JP3045086B2 - Optical character reading method and apparatus - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an optical character reading method and apparatus, and more particularly, to an optical character reading method in which a character written on a form is read while online determining whether the character is a handwritten character or a printed character. Method and apparatus.

[0002]

2. Description of the Related Art A conventional optical character reading method and apparatus,
In particular, as a conventional technique related to reading a form in which a form described in handwritten characters and a form described in print are mixed,
For example, (1) JP-A-4-273391, (2) JP-A-5-189604 and the like can be mentioned.

Japanese Unexamined Patent Publication No. Hei 4-273391 discloses a character reading apparatus for reading a character described in a handwritten type mixed entry field in which either a handwritten character or a type character is described. An image storage unit (1) for storing an area, a recognition target area extraction unit (2), and a handwritten type pre-judgment unit (3) for determining whether the target area is a handwritten character or a printed character based on the character candidate area group and its size ), A handwritten character recognition unit (4) for recognizing characters in a character candidate region determined to be handwritten characters and characters in a region determined to be printed characters, and a character in a region determined to be printed characters. A type recognition unit (5) for recognizing a type and a post-handwritten type determination unit that determines a determination result of a good recognition rate as a determination result based on the recognition rates of the determination results of both determination units when determining that the character is a type character. 6) Character reader, characterized in that. "
Has been proposed. Then, the characters in the area determined to be print characters are recognized as a whole without lowering the print character recognition rate by using the determination result with the higher recognition rate of the determination results of both determination units. The rate can be improved.

[0004] Japanese Patent Application Laid-Open No. Hei 5-189604 discloses that
`` In an optical character recognition device that recognizes character data obtained by reading a form on which a mixture of handwritten characters and printed characters are read by a scanner using either a handwritten dictionary or a printed dictionary, Based on the extracting means for extracting each character data in field units and the height of each character data extracted by the extracting means, it is determined whether each of these character data is a handwritten character or a printed type. Judgment means, when the character data is determined to be handwritten by the judgment means, perform character recognition of these character data using the handwritten dictionary,
An optical character reading device comprising: a recognition unit for performing character recognition of the character data using the type dictionary when it is determined that the character data is a type. "Has been proposed. Then, it is quickly determined whether the read character is handwritten or type, and only one of the handwritten character recognition and the type recognition is performed, and it is not necessary to perform both. It can be shortened.

[0005]

The above-mentioned conventional optical character reading method and apparatus, for example, Japanese Patent Laid-Open No. Hei 4-2733.
In the case of the character reading device disclosed in Japanese Patent Publication No. 91, characters in a character candidate area determined to be handwritten characters are only recognized by the handwritten character recognition unit, but characters in an area determined to be printed characters are recognized by handwritten character recognition. Since the judgment result of the judgment unit of the judgment unit and the type character recognition unit, which has the higher recognition rate, is used, handwritten character recognition must be executed for the type characters, and there is no waste in processing. However, there is a problem that the overall processing speed is reduced.

[0006] Japanese Patent Application Laid-Open No. Hei 5-189604 discloses that
Based on the height of each character data extracted by the extraction means, in order to improve the problem of processing speed reduction due to the combined processing of handwritten character recognition and type recognition having the conventional technology,
By providing a determination means for determining whether each of these character data is a handwritten character or a printed type, only one of the handwritten character recognition and the type recognition is performed. Although speed is improved, handwritten characters by an unspecified number of writers may be written as small as print, so if small handwritten characters are written in order, they are recognized as type. Is executed, and since the dictionary for recognition is different between handwritten characters and printed characters, most of the entered characters are determined to be unreadable or misread.

SUMMARY OF THE INVENTION It is an object of the present invention to provide an optical character reading method and apparatus capable of minimizing a reduction in processing speed and relieving mutual misunderstanding between handwritten characters and printed characters.

[0008]

According to the optical character reading method of the present invention, an image input means reads a paper image of a form in which a part described in handwritten characters and a part described in printed characters are mixed, and the paper input method is used. Only the partial image including the character entry field to be read from the image is cut out by the area cutout means, and the target character pattern included in the partial image is written based on the vertical / horizontal size and the character spacing. Which font is determined by the font determining means, the collating means performs character collation using either the handwritten character dictionary or the type dictionary based on the result of the decision, and the collation result output by the collating means The collation result determination means calculates the ratio of the unread character from the above, and if the ratio of the unread character is less than a predetermined threshold, the collation result is used as the final collation result. And force, the When the proportion of the non-read characters the predetermined threshold value or more requested typeface redetermination and collated rerun, and to perform the re-character matching by changing the font.

