308, and a counter c that counts the number of matched characters is incremented.
[0033]
When the process proceeds to step S309, it is determined whether or not the collation process for the last character of the search keyword is completed. If it is determined that the collation for the last character of the search keyword is not completed, the process proceeds to step S310, and the variable j is incremented. In step S311, it is determined whether or not there is a character in the next text data. If there is, a variable n is incremented in step S312 to advance the character position to be collated to the next character position, and in step S304. Return to.
[0034]
On the other hand, if it is determined in step S311 that the text data is terminated, it is determined in step S316 that there is no search keyword in the text data, the process is terminated, and the process returns to the process of FIG.
[0035]
If it is determined in step S309 that collation for the last character of the search keyword has been completed, the process proceeds to step S313, and the degree of matching m is calculated.
[0036]
Here, the calculation of the degree of coincidence m in the embodiment is based on, for example, the following equation.
[0037]
m = (number of matched characters) / (total number of characters) × 100 [%]
That is, the number of characters matched in the character string of the search keyword is shown as a percentage (the larger the value of m, the higher the possibility of matching the search keyword).
[0038]
When the process proceeds to step S314, the degree of coincidence m calculated as described above is compared with a preset value M, and if it is less than that, the character string starting from the nth character in the text data portion is: It is determined that the search keyword is not checked, and the process proceeds to step S317. In step S317, it is determined whether or not the end of the text data portion has been reached. If it is determined that the text data is not finished, the variable n is incremented by one, the next character position in the text data portion is set as the cut-out position, and the processes in and after step S302 are performed.
[0039]
If it is determined in step S317 that the text data portion has ended, the process proceeds to step S319, where it is determined that there is no character string corresponding to the search keyword in the second text data of interest. To return to the processing of FIG.
[0040]
If it is determined in step S314 that the degree of matching is m> M, the process proceeds to step S315, where the text data portion of the data of interest matches the search keyword (when m = 100), or If there is a high possibility of matching, it is determined that there is a character string, the result is returned to the upper routine, and the present process is terminated.
[0041]
In the above processing, for example, when “internal processing integrated type” is input as a search keyword and a search is performed, for example, the i-th text data portion in certain attention data is “inside” as shown in FIG. It is assumed that the characters “processing”, “reason”, “go”, and “type” are in the first candidate of the recognition result, the character “Tong” is in the third candidate, and the character “part” is not in the candidate. . At this time, the matching degree m is
6/7 × 100 = 85.7 [%]
It becomes.
[0042]
For example, if the threshold value is set to 50%, the threshold value is exceeded in the previous match, so it is determined that the text data has a “matching portion” with the search keyword, and the image data portion can be expanded and displayed. it can.
[0043]
As described above, according to the present embodiment, when a document image is read and character recognition is performed and the result is registered as a database, not only the first candidate for character recognition but also a plurality of candidates are registered. Since the keyword is searched with the combination of candidates, it is possible to increase the rate at which the search with the search keyword is as intended.
[0044]
Furthermore, even if there is no search keyword itself, even if a part of the character string of the keyword is different, if the degree of matching as a whole is to some extent, it becomes a search target. It becomes possible to increase the possibility of being searched even if there is a character that does not exist.
[0045]
Note that according to the above description, the greater the number of characters in the search keyword, the higher the certainty for the determination result. Therefore, when the number of characters in the character string of the search keyword is n or more, the determination based on the matching degree may be performed. In some cases, a threshold for determining the degree of coincidence may be set by the user. For example, if the threshold value is increased, if there are a small number of characters, it must be completely matched as a whole. Conversely, if there are many characters, it can be determined that the keyword is held even if several characters do not match.
[0046]
<Description of the second embodiment>
Next, a second embodiment will be described. In the present embodiment, as the first stage, individual characters constituting the search keyword are collated with the first candidate of text data in the text data portion to be searched, and the degree of coincidence is equal to or higher than the first threshold value M1. If it is determined whether or not there is a threshold value M1 or more, it is determined that there is a high possibility that the character string matches the search keyword. When the degree of coincidence is recalculated with a combination of the second and lower candidates of the corresponding character (lower than the second candidate of the character determined to be inconsistent) and it is determined that it is greater than the second threshold M2. Determines that there is a character string corresponding to the search keyword in the text data.
[0047]
Therefore, the relationship between the threshold values M1 and M2 is M1 <M2. That is, in the first stage, it is determined whether or not there is a character string that can be a search keyword, and if it is determined that there is a possibility, the collation described in the first embodiment is performed. Is.
