JPH04322390A - Character recognition device - Google Patents

Abstract

PURPOSE:To avoid the degradation of the recognition rate by accurately determining a character frame even in the case of coupling of blocks of separated characters. CONSTITUTION:Character validity information 34 is preliminarily prepared, and it is discriminated in a character validity discriminating part 31 whether a black point block in a character frame 29 meets character validity information 34 or not. If it does not meet this information 34, a threshold to generate a binarized pattern is changed by the control of a threshold change control part 33, and discrimination in the character validity discriminating part 31 is repeated. Thus, erroneous recognition due to coupling of black point blocks or a lack of the space between them is reduced.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition device for recognizing characters written on a medium.

0002

2. Description of the Related Art In so-called forms such as checks and slips handled by financial institutions, various information is displayed using characters of a predetermined standard called E13B characters. This E13B character is a numeric character (0 to 9, 10
It consists of a character) and a separator character indicating a symbol (4 characters) other than a numerical value. Furthermore, optical character readers (OCR) are widely used to automate the reading of E13B characters on forms.

FIG. 2 shows a conceptual diagram of a general optical character reading device. The image sensor 2 of the optical character reading device shown in the figure is composed of an optical sensor etc. that acquires image data corresponding to the characters on the form 1 and converts it into an electrical signal. It outputs multi-value data (gradation data indicating density) for each time.

The multi-value pattern memory 3 is composed of a RAM and the like that stores multi-value data output from the image sensor 2 for each dot read by the image sensor 2. Note that the storage format of the multi-value data in the multi-value pattern memory 3 is such that the multi-value data is stored in two formats so as to correspond to the positions on the form 1, that is, the positions on the form 1 of the dots constituting the image data.
stored dimensionally. The binarization circuit 4 is composed of a comparator and the like that divides the multi-value data stored in the multi-value pattern memory 3 into two parts (determining black and white) using a predetermined threshold set in advance and outputs binary data. It is something.

The binary pattern memory 5 is composed of a RAM etc. that stores the binary data output from the binarization circuit 4, and corresponds to the storage format of the multi-value pattern memory 3.
A two-dimensional storage format is used. The X address register 6 is a register indicating the address of the binary pattern memory 5 in the X direction. The Y address register 7 is a register indicating the address of the binary pattern memory 5 in the Y direction. The extraction control unit 8 controls the threshold value of the binarization circuit 4 and
It controls the extraction of predetermined binary data stored in the value pattern memory 5. Note that the predetermined binary data refers to binary data stored in the pattern memory 5 corresponding to a preset position of the E13B character on the form 1.

[0006] The format table 9 is provided by the binarization circuit 4.
This table is made up of a ROM, etc., in which various information for specifying the threshold value and the binary data extraction area on the binary pattern memory 5, that is, the writing position of the E13B character on the form 1, is set. The recognition unit 10 recognizes the content of the E13B character extracted under the control of the extraction control unit 8 and outputs a predetermined character code. The main control unit 11 performs editing processing such as displaying the recognition result of the recognition unit 10 on a display device (not shown).

FIG. 3 shows a conceptual diagram of the form 1. As shown in the figure, the position where the E13B character is written on the form 1 is set to the right of the position indicated by the vertical character position 20 and the horizontal character position 21 from the left edge P of the form 1. . It is assumed that the interval at which the E13B characters are arranged is character interval 22, and the height of the characters is character height 23. In other words, E23B characters with a character height of 23 are written at a character interval of 22 based on the vertical character position 20 and the horizontal character position 21.

FIG. 4 shows a conceptual diagram of the binary pattern memory 5. The figure shows the content and arrangement of binary data stored in the binary pattern memory 5 as a result of reading the form 1 explained in FIG. A storage format is adopted that allows the user to know the location on the top of the screen. The cutout control unit 8 determines the cutout area in order to cut out the area storing the binary data corresponding to the E13B character extracted by the recognition unit 10 from the binary pattern memory 5.

That is, the cutting control unit 8 determines the upper limit 25 in the binary pattern memory 5 based on the vertical character position 20 and character height 23 stored in the format table 9.
and determine the lower limit 26. After that, the cutting control section 8
Based on the horizontal character position 21 and character spacing 22 stored in the format table 9, the left limit 27 and the right limit 2 are
8 is determined, and an area in which binary data corresponding to one E13B character is stored is extracted. Furthermore, the cutout control unit 8 reduces the cutout area surrounded by the upper limit 25, lower limit 26, left limit 27, and right limit 28, and determines a character frame 29 that accommodates one E13B character. When the determination of the character frame 29 is completed, the processing for one E13B character by the extraction control unit 8 is completed, and then the recognition unit 10 performs character recognition within the character frame 29.