According to the optical character reading method of the present invention, the ratio of unread characters calculated by the collation result judging means based on the collation result obtained by changing the font and performing character collation again is the result of the collation in the first font. If it is the same value, the display means displays the result of the collation in the first font and the result of the collation performed by changing the font, and outputs the result of one of the collation according to a selection instruction from the outside. You may.

In the optical character reading method according to the present invention, the threshold value for determining the ratio of unread characters may be any value between 0 and 1 set externally at any time.

According to the optical character reading method of the present invention, the ratio of unread characters in the result of reading characters using the matching means is calculated, and the ratio of the unread characters is less than a predetermined threshold value. If so, a reading result may be output, and if the ratio of the unread characters is equal to or greater than a threshold, an alarm signal may be output.

The optical character reading apparatus according to the present invention is characterized in that image input means for inputting a paper image of a form in which a portion described in handwritten characters and a portion described in printed characters are mixed as a binary image, Area extracting means for extracting only a partial image including a character entry field to be read from the paper image input by the means; and obtaining a space between adjacent characters and a vertical / horizontal size of each character from the partial image. Character extracting means for extracting a character pattern character by character, and character type determination for determining whether the written character is a handwritten character or a printed character based on the vertical and horizontal sizes and character spacing of the character pattern extracted by the character extracting means Means, a handwritten character dictionary in which the character features of the handwritten characters are recorded in advance, and a character dictionary in which the character features of the type characters are recorded in advance, and the font determining means A collation unit that performs collation using either the handwritten character dictionary or the type dictionary based on the font determination result, and calculates a ratio of unread characters in the collation result output by the collation unit. If the proportion of characters is less than a predetermined threshold, the collation result is output.If the proportion of unread characters is greater than or equal to the threshold, a request for font re-determination and collation re-execution is requested. A comparison result determining unit that compares the ratio of the unread characters in both the comparison results and outputs a comparison result with a smaller number of unread characters, and a control that controls the operation of each unit. Means.

In the optical character reading apparatus according to the present invention, the collation result determination means calculates a ratio of unread characters in the collation result output from the collation means, and the ratio of the unread characters is less than a predetermined threshold. If so, it may have a function of outputting the collation result, and outputting an alarm signal if the ratio of the unread characters is equal to or greater than a threshold value.

[Operation] In the optical character reader (1) of the present invention, the image input means (2) reads a paper surface image of a form in which handwritten characters and printed characters are mixed and entered as a binary image. The area extracting means (3) cuts out only the partial image of the character entry field from the paper surface image of the form read by the image input means (2). Next, the character extracting means (4) obtains the interval between adjacent characters and the vertical and horizontal sizes of each character from the extracted partial image, and extracts character patterns one by one. The font determining unit (5) determines whether the entered character is a handwritten character or a printed character based on the vertical and horizontal sizes of the extracted character pattern and the character spacing. The collation unit (6) is configured to select one of a handwritten character dictionary (7) in which the character shape characteristics of the handwritten characters are recorded in advance and a character dictionary (8) in which the character shape characteristics of the type characters are recorded in advance, based on the result of the character style determination. Collation is performed using the dictionary. The collation result determination means (9) calculates the ratio of unread characters in the collation result output from the collation means (6), and outputs the collation result as it is when the ratio is less than a predetermined threshold. However, if the ratio of unread characters is equal to or greater than the threshold value, request font re-determination and collation re-execution, change the font and perform character collation again, and obtain the results of both handwritten character recognition and printed character recognition. If so, the ratio of unread characters in both the comparison results is compared, and the comparison result with the smaller number of unread characters is output.

[0015]

Next, a first embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of the present invention.

An optical character reader 1 of the present invention optically takes in a paper image of a form to be read, converts the paper image into a binary image, and converts the binary image into a character area. , A character extracting unit 4 for extracting a character pattern one character at a time from the character region extracted here, and a character area to be read based on the vertical and horizontal sizes of the character pattern and the character interval. A font determining unit 5 that determines whether the area is a print area or a handwritten character area;
Based on the result of the determination by the font determining unit 5, the handwritten character dictionary 7
Alternatively, a collating unit 6 for performing collation using one of the type dictionaries 8, and a collation result determining unit 9 for determining a collation result
And a control unit 10 for controlling the operation of each unit described above.

Hereinafter, detailed functions and operations of each unit of the optical character reading apparatus according to the present embodiment will be described.