[0048]
Hereinafter, the contents of the operation processing in the second embodiment will be described according to the flowchart of FIG. 10 (corresponding to the flowchart of FIG. 3). The apparatus configuration is the same as that of the first embodiment. Therefore, the program based on FIG. 10 is stored in the ROM 106.
[0049]
First, in steps S1001, 1002, and 1003, each variable is initialized. Next, in steps S1004 to S1006 and steps S1013 to S1014, collation processing of the number of characters given by the search keyword is performed. In this process, if the end of the text data of the data of interest is reached during the collation of the number of characters of the search keyword, it is determined that there is no matching character string, and this process is terminated (step S1015).
[0050]
When the collation of the number of characters of the search keyword is completed, the first step of matching m1 is calculated in step S1007, and compared with a preset threshold value M1 in the next step S1008.
[0051]
Here, if it is determined that m1 <M1, the character string from the cutout position (given by the variable n) in the text data portion is unlikely to be a search keyword, so the processing is step S1017. To determine whether or not the end of the text data portion has been reached. If it is determined that the text data portion has not been reached, the cut-out position is advanced by one, and the flow returns to step S1002. The determination in step S1017 is sufficient if it is determined whether or not the position is obtained by subtracting the number of characters of the search keyword from the number of characters in the text data portion. This is because, after that, the number of characters of the search keyword is smaller, so the determination result in step S1014 is always yes.
[0052]
If it is determined that the relationship between the matching level m1 and the threshold value M1 in the first stage is m1> M1, the process proceeds to step S1009, and the second candidate character and subsequent characters of the first candidate character that did not match are determined. The collating process is also performed with reference to FIG. Since this collation processing is the same as that of the first embodiment, its description is omitted.
[0053]
Thus, when the collation process including the characters after the second candidate is completed, the final matching degree m2 is calculated (step S1010), and the comparison determination between m2 and the threshold value M2 is performed (S1011).
[0054]
As a result, if it is determined that m2> M2, it is determined that there is a search keyword in the text data portion of the data of interest, and the process is terminated (step S1012).
[0055]
On the other hand, if it is determined that m2 ≦ M2, the process proceeds to step S1017.
[0056]
As a result of the above, it is determined whether or not there is a possibility that the search keyword is matched, and only when it is determined that there is a possibility that the search keyword is matched, the comparison process is performed, thereby comparing with the first embodiment. Thus, the search process can be performed at high speed.
[0058]
<Description of the third embodiment>
FIG. 11 shows that the degree of coincidence and the list of document files are displayed in order of the degree of coincidence so that the user can select them. In the present embodiment, “internal processing integrated type” is input as a search keyword, and the search results are displayed together with the matching degree in descending order of matching degree. A desired document file is selected from the list, and an “open” button is clicked with a pointing device to expand and display the image data portion. As a result, it is possible to prevent oversight of document filing, eliminate unnecessary time for decompressing unnecessary image data portions, and improve usability.
[0059]
In order to realize the third embodiment, when it is determined that the search results match, the corresponding data is not read and decompressed at that time, but is performed on the whole ( This can be dealt with by temporarily creating a file of the degree of coincidence and data number list), and the list shown in the figure may be displayed based on the file. Therefore, in this case, the degree of coincidence is not compared with the threshold value, and the threshold value is set to a low value to make the user judge.
[0060]
<Description of the fourth embodiment>
A fourth embodiment will be described. FIG. 12 is a flowchart showing the processing contents in the fourth embodiment. FIG. 13 is a diagram for explaining the operation. FIG. 14 shows data to be stored. Similarity data is stored in addition to text data and image data.
[0061]
In the fourth embodiment, the collation range is limited using the similarity. If the similarity of the first candidate is equal to or greater than a predetermined threshold value X1, the similarity of the second and subsequent candidates is up to (similarity of the first candidate- α) (α is a predetermined first collation range) If the similarity of the first candidate is less than X1, the similarity of the second and subsequent candidates is (similarity of the first candidate− β) (β is a predetermined second matching range, By setting the range up to α <β ), when the probability of the first candidate of the recognition result is high, the character candidate that is wrong as much as possible is excluded from the collation range, and conversely the first candidate of the recognition result When the probability is low, there is a collation range limiting method in which the collation range is expanded so as not to exclude correct recognition candidates.
[0062]
In the example of FIG. 13 , the threshold similarity X1 is 90, the first collation range α is 10, and the second collation range β is 20, and the similarity of the first candidate “inside” (reference numeral 1401) of the recognition result is 95. Therefore, the collation range is up to recognition candidates having a similarity of 85 or more, and up to “meat” is the collation range. Moreover, the similarity of the first candidate of the recognition result "vertical" (1402) than a 78, matching range is up to recognition candidate having 58 or more similarity, to "integrate" is the collation range . As a result, it is possible to reduce processing time and shorten processing time.
[0063]
In general, in the character recognition process, a feature amount is extracted from a character image to be recognized, and the character closest to the feature amount and the feature amount stored in the recognition dictionary is recognized as the first candidate. Here, the similarity is a value used to determine the rank of candidate characters obtained by the recognition process.
[0064]
In each of the above embodiments, an apparatus for optically reading a document image has been described. However, the present invention is not limited to this, and an image is input via a communication line or stored in a storage medium. You may input the image. Further, it is possible to adapt to a single device, or a system constituted by a plurality of devices. Further, although each processing program has been described as being stored in the ROM, it can be realized by being supplied from the outside (loaded into the RAM), and therefore the present invention is limited by the above embodiment. is not.
[0065]
As described above, according to this embodiment, even when character recognition is not 100% correct when registering document image data, one or more recognition candidates of each recognition target are held as text data, and the text data Search the image data containing the search keyword with high accuracy by matching the recognition candidates with the search keyword, obtaining the degree of match with the search keyword, and determining that the match is found when the match exceeds the threshold There is an effect that can be.
[0066]
【The invention's effect】
As described above, according to the present invention, it is possible to improve the search rate based on the search keyword in consideration that character recognition is not complete.
[0067]
[Brief description of the drawings]
FIG. 1 is a block configuration diagram of an apparatus according to an embodiment.
FIG. 2 is a flowchart of a search according to the first embodiment.
FIG. 3 is a detailed flowchart of a search according to the first embodiment.
FIG. 4 is an example of text data in the first embodiment.
FIG. 5 is an example of text data candidates according to the first embodiment.
FIG. 6 is a flowchart of conventional registration.
FIG. 7 is a schematic diagram of conventional registration.
FIG. 8 is a structure of conventional registration data.
FIG. 9 is a flowchart of a conventional search.
FIG. 10 is a detailed flowchart of a search according to the second embodiment.
FIG. 11 is a display example of a document file list according to the third embodiment.
FIG. 12 is a flowchart of search in the fourth embodiment.
FIG. 13 is an example of recognition result similarity and search target limitation according to the fourth embodiment.
FIG. 14 shows the structure of registration data in the fourth embodiment.

Claims (3)

A filing device that stores document images and searches for document images based on a search keyword using data obtained by recognizing characters in the document images,
Storage means for storing candidate character groups obtained as a result of recognition of characters in the input document image in association with similarities;
First calculating means for calculating a degree of coincidence between a given search keyword and a first combination in which the similarity of the recognition result is a combination of first candidate characters;
If the degree of coincidence calculated by the first calculating means is greater than or equal to a predetermined first threshold , the second and subsequent candidates in the search keyword and a candidate character group determined to be greater than or equal to the predetermined first threshold Second calculating means for calculating the degree of coincidence with the second combination including the character of
A filing apparatus comprising: a determination unit that determines that the search keyword is hit when the degree of coincidence calculated by the second calculation unit is equal to or greater than a predetermined second threshold value. The second calculation means compares the similarity of the first candidate character of the target character with a predetermined threshold,
Condition: 1st similarity score ≧ predetermined threshold
In the case of satisfying the above, a combination including a candidate character having a similarity of 1st-ranked or higher among candidate characters after the second-ranked character of interest is defined as the second combination,
If the above condition is not satisfied, a combination including candidate characters having a similarity of β-first (where α <β) or more among the candidate characters after the second-ranked character of interest is included. The second combination,
The filing apparatus according to claim 1, wherein the degree of coincidence with the search keyword is calculated. A method for controlling a filing apparatus that stores document images and uses data obtained by recognizing characters in the document images to search for document images based on a search keyword,
A storage step provided in the filing device stores a candidate character group obtained as a result of recognition of characters in the input document image in association with a similarity, and
A first calculation step in which a first calculation unit included in the filing device calculates a degree of coincidence between a given search keyword and a first combination obtained by combining candidate characters ranked first in the recognition result; ,
When the degree of coincidence calculated in the first calculation step is greater than or equal to a predetermined first threshold , the second calculation means included in the filing device determines that the search keyword is greater than or equal to the predetermined first threshold. a second calculation step of calculating the degree of coincidence between the second combination including the character of the second position subsequent candidate in the candidate character group has,
The determination means included in the filing device includes a determination step of determining that the search keyword is hit when the degree of coincidence calculated in the second calculation step is equal to or greater than a predetermined second threshold. Control method of filing device.

Images