By the way, the cutout control section 8 corrects the left limit 27 and the right limit 28 in order to accurately determine the character frame 29 even when the printing position of the E13B character on the form 1 is shifted. FIG. 5 shows an explanatory diagram of correction by the cutout control section 8. The cutting control unit 8 controls the width W of the leftmost block (component of the E13B character) in the character frame 29.
is measured, and this width W is a preset width threshold TH.
, the position of the character frame 29 was determined. Further, if the width W is less than or equal to the width threshold TH, the character frame 29
Reset the position of , for example, move it to the right.

FIG. 6 shows an explanatory diagram of the width threshold TH. As shown in the figure, the separation character D1 is divided into three blocks BL1
, BL2, and BL3. The width of the block BL1 of the separated character B1 is a, the width of the numerical character S1 indicating the numerical value "1" is b, and the width threshold TH is (a+b)/c
This is the value obtained by . Note that c is a value selected as appropriate, and is 2 in this case.

0012

However, depending on the printing density of E13B characters on form 1, blocks may be combined. When the blocks are combined, it is determined that the blocks are not separate character blocks but numeric characters, and the correction of the character frame 29 is no longer executed. For this reason, a problem has arisen in that recognition by the recognition unit 10, which is performed subsequently, cannot be performed, leading to a decrease in the recognition rate. The present invention has been made with attention to the above points, and it is possible to accurately determine a character frame even when blocks of separated characters are combined, and to avoid a decrease in recognition rate. This was done for the purpose of providing a recognition device.

[0013]

[Means for Solving the Problems] A character recognition device of the present invention recognizes a character composed of a plurality of black dot blocks written on a medium.
a character validity determination unit that inspects dimensional features and determines whether the black dot block constitutes the character; and a case where it is determined that the black dot block constitutes the character; , a cutout position control unit that specifies a cutout position of a character frame for cutting out an area in which the characters are written from the medium; The image data processing apparatus includes a threshold value change control unit that changes the threshold value when reading and converting into image data.

[0014]

[Operation] Character validity information is prepared in advance, and the character validity determination section determines whether a black dot block within a character frame corresponds to the character validity information. If it is not applicable, the threshold value for generating the binarized pattern is changed under the control of the threshold value change control unit, and the determination by the character validity determination unit is repeated. This makes it possible to reduce erroneous recognition due to the combination of black dot blocks or insufficient spacing.

[0015]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The character recognition device of the present invention has the same components as the character recognition device described above with reference to FIG. FIG. 1 shows a schematic diagram of a character recognition device according to the present invention. The figure shows the cutout control section 8 previously explained in FIG.
and a format table 9 are shown. The cutout control section 8 is provided with a character validity determination section 31, a cutout position control section 32, and a threshold value change control section 33. The character validity determination unit 31 uses a format table 9, which will be explained later.
The character frame 29 extracted by referring to the character validity information 34 of
This is to determine whether the black dot block within is valid as an E13B character.

The cutout position control section 32 controls changing the position of the character frame 29. The threshold value change control unit 33 controls the threshold value of the binarization performed in the binarization circuit 4. Format table 9 contains
Character validity information 34 is provided. The character validity information 34 is information used when the character validity determination unit 31 determines whether a black dot block within the character frame 29 is valid as an E13B character.

FIG. 7 shows an explanatory diagram of the character validity information 34. As shown in the figure, for each character that makes up the E13B character, the horizontal (result of vertical scanning) and vertical (
This numerical value is a numerical representation of the feature (result of horizontal scanning), and this numerical value will be referred to as the intersection feature here. For the intersection feature, raster scanning is performed for each E13B character, and the number of transition points from a white point to a black point (black point block), that is, the number of edge points, is counted.

For example, in the case of the numerical character "0", results such as "1, 2, 1" can be obtained in both the horizontal and vertical directions. For the numerical character “5”, the horizontal intersection feature is “2, 3, 2” and the vertical intersection feature is “2, 3, 2”.
1". Similarly, in the case of the separated character D1, the horizontal intersection feature is "1, 0, 2", and the vertical intersection feature is "1, 2, 1, 2, 1". Separated character In the case of D3, the intersection feature in the horizontal direction is "1, 0, 1, 0, 1", and the intersection feature in the vertical direction is "1, 3, 2". The intersecting features are obtained and stored in the format table 9 as character validity information.