When the optical character reading apparatus 1 starts processing, first, the control section 10 activates the image input section 2. The image input unit 2 receives the reflection of the light illuminated on the form paper surface with a sensor, thereby taking in the paper surface image of the form to be read as a multi-valued image in which each pixel is represented by a density value. Next, this density value is compared with a preset threshold value, and a pixel darker than the threshold value is set to 1 and a pixel lighter than the threshold value is set to 0, thereby performing conversion into a binary image. After ending the processing of the image input unit 2, the control unit 10 activates the area extracting unit 3 next.

The area extracting section 3 previously stores positional information (distance from the top of the form, distance from the right or left end of the form to the end point of the area) regarding the character entry field on the form, and the vertical and horizontal sizes of the entry field. Referring to a read information file not shown in FIG. 1, the position of a character entry field is specified on a form image, a character area to be read is cut out, and the partial image is not shown in an area not shown. Performs copying to the buffer. In this process, in order to reduce the time for copying to the buffer, the coordinates of the reading area on the actual form image may be calculated, and the coordinate values may be stored.

The area cutout unit 3 projects black pixels in the character area in the vertical direction, obtains the cumulative value of the black pixels in the vertical direction, and examines the distribution of the cumulative values to obtain the character pitch shown in FIG. And character pitches (p1 to p3) shown on the explanatory diagram for explaining the circumscribed frame are extracted. (Such a method is generally used in character extraction.) Next, as shown in FIG. 5, a circumscribed frame for each character is determined, and the vertical and horizontal sizes of each character are extracted. . The character pitch and the vertical / horizontal size determined here are stored in the control unit 10 for reference by the character style determination unit 5 described later. After ending the processing of the area extracting section 3, the control section 10 activates the character extracting section 4 next.

The character extracting section 4 performs a process of copying the image in the circumscribed frame set by the area extracting section 3 into a character buffer (not shown in FIG. 1), and uses this as a character pattern for one character. . The control unit 10 activates the font determining unit 5 after the processing in the area extracting unit 4 is completed.

The font determining unit 5 determines whether the character pattern is a handwritten character or a printed character based on the vertical and horizontal sizes and the pitch of the character pattern extracted by the area extracting unit 3. In this embodiment, if both the vertical and horizontal sizes of the character pattern are 4 mm or less, and the variance of the pitch is almost equal to the experimental value, the character pattern is determined to be a printed character. It shall be determined. Note that the experimental value indicates a variance value which is obtained in advance by separately conducting a preliminary experiment and checking the pitch of the type font to be read. The result of the determination is output to the control unit 10.

The control unit 10 saves the font determination result and activates the collation unit 6. At this time, the font of the determination result is notified to the collation unit 6. The collating unit 6 performs the collating process using either the handwritten character dictionary 7 or the type dictionary 8 based on the font determination by the font determining unit 5. In the case of print character collation, the read character pattern and the dictionary pattern are superimposed, and pattern matching for calculating the distance is performed.
In the type dictionary 8, the outline features of the type (only the outline is extracted from the character image) are recorded in advance.

For handwritten character collation, pattern matching may be used in the same way as for printed character collation. , Connection, etc.) may be used. Regardless of which method is used, it is assumed that features required for collation are recorded in the handwritten character dictionary in advance. After the processing in the collating unit 6 is completed, the control unit 10
The collation result determination unit 9 is started.

FIG. 2 is a flow chart for explaining the operation of the collation judging section according to the first embodiment.

The operation of the collation result judgment section 9 will be described with reference to FIG. The description uses the names and reference numerals in FIG.

The collation result judging unit 9 first receives the collation result from the collation unit 6 in step (hereinafter referred to as S) 1,
In S2, it is determined whether or not this collation result is the first collation result for the reading area.
And compares the ratio R1 occupied by unread characters in the collation result with a preset threshold value. If R1 is less than the threshold value, S
In step 4, the comparison result is output, and the process ends. Hereinafter, the ratio of unread characters in the result of the i-th comparison is described as Ri.

On the other hand, if the result of comparison in S3 is that the ratio R1 of unread characters is equal to or greater than the threshold, the font is changed in S5 and the control unit 10 is requested to re-execute collation. Upon receiving the re-execution request, the control unit 10 changes the stored font determination result and restarts the collation unit 6.

