JP2000011091A - Character recognition method, device therefor and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make it possible to properly and efficiently read out an input character string even in the case of an input pattern mixing a pattern such as a noise other than characters and having a size similar to the size of characters in a pattern for correcting a noise due to the low quality of an input system, a double line, an erasing/paint-out, and the like. SOLUTION: Plural character strings to be recognized about characters are previously registered in a word dictionary 9 together with the order of respective characters constituting these character strings, a segmentation processing part 4 segments candidate segments based on the image of character strings written in an inputted paper medium and a character recognition part 5 recognizes characters in each candidate segment by referring to an individual character recognition dictionary 8 including at least respective characters constituting the previously registered character strings and allows candidate segments having a character recognition result more than prescribed matching quantity to correspond to respective characters of plural character strings coincident with the character recognition result. A knowledge processing part 6 extracts a character string candidate consisting of the combination of candidate segments suited to the order relation of plural character strings in each plural character strings, calculates the evaluation value of each character string candidate based on the matching quantity of the extracted character string candidate and judges the character strings written in the paper medium.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition method and apparatus used in a system for reading a character or a character string written on a paper medium and sorting the paper medium based on the read result, and an apparatus and a recording medium therefor. In particular, a character recognition method and apparatus capable of properly recognizing a character or a character string based on a word dictionary of a character string registered in advance even when noise is mixed in an image read from a paper medium, and It relates to the recording medium.

[0002]

2. Description of the Related Art Conventionally, there is a character recognition system for automatically reading characters or character strings printed or handwritten at a high speed. However, in these character recognition systems, the accuracy of cutting out and reading each character is limited. Therefore, the reading accuracy is improved by using the knowledge about the context of the character string.

[0003] As these character recognition methods, for example,
"Character segmentation and recognition from handwritten character strings using linguistic information" (Hiroshi Murase et al., IEICE Transactions (D), j69-D, 9, pp.1292-1301), which is described in this paper After the input pattern is decomposed into the minimum processing units called basic segments, a search is made for a combination having the maximum character similarity obtained by character recognition from among the comprehensive combinations of the basic segments, and the word is searched for. The final character string recognition result is determined using the knowledge.

A character extracting method described in Japanese Patent Application Laid-Open No. 6-195508 also discloses a method of performing a knowledge process to recognize a character string. That is, first, characters are cut out using shape knowledge, character recognition is performed on the character cutout results, and character validity is evaluated using word-like concatenation relations. For only the area evaluated as being present, character extraction is again performed using the best-priority search method to perform an optimal pattern search, and a reliable character string is read.

[0005]

However, the above-described character recognition method is based on the premise that some character category is assigned to all segments cut out from an input pattern in which a character string is written. The size of the pattern and the shape of the aspect ratio, such as a pattern other than a character, such as noise due to poor quality of the input system, or a pattern of correction writing by double line, erase, fill, etc. When noise that cannot be removed by the feature alone is mixed, the noise is assigned as one character or a part of one character, and the character string cutout processing and reading (recognition) performance are performed. Is extremely reduced.

In this case, it is conceivable to perform character recognition processing by excluding some of the above-described noise-containing segments from the input pattern on the assumption that they are noise in sequence. When the processing is added, the character recognition processing amount becomes very large, and as a result, there arises a problem that the character recognition processing efficiency is reduced.

Therefore, the present invention eliminates such a problem,
This is an input pattern in which noise due to poor quality of the input system, or a non-character pattern such as noise close to the character size in a pattern such as double line, erase, fill, etc. It is another object of the present invention to provide a character recognition method and apparatus capable of reading an input character string properly and efficiently, and a recording medium therefor.

[0008]

According to a first aspect of the present invention, there is provided a first step in which each character string is registered in advance together with the order of each character constituting a plurality of character strings to be recognized, and a medium to be read. A second step of cutting out a candidate segment based on the image of the character string entered above, and referring to at least each character constituting the plurality of pre-registered character strings for each candidate segment, A third step of recognizing and associating a candidate segment having a character recognition result equal to or greater than a predetermined matching amount with each character of the plurality of character strings matching the character recognition result; A fourth step of extracting a character string candidate composed of a combination of the candidate segments conforming to the order relation of the character strings, and evaluating each character string candidate based on the matching amount of the extracted character string candidate. Characterized in that to calculate the value it was and a fifth step of determining character string written on the reading target medium.