A method for calculating intersection features will be explained with reference to FIGS. 8 and 9. FIG. 8 is a block diagram of the character validity determination unit 31 according to the present invention. As shown in the figure,
The character validity determination unit 31 is provided with a scan control unit 41, an edge counter 42, an old edge register 43, a comparator 44, a horizontal crossing feature storage circuit 45, and a vertical crossing feature storage circuit 46. The scan control unit 41 controls the X address register 6 and the Y address register 7 to read the binary pattern memory 5 and store the intersection feature, that is, a horizontal intersection feature storage circuit 45 and a vertical intersection feature storage circuit. 46 switching is performed.

The edge counter 42 is a counter that counts the number of edges calculated from the read result of the binary pattern memory 5. The old edge register 43 is a register that stores the counting results that have already been stored in the edge counter 42 when the edge counter 42 performs new counting. The comparator 44 compares the contents of the edge counter 42 and the old edge register 43. The horizontal direction intersection feature storage circuit 45 is made up of a register and the like that stores the horizontal direction intersection feature output from the comparator 44. The vertical crossing feature storage circuit 46 is made up of a register and the like that stores the vertical crossing feature output from the comparator 44.

FIG. 9 shows an explanatory diagram of the operation of the character validity determining section. Here, an example will be explained in which the vertical intersection feature of the numerical character "0" is determined. Numerical character “0” is 8×
It is assumed that 8 pixels are stored in the binary pattern memory 5. When obtaining a vertical intersection feature, the main scan of the raster scan is taken in the vertical direction, and the sub-scan is taken in the horizontal direction. (When determining the intersection feature in the horizontal direction, the scanning direction is reversed.) First, the scanning control unit 41 sets the value of the Y address register 7 to the first row in the main scanning direction. Thereafter, the value of the X address register 6 is incremented from the first column to the eighth column in the sub-scanning direction.

As a result, the edge counter 42 counts the change from a white point to a black point for each line (step S1).
, is output to the comparator 44 (step S2). Since there is no change in the first row of numerical characters "0", the output of the edge counter 42 becomes 0. The output of the edge counter 42 is stored in the old edge register 43 and held until the edge counter 42 outputs a new value.

The old edge register 43 outputs the stored contents to the comparator 44 before storing the output of the edge counter 42 . In the comparator 44, the edge counter 42
The latest output is compared with the old output one scan before the latest output. When the edge counter 42 outputs the first row count value "0", the comparator 44 compares the initial value (invalid value) of the old edge register 43 with the first row count value.

When the edge counter 42 outputs the second row count value "1", the comparator 44 outputs the first row count value "0" stored in the old edge register 43 and the second row count value. "1" will be compared. Comparator 44
In the comparison operation in , first, if the outputs of the edge counter 42 and the old edge register 43 match, invalid data is output. If the outputs of the edge counter 42 and the old edge register 43 do not match, the value accepted from the edge counter 42 is output (step S4). In this case, the output of the comparator 44 is stored in the vertical cross feature storage circuit 46 under the control of the scan control section 41.

The vertical intersection feature storage circuit 46 deletes invalid data and the first and last data whose value is "0" from the output of the comparator 44, and determines the final vertical intersection information. (Step S5). In the case of the numerical character "0", the vertical intersection feature is "1, 2, 1" and is stored as character validity information 34 in the format table 9. A similar calculation is performed when determining the horizontal intersection feature of the numeric character "0", and calculations are also performed in advance for other characters, and the character validity information 34 shown in FIG. 5 is prepared in the format table 9. be done.

Next, a case will be described in which the E13B characters printed on form 1 are recognized. 10 to 12 are first to third flowcharts according to the present invention. It is assumed that the binary pattern memory 5 stores the reading result of the form 1. As shown in FIG. 10, the threshold change control unit 33 sets a preset reference binarization threshold in the binarization circuit 4 (step S11). Thereafter, the character validity determination unit 31 is activated and performs vertical and horizontal intersecting feature calculation (first block feature extraction) for the leftmost black dot block within the character frame 29 (step S12).

[0027] When the first block feature extraction is completed, the character validity determination unit 31 uses the previously prepared character validity information 3.
4 (step S13). If the determination in step S13 is OK, that is, if it matches any of the character validity information 34, a character frame determination process is performed to determine the character frame 29 (step S14), and the process related to cutting out the character frame is ended. . If the result of step S13 is NG,
That is, if it does not match any of the character validity information 34, the process moves to step S21 in FIG. 11.