Upon receiving the restart, the collation unit 6 performs re-collation and outputs the collation result to the collation result determination unit 9. Upon receiving this collation result in S1, the collation result determination unit 9 determines whether or not the collation result is the first collation result for the reading area in S2, and two collation results have already been obtained for the reading area. If so, it is determined that the result is the second collation,
6, the comparison result of the two times, that is, the ratio of the unread characters of both the type recognition and the handwriting recognition is compared. If R1 <R2, the first comparison result is output as the read result in S7.
If R1 ≧ R2, the second collation result is output as a read result in S8, and the process ends.

Next, a second embodiment of the present invention will be described with reference to the drawings.

The block diagram showing the second embodiment has exactly the same shape as the block diagram showing the first embodiment. However, since the operation of the collation result judging unit is different, the signs of the optical character reader and the collation result judging unit are changed and displayed as the optical character reader (11) and the collation result judging unit (12). . The other components are the same as those in the first embodiment, and thus description thereof is omitted.

FIG. 3 is a flowchart for explaining the operation of the collation result judging unit according to the second embodiment.

The operation of the collation result determination section 12 will be described with reference to FIG. The description uses the names and reference numerals in FIG.

In the collation result judging section 12 of the second embodiment, when the collation result is received from the collation section 6 in S11,
In S12, it is determined whether or not this collation result is the first collation result for the reading area.
In step 13, the ratio R1 of unread characters in the collation result is compared with a preset threshold value. If R1 is less than the threshold value, the collation result is output in step S14, and the process ends.
On the other hand, if the result of the comparison in S13 is that the ratio R1 of unread characters is equal to or greater than the threshold value, the font is changed in S15 and the control unit 10 is requested to re-execute the comparison. Control unit 1 that has received a re-execution request
0 changes the saved font determination result to the collation unit 6
To restart.

Upon receiving the restart, the collation unit 6 performs re-collation and outputs the collation result to the collation result determination unit 12. Upon receiving this collation result in S11, the collation result determination unit 12
It is determined whether or not the collation result is the first collation result for the reading area. If two collation results are already obtained for the reading area, it is determined that the collation result is the second collation result. The comparison results of two times, that is, the ratio of unread characters in both type recognition and handwriting recognition are compared, and R1 = R
If it is 2, in step S17, the collation result of the read characters for both the character recognition and the handwriting recognition is output and displayed on a display device (hereinafter referred to as CRT) not shown in FIG. The user selects and instructs which of the read characters to use in handwriting recognition, outputs the collation result of the one selected by the operator in S19, and ends the processing.

If the determination in S16 is not R1 = R2, the ratio of unread characters in both type recognition and handwriting recognition is compared in S20. If R1 <R2, 1 is determined in S21.
The result of the second verification is output, the processing is terminated, and if R1 <
If it is not R2, the second matching result is output in S22,
The process ends.

Next, a third embodiment of the present invention will be described with reference to the drawings.

The block diagram showing the third embodiment has exactly the same shape as the block diagram showing the first embodiment. However, since the operation of the collation result judging unit is different, the signs of the optical character reader and the collation result judging unit are changed and are displayed as the optical character reader (13) and the collation result judging unit (14). . The other components are the same as those in the first embodiment, and thus description thereof is omitted.

FIG. 4 is a flow chart for explaining the operation of the collation result judging section according to the third embodiment.

The operation of the collation result determination section 14 will be described with reference to FIG. The description uses the names and reference numerals in FIG.

In the collation result determination unit 14 of the third embodiment, first, when the collation result is received from the collation unit 6 in S21,
In S22, the ratio R1 of unread characters in the collation result output from the collation unit 6 is compared with a preset threshold value, and R1 <
If it is the threshold, the collation result is output in S23, and the processing is terminated. If R1 <threshold, the font determination by the font determination unit 5 may have been erroneous. Output and end the process.

[0043]

As described above, according to the present invention, a paper image of a form in which a part written in handwritten characters and a part written in type are mixed is read by the image input means and read from the paper image. Only the partial image including the character entry box to be cut out is extracted by the region extracting means, and the target character pattern included in the partial image is a handwritten character or a typeface based on the vertical and horizontal sizes and the character spacing. Based on the determination result, the matching unit performs character matching using either the handwritten character dictionary or the type dictionary, and based on the matching result output from the matching unit, the matching result determination unit determines the ratio of unread characters. Is calculated, and if the ratio of the unread characters is less than a predetermined threshold, the result of the collation is output as the final collation result. If this is the case, a request for font re-determination and collation re-execution is requested, and the font is changed and the character collation is performed again. Thus, even if the character size or the interval (pitch) between adjacent characters is ambiguous, the font There is an effect that unreading / misreading due to an error can be avoided. Further, even when it is difficult to distinguish between a printed character and a handwritten character due to a small character size, the character reading process is executed at a higher speed without always performing both the type recognition and the handwritten character recognition as in the related art. There is an effect that it becomes possible. further,
If the ratio of unread characters is greater than or equal to the threshold value, there is a possibility that the determination result of the font may be erroneous. This also has the effect of notifying the operator of the apparatus that the possibility of misreading is high and alerting the user. In addition, there is an effect that reduction in processing speed can be reduced as much as possible and mutual misunderstanding between handwritten characters and printed characters can be relieved.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention.

FIG. 2 is a flowchart illustrating an operation of a matching determination unit according to the first embodiment.

FIG. 3 is a flowchart illustrating an operation of a collation result determination unit according to the second embodiment.

FIG. 4 is a flowchart illustrating an operation of a collation result determination unit according to the third embodiment.

FIG. 5 is an explanatory diagram illustrating a character pitch and a circumscribed frame.

[Explanation of symbols]

 1,11,13 Optical character reader 2 Image input unit 3 Area cutout unit 4 Character cutout unit 5 Font determination unit 6 Collation unit 7 Handwritten character dictionary 8 Type dictionary 9, 12, 14 Collation result determination unit 10 Control unit

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G06K 9/20 G06K 9/62 G06K 9/68

Claims (6)

(57) [Claims] An image input means reads a paper image of a form in which a part described in handwritten characters and a part described in print are mixed, and only a partial image including a character entry field to be read from the paper image Is cut out by the area cutout means, and the character pattern determination means determines whether the character pattern to be included in the partial image is a handwritten character or a typeface based on the vertical / horizontal size and the character spacing, Based on the determination result, the matching unit performs character matching using one of the handwritten character dictionary and the type dictionary, and the matching result determination unit calculates the ratio of unread characters from the matching result output by the matching unit. If the ratio of the unread characters is less than a predetermined threshold, the result of the comparison is output as a final comparison result, and the ratio of the unread characters is the predetermined threshold. An optical character reading method comprising: requesting re-determination of the font and re-executing the collation, changing the font, and performing character collation again. 2. The display means if the ratio of unread characters calculated by the collation result determining means from the collation result obtained by changing the font and performing character collation again is the same value as the collation result in the first font. Displaying the result of the collation in the first font and the result of the collation performed by changing the font, and outputting one of the collation results in accordance with an external selection instruction. The optical character reading method described in the above. 3. The optical character reading method according to claim 1, wherein the threshold value for determining the ratio of unread characters is an arbitrary value between 0 and 1 set externally at an arbitrary time. And calculating a ratio of unread characters in a result of the character reading using the collating means, and outputting the read result if the ratio of the unread characters is less than a predetermined threshold. 2. The optical character reading method according to claim 1, wherein an alarm signal is output if the ratio of the unread characters is equal to or greater than a threshold value. 5. An image input means for inputting, as a binary image, a paper surface image of a form in which a part described in handwritten characters and a part described in printed characters are mixed, and An area extracting means for extracting only a partial image including a character entry field to be read, and a character for extracting a character pattern one character at a time from the partial image by calculating an interval between adjacent characters and a vertical and horizontal size of each character A cut-out unit, a font determining unit that determines whether the written character is a handwritten character or a printed character based on the vertical and horizontal sizes and character intervals of the character pattern cut out by the character extracting unit, and a character shape characteristic of the handwritten character Is connected to a handwritten character dictionary in which characters are pre-recorded and a character dictionary in which character characteristics of characters are pre-recorded. A collating unit that performs collation using one of a written character dictionary and a type dictionary, and a ratio of unread characters in a collation result output by the collating unit is calculated. If the ratio is less than the threshold, the collation result is output.If the ratio of the unread characters is equal to or greater than the threshold, a request for font re-determination and collation re-execution is requested, and the collation results of both handwritten character collation and type collation are obtained. Collation result determination means for collating the ratio of the unread characters in both collation results and outputting the collation result with the smaller number of unread characters,
An optical character reading device comprising: a control unit for controlling operation of each of the units. 6. A collation result determination unit calculates a ratio of unread characters in the collation result output from the collation unit, and outputs the collation result if the ratio of the unread characters is less than a predetermined threshold. 6. The optical character reading device according to claim 5, further comprising a function of outputting an alarm signal when the ratio of the unread characters is equal to or greater than a threshold value.