According to the first aspect of the invention, noise other than characters such as noise due to low quality of the input system or noise close to the character size in a pattern such as double line, erase, fill, etc. Even if the input pattern is a character string in which these patterns are mixed, the input character string can be read properly and efficiently based on the order relation of the character strings.

According to a second aspect of the present invention, in the first aspect of the present invention, the fourth step includes, for each of the plurality of character strings, a character string comprising a combination of the candidate segments adapted to the order relation of the plurality of character strings. When extracting column candidates, it is necessary to delete candidate segments that do not conform to the order of the plurality of character strings, and to delete in advance character strings that are combinations of candidate segments less than the number of characters of the character strings as a result of this deletion. Features.

According to the second aspect of the present invention, when extracting a character string candidate, a character string having a character associated only with a candidate segment that does not satisfy the order relation of the character string is deleted from the extraction target in advance. Character string candidate extraction processing can be performed efficiently.

According to a third aspect of the present invention, in the first or second aspect of the invention, in the fourth step, as a result of extracting the character string candidates, candidate segments that are not included in the character string candidates are converted into each character segment. The fifth step is that, when calculating the evaluation value of the extracted character string candidate, the matching amount of each candidate segment constituting the character string candidate and the character string candidate The evaluation is made based on a matching amount less than the predetermined matching amount given to candidate segments not included and the size of the area of each candidate segment.

According to the third aspect of the present invention, when evaluating the extracted character string candidates, evaluation is performed including candidate segments that are not included in the character string candidates. Can be.

According to a fourth aspect of the present invention, there is provided the first to third aspects.
In the invention, the fifth step includes, if the character string candidate has an evaluation value equal to or more than a predetermined value, and the only character string candidate having an evaluation value equal to or more than a predetermined value is the only character string candidate, The determination is made as a character string entered on the medium to be read.

According to the fourth aspect of the present invention, the character string corresponding to the character string candidate is determined as a character recognition result only when the evaluation value of the character string candidate is equal to or more than a predetermined value and is unique.
Furthermore, proper character string recognition can be performed.

According to a fifth aspect of the present invention, there is provided a registration means for pre-registering each character string together with the order of each character constituting a plurality of character strings to be recognized, and an image of the character string written on a medium to be read. Extracting means for extracting a candidate segment based on the character string; character recognition means for recognizing a character by referring to at least each character constituting the plurality of character strings registered in advance for each candidate segment; Means for associating a candidate segment having a character recognition result equal to or greater than the amount with each character of the plurality of character strings that match the character recognition result, and matching the order relationship of the plurality of character strings for each of the plurality of character strings Extracting means for extracting a character string candidate composed of a combination of the candidate segments to be calculated; and calculating an evaluation value of each character string candidate based on a matching amount of the extracted character string candidates, and Characterized by comprising a determining means for determining character string written on the body.

Thus, the same effects as those of the first to fourth aspects of the invention are obtained.

According to a sixth aspect of the present invention, there is provided a first step of pre-registering a plurality of character strings to be subjected to character recognition together with the order of the characters constituting the character strings; A second step of extracting candidate segments based on the image of the column, and character recognition for each candidate segment by referring to at least each of the characters constituting the plurality of pre-registered character strings; A third step of associating a candidate segment having the above character recognition result with each character of the plurality of character strings matching the character recognition result; and, for each of the plurality of character strings, an order relationship between the plurality of character strings. A fourth step of extracting a character string candidate comprising a combination of the matching candidate segments, and calculating an evaluation value of each character string candidate based on the amount of matching of the extracted character string candidates. Characterized in that it is a storage medium storing software and a fifth step of determining character string written on the target medium collected.

Thus, the same effects as those of the first to fifth aspects of the invention can be obtained.

[0020]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a schematic configuration diagram of a sorting system using a character recognition device according to an embodiment of the present invention. In FIG. 1, a plurality of paper media to be sorted whose free handwritten characters are written on the surface thereof are set in the feeding device 1.
The feeding device 1 sequentially feeds paper media one by one to the character recognition device 2.

Under the control of the control unit C, the character recognition device 2 reads the input image of the paper medium, performs character recognition, and outputs a character string recognition result to the sorting device 7. This character recognition processing will be described later.

The sorting device 7 sorts the paper medium conveyed through the character recognition device 2 based on the result of character recognition. Here, if the character recognition result is not unique, the sorting device 7 rejects the paper medium. This reject is also one sort.

The free handwritten characters written on the paper medium are a plurality of types of character strings set in advance. The individual characters constituting the character strings are registered in the individual character recognition dictionary 8 and , And the character string are registered in the word dictionary 9 together with the character arrangement order. This registration process
The image is read by the image input unit 3 under the control of the control unit C and registered in the word dictionary 9 and the individual character recognition dictionary 8. The control unit C, the image input unit 3, the word dictionary 9, and the individual character recognition dictionary 8 correspond to a registration unit.

FIG. 2 is a flowchart showing a character reading processing procedure of the character recognition device 2. FIG. 3 is a diagram showing a specific example for explaining processing up to individual character recognition processing in this character reading processing procedure. Here, the character reading process is performed by four words (character strings), that is, “Akame Himeji”,
The processing will be described on the assumption that any one of "Ako Hyotaro", "Hyogo Ichiro", and "Himeji Kazuko" is entered on a paper medium. Then, as shown in FIG. 3A, it is assumed that a noise pattern corresponding to “one” is mixed between the images of “road” and “red”. This noise pattern often occurs when there is a fold in the paper medium or in a low-quality image in which the difference in density between the character and the background is small.

In FIG. 1 to FIG. 3, first, paper media are fed out one by one from the feeding device 1, and the entire image of one paper medium is captured by the image input unit 3 of the character recognition device 2 (S
1) The image input unit 3 performs the inclination correction to extract the area of the character string portion of the captured whole paper medium image (S).
2). This skew correction process is a process for correcting the skew of the paper medium when the paper medium is transported, detects the skew angle of the edge of the paper medium, and rotates the image of the character string region based on the skew angle. to correct. Further, the image input unit 3
A process of binarizing the extracted character string area is performed (S3). For example, when a free handwritten character string as shown in FIG. 3A is entered in a certain area of a paper medium, the image input unit 3 extracts an area of a free handwritten character string portion corresponding to “Akako Himeji”. Then, a binarized image E1 is generated.

The binarized image E1 is input to the cutout processing section 4, and the cutout processing section 4 performs a cutout processing of the binarized image E1. Here, small black pixel areas whose height and width are less than the predetermined values are removed. However, a pattern corresponding to “one” between “road” and “red” is not removed because it is equal to or greater than a predetermined value. In the clipping process, first, a Y-axis projected feature extraction process is performed as shown in FIG. This Y-axis projected feature extraction processing is performed because the character string is written vertically. The Y-axis projection is a histogram of the distribution of black pixels as viewed from the Y-axis direction. The cutout processing unit 4 creates a basic segment based on the Y-axis projection E2 subjected to the Y-axis projection feature extraction processing (S4). The basic segment is determined as a unit of Y-axis projection, that is, a portion where a histogram of black pixel distribution is continuous. For example, the Y-axis projection E2 in FIG. 3B can be divided into six groups of partial Y-axis projections E21 to E26, and the basic segments BS1 to BS6 corresponding to the respective partial Y-axis projections E21 to E26 are basic. Determined as a segment. The subscript i of the basic segment BSi is sequentially assigned from the smaller Y axis value (from the top).

Further, the cutout processing section 4 merges the adjacent basic segments BS1 to BS6 to perform a candidate segment creation process (S5). The criteria for merging the basic segments are as follows. That is, (1) the size in the Y-axis direction after the merge is not more than a predetermined multiple P1 of the size in the X-direction after the merge, and (2) the size in the Y-axis direction after the merge is not more than the fixed value P2. (3) The sum of the areas of the basic segments used for merging is equal to or less than a predetermined percentage P3 of the total area of all the basic segments. (4) The distance between the basic segments to be merged is all the basic segments. It is determined on the basis of being equal to or less than the average of the distance between them. Note that it is required that all of these criteria be satisfied, and the values of P1 to P3 are set in advance. Both the new merged segment and the base segment are candidate segments.

Based on such a criterion, for example, FIG.
Candidate segments CS1 (1,1) to CS8 shown in (c)
Eight candidate segments (4, 5) are created. The candidate segments CS1 (1,1) to CS6 (6,6) correspond to the basic segments BS1 to BS6 and have the same contents. In addition, candidate segment 7 (3,
4) is a basic segment BS3 and a basic segment BS4
Are merged, and the sign in parentheses of the candidate segment is shown as (3, 4) in that sense. The first value “3” in parentheses is the basic segment BS3
Is the value of “3”, and the last value “4” in parentheses is the value of “4” of the basic segment BS4. That is, the first value in parentheses of the candidate segment indicates the number of the first merged basic segment, and the last value in parentheses indicates the number of the last merged basic segment. Candidate segment CS8 (4,5) is the basic segment BS
4 and BS5. Although not present in FIG. 3, for example, when the value of the parenthesis of the candidate segment is (4, 6), the basic segments BS4 and BS4
5 and BS6 have been merged. Hereinafter, the value in the parentheses is indicated as (starti, endi), which indicates the merged range. However, starti
And the value of i of endi may be different like CS7 or CS8.

Next, the cutout processing section 4 generates candidate segments CS1 (1,1) to CS5 based on the merge criteria.
CS8 (4, 5) is sent to the character recognition unit 5, and the character recognition unit 5 performs individual character recognition for each candidate segment based on the individual character recognition dictionary 8 (S6). In the individual character recognition dictionary 8, all types of characters constituting at least the four words described above are registered. The character recognizing unit 5 performs a matching process between the registered characters and each candidate segment, and sets a matching amount. Is associated with a character that is equal to or greater than a predetermined value as a possible character corresponding to each candidate segment. Note that, as the registered character, at least a character used for a plurality of character strings corresponds, but it is preferable to further register various characters including this character. The type of character and the number of registrations may be set arbitrarily according to the system. For example, in FIG.
As a candidate segment CS1 (1, 1) having a matching amount equal to or greater than a predetermined value, “hime”, “ro”,
This indicates that the four characters “Koko” and “Soldier” have been selected. The number of characters that have a matching amount
Assuming J, here J is 4. Note that the matching amount in this case is “princess” → “roo” → “storage” → “soldier”. This order is represented by CODEj using “j” (j = natural number equal to or less than J), and the matching amount is represented by σj
Can be shown as For example, CODE1 (CS
1) is “princess”, and σ2 (CS1) indicates a matching amount of “ro”.

Next, each candidate segment and the individual character recognition result are output to the knowledge processing unit 6, and the knowledge processing unit 6
Based on the correspondence and the word dictionary 9, an optimal combination of candidate segments is selected to recognize a freely handwritten character string (S7). The outline is to generate a correspondence table 6a in which the relationship between the character recognition results with respect to the candidate segments and the correspondence relationship between the candidate segments with respect to the words (character strings). The candidates are extracted, and the evaluation / determination unit 6c evaluates each character string candidate based on the matching amount as the character recognition result, and recognizes the character string based on the evaluation result. Details of this knowledge processing will be described later. Note that the extraction processing unit 4 corresponds to the extraction unit, the character recognition unit 5 corresponds to the character recognition unit, the character string candidate extraction unit 6b corresponds to the extraction unit, the evaluation / determination unit 6c corresponds to the determination unit, and the knowledge processing unit 6 corresponds to the extraction unit. Each corresponds to a means.

The result recognized by the knowledge processing section 6 is output to the sorting device 7 (S8), and the sorting device 7 sorts the conveyed paper medium based on the recognition result.

It is determined whether or not all the paper media have been processed (S9). If there is still a paper medium to be processed, the process proceeds to step S1, and the above-described steps S1 to S1 are performed.
The processing of S8 is repeated, and when the processing of all the paper media is completed, this processing is completed. Incidentally, actually, step S5
The next paper medium is fed out in the middle of steps S8 to S8, and the entire image is captured (S1).

Next, the knowledge processing in step S7 will be described in detail with reference to FIGS. Hereinafter, the description will be made in the order of the knowledge processing procedure of step S7 shown in the flowchart of FIG. 4. First, the initialization processing is performed before this processing. That is, an index table shown in FIG. 6 is created.
This index table is created based on the word dictionary 9 shown in FIG. 5. For all the characters constituting the word dictionary 9, each character is the word of the word dictionary 9 (that is, the word number), and The number of the character (that is, the character number) is shown as a pair. For example, “princess” is the first character of the first word in the word dictionary 9 (word number =
1, character number = 1) and the first character of the fourth word (word number = 4, character number = 1). Therefore, in the index table, (1,1), ( 4,
The address information of 1) is associated.

(1) Creation of Correspondence Table (Step S11) The knowledge processing unit 6 creates a correspondence table as shown in FIG.
This correspondence table is a work sheet indicating, for each word, which candidate segment each character constituting the word corresponds to.

First, when the candidate segment CSi (i = a natural number, the number up to the number of candidate segments) has a character CODEj as a result of character recognition, the correspondence table is prepared by referring to the index table and referring to the index table. Character CODE
The subscript i of the candidate segment CSi is written in the row j.

For example, the character CODE1 = “princess”, which is one of the character recognition results of the candidate segment CS1, is the information (1,1),
Based on (4,1), the subscript “1” of CS1 is added to the first character row of word number 1 and the first character of word number 4 in the correspondence table.
Write. CODE2 of candidate segment CS1 =
For “ro”, the subscript “1” of CS1 is written in the fifth character of word number 2 and the fourth character of word number 3. Similarly, for the character CODE3 = “Koko” of the candidate segment CS1, the subscript “1” of CS1 is added to the second character of the word number 3.
Write. Finally, the character CO of the candidate segment CS1
For DE4 = “Soldier”, the subscript “1” of CS1 is written in the third character of word number 2 and the first character of word number 3. Thus, the process for the candidate segment CS1 is completed, and similarly, the other candidate segment CS1 is left.
Perform the process of i.

Here, the numbers in the corresponding candidate segment column in FIG. 7 indicate the number of the candidate segment corresponding to the character in the word for each word, and the number of the candidate segment CSi (starti, end
When the value of “i” of “starti” in i) is smaller than the character number, such a correspondence cannot exist, so the subscript of this candidate segment is excluded from the correspondence table in advance. FIG.
In the example, a cross mark is added to the subscript of the corresponding candidate segment to be excluded. This reduces post-processing.

(2) Rejection of Word Candidates Containing Unsupported Characters (Step S12) As a result of the creation of the correspondence table in step S11, each character of each word has a plurality of corresponding candidate segments, or has no corresponding segment. There is. For example, three candidate segments correspond to the third character “red” of word number 1 in FIG. 7, and the fifth character “ro” of word number 2 has no corresponding candidate segment (see above). (Because it is excluded in (1)).
If the corresponding candidate segment contains no characters, it is determined that such a word will not be described. For example, since there is no candidate segment corresponding to the fifth character "ro" of "Ako Hyotarou" of word number 2, the word of "Ako Hyotarou" is recognized as not being described, and the word is recognized. Excluded from candidates. In general,
When the number of characters in which the corresponding candidate segment exists is smaller than the number of words, the word is excluded from the recognition candidates. FIG. 8 is a diagram showing such a correspondence.
Words with word numbers 2 and 3 are excluded from the recognition candidates. Therefore, by this processing, word number 1
And the word of word number 4 are recognition candidates, and any of the words is a word described on a paper medium.

(3) Checking the Order Relationship (Step S1)
3) Next, it is checked whether or not the order relation between corresponding candidate segments is established between adjacent characters. The order of the letters in the word is already listed in the correspondence table in the order of the strings,
Subscript values indicate the order of the corresponding candidate segments, and the order relation is checked based on these orders and values. At this time, considering that the above-described noise is mixed, the conditions for establishing the order relation are as follows in specific calculations. That is,
"Endj1 <startj2 is established between the candidate segment CSj1 (startj1, endj1) corresponding to the i-th character and the candidate segment CSj2 (startj2, endj2) corresponding to the (i + 1) -th character in a word. That ". For example, between the candidate segment CS1 (1,1) corresponding to the first character “princess” in the word number 1 in FIG. 7 and the candidate segment CS2 (2,2) corresponding to the second character “ro”. , End1 = 1 <start2 = 2, and the order relation is established. Also, between the candidate segment CS2 (2,2) corresponding to the second character “Ro” in word number 1 and the candidate segment CS8 (4,5) corresponding to the third character “Red”, end2 = Since 2 <start8 = 4, the order relation is established. However, between the candidate segment CS2 corresponding to the first character “hime” in the word number 1 and the candidate segment CS2 corresponding to the second character “ro”, en
d2 = 2 = start2 = 2, and the order relation is not established.

As a result, a link that satisfies the order relation as shown by the broken line in FIG. 7 is created.

(4) Creation of Linked List and Extraction of Unsupported Candidate Segments (Step S14) A list is created by connecting only the candidate segments whose permutation relationship is OK by the process of step S13, and this is called a linked list. As shown in FIG. 9, the linked list CL1_
1 to CL1_3 and CL4_1 to CL4_3 are generated. Then, for each linked list, candidate segments that are not used, that is, unsupported candidate segments are extracted. However, the unsupported candidate segments are selected from those having one basic segment.
For example, the linked list CL1_3 includes the candidate segment CS
1, CS2, CS8, and CS6 and does not include the candidate segments CS3, CS4, CS5, and CS7, but does not set all of the candidate segments CS3, CS4, CS5, and CS7 as unsupported candidate segments, and only CS3 is unsupported. Select as a candidate segment. This is because CS3, CS4, and CS5 are composed of one basic segment. Since CS4 and CS5 are used in the candidate segment CS8, only the candidate segment CS3 needs to be unsupported after all.

(5) Checking the arrangement relation (step S1)
5) With respect to the character strings constituting each linked list, the arrangement relationship is checked based on the width or height of the circumscribed rectangle, the position of the center of gravity, and the like. For a character string having a poor arrangement balance such as non-uniformity, processing for excluding the linked list is performed. This processing can be achieved by using a known technique. In the example of the linked list (FIG. 9) created from the embodiment shown in FIG. 3, nothing is excluded.

(6) Calculation of Evaluation Amount (Step S16) The evaluation amount for each of the linked lists remaining in the processing up to step S15 is calculated. Then, the linked list having the maximum value in each word is set as the evaluation value for the word and the correspondence between the candidate segments. The evaluation amount T of each linked list is given by the following equation. That is, It is. Here, Ω1 is a set of candidate segments forming the linked list, and Ω2 is a set of unsupported candidate segments not included in the linked list. Also, L (C
S) is an amount representing the size of the candidate segment CS, for example, the height of CS or the number of black pixels. Si (CS)
Is the matching amount when the candidate segment CS is read with the i-th character of the word (the i-th character in the linked list).
The second candidate segment and CS match). Also, N
(CS) is an evaluation value when the candidate segment CS is regarded as noise. For example, a constant value smaller than the matching amount when the candidate segment CS is recognized as any character of the word, or σJ + 1 (CS) is Evaluation value.

Here, the linked list CL1_1 shown in FIG.
For example, the linked list is composed of candidate segments CS1, CS2, CS4, and CS6, and the unsupported candidate segment is candidate segment C.
S3 and CS5. Therefore, when applied to the above expression of the evaluation amount T, the first term of the numerator is S1 (CS1) L
(CS1) + S2 (CS2) L (CS2) + S3 (CS
4) L (CS4) + S4 (CS6) L (CS6). S1 (CS1) is equal to the matching amount σ1 (CS1) when the candidate segment CS1 is read as “Princess” of the first character of “Himeji Akako” of word number 1. That is, S1
(CS1) = σ1 (CS1). Similarly, S2
(CS2) = σ1 (CS2), S3 (CS4) = σ1
(CS4), S4 (CS6) = σ1 (CS6).

The second term of the molecule is N (CS3) L
(CS3) + N (CS5) L (CS5) and N (C
S3) is an evaluation amount when the candidate segment CS3 is regarded as noise, for example, σ3 (CS3). This is C
This is because the value of “J” in S3 is up to 2, and σJ + 1 = σ3. That is, since the candidate segment CS3 is regarded as noise, the third matching amount for which the character recognition result did not exceed the predetermined value is used. Similarly, N (CS
5) = σ3 (CS5).

Therefore, the evaluation amount T for the linked list CL1_1 is as follows. That is, T = (σ1 (CS1) L (CS1) + σ1 (CS2) L
(CS2) + σ1 (CS4) L (CS4) + σ1 (CS
6) L (CS6) + σ3 (CS3) L (CS3) + σ3
(CS5) L (CS5)) / (L (CS1) + L (CS
2) + L (CS4) + L (CS6) + L (CS3) + L
(CS5)). In this way, the evaluation amount can be obtained based on the matching amount and the amount indicating the size of the candidate segment. As a result, in the linked list shown in FIG. 9, the linked list CL1_3
= The pair of (CS1, CS2, CS8, CS6) and the unsupported candidate segment CS3 is selected as having a large evaluation amount, and the linked list CL4_2 = (CS1, CS2, CS3, CS6) and the unsupported candidate segments CS4, CS5 are selected as having a large evaluation amount.

(7) Determination of Word Recognition Result (Step S1)
7) When the value of the evaluation amount T of the linked list selected in step S16 is the only linked list having a predetermined threshold or more, the words of this linked list are output as a recognition result, and the predetermined threshold is output. If there is no linked list having the above, it is rejected as having no read word candidate, and if there are two or more linked lists having a predetermined threshold value, it is rejected as undeterminable.

From the linked list shown in FIG. 9, it is assumed that linked list CL1_3 has an evaluation value equal to or greater than a predetermined threshold value and linked list CL4_2 does not have an evaluation value equal to or greater than a predetermined threshold value. Shows "Akako Himeji"
Is output as the recognition result, and "Himeji Kazuko" indicated by the linked list CL4_2 is not adopted as the recognition result, and as a result, "1" is interpreted as a noise pattern.

As described above, according to the knowledge processing according to the present embodiment, it is possible to reduce the quality of the input system in which the fold of the paper medium and the difference in shading between the character and the background are small, such as the above-mentioned "1". Even if the input pattern includes noise caused by the noise, the input character string can be read properly and efficiently.

Not only the noise pattern due to the low quality of the input system as described above, but also a pattern such as a double line, a correction, such as erase, fill, etc. Even an input pattern in which patterns other than the above are mixed can be applied to the above-described knowledge processing.

Further, in the above-described embodiment, the description has been made as a vertically written character string. However, it is apparent that the present invention can be applied to a horizontally written character string. In this case, X-axis projected feature extraction processing may be performed. Furthermore, although the basic segment is determined using only the projected feature extraction, the basic segment is determined based on the analysis of the connected components of the black pixels, or the basic segment is determined by various methods such as a method of fusing these. Is also good.

Further, even if the length and the number of words in the word dictionary increase, the above-described knowledge processing only increases in a linear order, and there is no large increase in the processing amount. The time required for media sorting does not increase.

[0054]

As described in detail above, according to the first to sixth aspects of the present invention, the noise caused by the low quality of the input system or the correction writing such as double line, erase, fill, etc. Even if the input pattern is a character string in which non-character patterns such as noise close to the character size are mixed in the pattern, the character string input properly and efficiently based on the order relation of the character strings This has the effect of being able to read.

In particular, since character string candidates are extracted based on such an order relationship, even if the number of characters in a character string increases, the number of combinations of character string candidates increases in a linear order. However, the processing amount does not suddenly increase with an increase in the number of characters in the character string, and the effect is that efficient knowledge processing can be achieved.

According to the second aspect of the present invention, when extracting a character string candidate, a character string having a character associated with only a candidate segment that does not satisfy the order relation of the character string is deleted in advance from the extraction target. Thus, there is an effect that extraction processing of a character string candidate can be efficiently performed.

Further, according to the third aspect of the present invention, when the extracted character string candidates are evaluated, evaluation is performed including candidate segments that are not included in the character string candidates. It has the effect that it can be performed.

According to the fourth aspect of the present invention, the character string corresponding to the character string candidate is determined as a character recognition result only when the evaluation value of the character string candidate is equal to or more than a predetermined value and is unique. This has the effect that it is possible to recognize a simple character string.

[Brief description of the drawings]

FIG. 1 is a diagram showing a schematic configuration of a sorting system using a character recognition device according to an embodiment of the present invention.

FIG. 2 is a flowchart showing a character reading processing procedure of the character recognition device 2.

FIG. 3 is a diagram illustrating a specific example for explaining processing up to individual character recognition processing in a character reading processing procedure.

FIG. 4 is a flowchart showing a knowledge processing procedure in step S7.

FIG. 5 is a diagram illustrating an example of a word dictionary.

FIG. 6 is a diagram illustrating an example of an index table.

FIG. 7 is a diagram showing an example of a correspondence table.

FIG. 8 is a diagram illustrating an example of rejection of a word including an unsupported character.

FIG. 9 is a diagram showing a set of a linked list and unsupported candidate segments.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Feeding device 2 ... Character recognition device 3 ... Image input part 4
... Extraction processing unit 5 Character recognition unit 6 Knowledge processing unit 6a Correspondence table 6b
... Character string candidate extraction unit 6c ... Evaluation / judgment unit 7 ... Sorting device C ... Control unit

Claims (6)

[Claims] A first method in which each character string is registered in advance together with the order of each character constituting a plurality of character strings to be recognized.
And a second step of cutting out candidate segments based on the image of the character string written on the medium to be read; and forming at least the plurality of character strings registered in advance for each candidate segment. A third step of performing character recognition with reference to each character and associating a candidate segment having a character recognition result equal to or greater than a predetermined matching amount with each character of the plurality of character strings that match the character recognition result; A fourth step of extracting a character string candidate composed of a combination of the candidate segments that conforms to the order relation of the plurality of character strings for each of the character strings of: A fifth step of calculating an evaluation value of a character string candidate to determine a character string written on the medium to be read. 2. The method according to claim 1, wherein the extracting the character string candidate comprising a combination of the candidate segments that conforms to the order relation of the plurality of character strings for each of the plurality of character strings. 2. The character recognition method according to claim 1, wherein candidate segments that do not conform to the order are deleted, and a character string that is a combination of candidate segments less than the number of characters of the character string as a result of the deletion is deleted in advance. . 3. The method according to claim 3, wherein, as a result of extracting the character string candidates, candidate segments not included in the character string candidates are held in association with each of the character string candidates. When calculating the evaluation value of the extracted character string candidate, the matching amount of each candidate segment constituting the character string candidate is less than the predetermined matching amount given to the candidate segments not included in the character string candidate. The character recognition method according to claim 1 or 2, wherein the evaluation is made based on the matching amount of (i) and the size of the area of each candidate segment. 4. The method according to claim 1, wherein the character string candidate has an evaluation value equal to or more than a predetermined value, and when the character string candidate having the evaluation value equal to or more than a predetermined value is the only one, The character recognition method according to any one of claims 1 to 3, wherein the character string is determined as a character string written on the medium to be read. 5. A registration means for pre-registering each character string together with the order of each character constituting a plurality of character strings to be recognized, and a candidate based on an image of the character string written on a medium to be read. Segmenting means for extracting a segment; character recognizing means for recognizing a character by referring to at least each character constituting the plurality of pre-registered character strings for each candidate segment; character recognizing more than a predetermined matching amount Means for associating a candidate segment having a result with each character of the plurality of character strings that matches the character recognition result; and for each of the plurality of character strings, the candidate segment matching the order relationship of the plurality of character strings. Extracting means for extracting a character string candidate consisting of a combination; calculating an evaluation value of each character string candidate based on a matching amount of the extracted character string candidates; Character recognition apparatus characterized by comprising a determination means for determining character string. 6. A first step of pre-registering a plurality of character strings to be subjected to character recognition together with the order of characters constituting the character strings, and an image of the character strings written on a medium to be read. A second step of cutting out the candidate segments into: and character recognition for each candidate segment with reference to at least each of the characters constituting the plurality of pre-registered character strings, and a character recognition result equal to or more than a predetermined matching amount. A third step of associating a candidate segment having the following with each character of the plurality of character strings that matches the character recognition result; and, for each of the plurality of character strings, the candidate segment that conforms to the order relationship of the plurality of character strings. A fourth step of extracting a character string candidate consisting of a combination of the following: calculating an evaluation value of each character string candidate based on the amount of matching of the extracted character string candidates; Recording medium storing software and a fifth step of determining entry string.