In step S21, the character validity determining unit 31 determines whether the black dot block within the character frame 29 is part of a separated character. This determination determines whether the intersection feature of the black spot block in the horizontal and vertical directions is "1" and the width W of the black spot block is less than or equal to the width threshold TH. If the result of the determination is YES, the character validity determination unit 31 extracts the features of the first block and second block (the second black dot block from the left) in the character frame 29, similarly to step S12 (step S12). S22).

When the character validity determination section 31 completes the first and second block feature extraction, it performs a comparison determination with the previously prepared character validity information 34 (step S23). The determination in step S23 is OK, that is, the character validity information 34
If it matches any of the above, a character frame determination process is performed to determine the character frame 29 (step S24), and the process related to cutting out the character frame is ended.

If the result of step S21 is NO, the character validity determining unit 31 starts the threshold value change control unit 33, and
The binarization threshold is changed so that the operation result of the digitization circuit 4 is an increase in white dot blocks and a decrease in black dot blocks, and step S
Execute steps 11 and later again. The result of step S23 is NG
In the case of , that is, when there is no match with any of the character validity information 34, the process moves to step S31 in FIG. 12. In step S31, the character validity determination unit 31 determines the first and second blocks and the third block (the third black dot block from the left) in the character frame 29, as in steps S12 and S22.
Perform feature extraction.

When the character validity determination section 31 completes the first to third block feature extraction, it performs a comparison determination with the previously prepared character validity information 34 (step S32). If the determination in step S32 is OK, that is, if it matches any of the character validity information 34, a character frame confirmation process is performed to confirm the character frame 29 (step S33), and the character frame 29 is confirmed.
9 ends. Also, step S3
If the result of step 2 is NG, it is assumed that the form cannot be recognized with E13B characters, and a reject process such as ejecting and returning the form is performed (step S34), and the processing related to E13B character recognition is ended.

As described above, the character frame 29 is determined by referring to the horizontal and vertical intersection features, that is, the character validity information 34, and sequentially determining whether or not the black dot block is valid as an E13B character. In other words, in order to perform the cutting process,
It is also possible to recognize with high accuracy even forms in which black dot blocks are combined or lack spacing due to differences in print density. The present invention is not limited to the above embodiments. In the embodiment, the case where E13B characters are recognized is explained as an example, but the present invention can also be applied to the case where recognition of kana characters and the like having separated characters is performed.

[0033]

[Effects of the Invention] As explained above, according to the present invention, it is determined whether or not it is valid as a character, and if necessary, the threshold value for generating a binary pattern is changed, and the binary pattern is generated again. Therefore, it is possible to reduce the occurrence of recognition as one block due to combination of patterns, insufficient spacing, etc. Therefore, it is possible to similarly reduce the number of situations in which separated characters are treated as numeric characters.

[Brief explanation of drawings]

FIG. 1 is a conceptual diagram of a character recognition device according to the present invention.

FIG. 2 is a conceptual diagram of a general optical character reading device.

FIG. 3 is a conceptual diagram of a form.

FIG. 4 is a conceptual diagram of a binary pattern memory.

FIG. 5 is an explanatory diagram of correction by the cutout control unit.

FIG. 6 is an explanatory diagram of a width threshold.

FIG. 7 is an explanatory diagram of character validity information.

FIG. 8 is a block diagram of a character validity determination unit according to the present invention.

FIG. 9 is an explanatory diagram of the operation of a character validity determination section.

FIG. 10 is a first flowchart according to the present invention.

FIG. 11 is a second flowchart according to the present invention.

FIG. 12 is a third flowchart according to the present invention.

[Explanation of symbols] 8 Cutting control section 9 Format table 31 Character validity judgment section 32 Cutting position control section 33 Threshold value change control unit 34 Character validity information

Claims (1)

[Claims] Claim 1: A device that recognizes a character composed of a plurality of black dot blocks written on a medium, which inspects two-dimensional features of the black dot blocks and determines whether the black dot blocks constitute the character. a character validity determination unit that determines whether or not the black dot block constitutes the character, and a character frame cutout for cutting out an area in which the character is written from the medium when it is determined that the black dot block constitutes the character; a cutout position control unit that specifies the position; and a threshold change control unit that changes a threshold value when reading the black dot block and converting it into image data when it is determined that the black dot block does not constitute the character. A character recognition device characterized by: