JP3055484B2 - Character recognition apparatus and method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a character recognition device and method, and more particularly to a character recognition device and method for recognizing a character having a predetermined height.

[0002]

2. Description of the Related Art Heretofore, a character recognition device has been used for recognizing characters marked on a sheet by printing or the like. A conventional technique of such a character recognition device will be described below with reference to FIG. FIG. 7 is a schematic diagram showing a logical structure of a conventional character recognition device.

An OCR exemplified here as a character recognition device
For example, the system 1 connects an image scanner to a computer system, and connects the computer system to an OC.
It is realized as a structure (not shown) in which an application program for R is installed.

In the OCR system 1 having such a structure, the computer system executes an appropriate processing operation in accordance with a program, whereby an image extracting unit 2, a multi-valued image memory 3, a character extracting unit 4, a character normalizing unit 5, It logically includes a standard pattern memory 6, a character recognition unit 7, and the like.

The image extraction unit 2 corresponds to a part of the computer system which controls the operation of the image scanner. The image extraction unit 2 optically reads characters printed on a form, photoelectrically converts the characters, and digitizes the multi-valued levels. Is generated and stored in the multi-valued image memory 3. The multi-valued image memory 3 corresponds to, for example, a predetermined storage area of a RAM, and temporarily stores multi-valued image data.

The character extracting section 4 binarizes the multi-valued image data, extracts the binary data in units of characters, and detects the height and width of each character. The character normalizing section 5 corrects the position of the image data vertically and horizontally with respect to a predetermined image area by referring to the character height and character width after removing minute unevenness of the image data. In the standard pattern memory 6, a standard pattern of a character to be recognized is set in advance, and the character recognizing unit 7 sequentially stores a large number of standard patterns in a recognition pattern which is image data whose position is corrected as described above. Matching is performed, and the characters of the standard pattern with the highest matching accuracy are output as recognition results.

[0007] The character recognition device 1 having the above structure can recognize characters printed on a form. In this case, the characters printed on the form are optically read by the image extracting unit 2 and photoelectrically converted, and the digitized multi-level image data is stored in the multi-level image memory 3. In the character cutout unit 4, of the multi-valued image data, a black pixel portion which is a character itself is set to 1 and a black pixel portion which is not present in the background is set to 0, and the binarized image data is converted into a character unit. Then, the character height and character width of the image data are detected.

The character normalizing unit 5 removes minute irregularities from the image data cut out in units of characters as described above, and refers to the character height and character width detected by the character cutting unit 4 to determine the black level. The recognition pattern is generated by matching the center position of the pixel with the center position of the image area of the standard pattern. Next, the character recognition unit 7 sequentially matches the recognition pattern output from the character normalization unit 5 with a large number of standard patterns set in the standard pattern memory 6, and determines the character of the standard pattern with the highest matching accuracy. Output as recognition result.

In the above-described OCR system 1, when the recognition pattern detected from the form is matched with the standard pattern as described above, the center position of the image data is made to coincide with the center position of the image area of the standard pattern prior to this. ing. For this reason, it is possible to prevent a decrease in matching accuracy due to a positional shift between the standard pattern and the recognition pattern during pattern matching, and it is possible to satisfactorily recognize characters printed on a form.

[0010]

In the conventional character recognition device as described above, in order to improve the accuracy of character recognition,
The position of the recognition pattern to be matched with the standard pattern is corrected.

However, as shown in FIG. 8A, for example, characters printed on a form may have an upper portion or a lower portion missing due to missing dots of a printer or misregistration of a rubber stamp. When the center position of the image data with the upper and lower portions missing is matched with the center of the area of the standard pattern, the shape becomes different from the standard pattern, as shown in FIGS. 8B and 8C. Will decrease.

SUMMARY OF THE INVENTION The present invention has been made in consideration of the above-described problems, and has as its object to provide a character recognition apparatus and method capable of satisfactorily recognizing a character whose upper or lower part is missing.

[0013]

First, a character recognition device of the present invention sets a standard pattern of characters to be recognized in advance, and optically reads characters of a predetermined height marked on a form. The read image data is cut out in character units, the center position is aligned with the center position of the image area of the standard pattern, and the image data with the aligned center position is sequentially matched as a recognition pattern with a number of standard patterns, achieving the highest matching accuracy. In a character recognition device that outputs characters of a standard pattern as a recognition result, a missing determining means for determining whether there is a missing in the vertical direction of image data cut out in character units, and detecting a direction of the missing detected presence Direction detecting means, height detecting means for detecting the height of the detected missing part, and corresponding to the direction and height of the detected missing part. Correcting the position for the image area of the image data in the vertical direction Te, the direction of the missing Do found
Forming the have image data for selecting the position correction means Ru is displaced vertically both, one compares the two matching results
Fruit selection means .

According to the character recognition method of the present invention, a standard pattern of a character to be recognized is set in advance, characters of a predetermined height marked on a form are optically read, and the read image data is read. Cut out in character units, align the center position with the center position of the standard pattern image area, match the image data with the aligned center position as a recognition pattern, sequentially with a number of standard patterns, and select the characters of the standard pattern with the highest matching accuracy. In a character recognition method that is output as a recognition result, it is determined whether or not there is a vertical drop in image data cut out in character units, and the presence of the missing is detected, and the presence is detected. Detects the height of the missing part and moves the position of the image data with respect to the image area in the vertical direction according to the detected direction and height of the missing part. Corrected, stomach direction of the missing is not found
Image data is displaced both up and down , two matching
The results were compared and one was selected .

Therefore, when recognizing a character of a predetermined height marked on a form, image data optically read from the form is cut out in character units. The presence or absence of a vertical drop in the image data cut out in character units in this way is determined, and if the presence of a drop is not detected, the center position of the cut out image data is matched with the center position of the standard pattern. Then, the image data whose center positions match are sequentially matched with a large number of standard patterns as recognition patterns.

However, when the presence of a missing portion is detected, the direction and height of the missing portion are detected, and the direction and height of the missing portion are detected and the image data of the image area corresponding to the missing portion is detected. Correct the position vertically. By correcting the alignment operation of the center position in this way, the image data with the upper and lower parts missing is arranged at the original position, so that the recognition pattern with the upper and lower parts missing matches the standard pattern at the appropriate position. Is done. Furthermore, characters
If it is not clear whether the direction of the
The center position of the page data is displaced upward and downward
Is matched with the pattern, and the accuracy is determined from the two results.
The higher one is selected.

Note that the characters referred to in the present invention may be characters having a constant height, and may include specific symbols in addition to numbers and alphabets. The form means a medium on which characters are marked, for example, a sheet on which characters are printed is allowed. The various means of the character recognition device may be those having the function logically, and are realized, for example, as dedicated hardware, a general-purpose computer operating according to an appropriate program, a combination thereof, or the like.

In another aspect of the character recognition apparatus as described above, a character height storage unit in which a prescribed value of the character height is set in advance is provided, and the lack determination unit determines the height of the image data in character units. The presence / absence of a missing part is determined by comparing with a specified value. Therefore, a character with a missing upper or lower part has a shortage, and the presence or absence of a missing part is determined by comparing the character height with a specified value.

In another aspect of the character recognition apparatus as described above, there is provided a data correction means for correcting a prescribed value of the height based on the height of the image data determined to have no loss. Therefore, the height of the image data may become higher or lower as a whole due to a malfunction of the printer or the scanner, but even in such a case, the height of the image data is determined based on the height of the image data determined to have no omission. Since the specified value is updated, the lack of some image data is relatively detected.

[0020]

[0021]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to FIGS. The same parts as those in the conventional example described above with reference to the present embodiment are denoted by the same names and reference numerals, and detailed description is omitted. FIG. 1 is a schematic diagram showing a logical structure of the character recognition device of the present embodiment,
FIG. 2 is a block diagram showing a physical structure, FIGS. 3 and 4 are schematic diagrams showing recognition patterns, and FIGS. 5 and 6 are flowcharts showing a character recognition method according to the present embodiment.

As shown in FIG. 2, an OCR system 11 exemplified as a character recognition device in this embodiment includes a computer system 12 as a data processing device and an image scanner 13 as an image reading device. .

The computer system 12 includes a CPU 101 as a main body of the computer.
03, RAM 104, HDD 105, FDD 107 in which FD 106 is loaded, CD drive 109 in which CD-ROM 108 is loaded, keyboard 110, mouse 11
1, a display 112, a communication I / F 113, and the like are connected. A connector 114 is connected to the communication I / F 113, and the image scanner 13 is connected to the connector 114.

In the OCR system 11 of the present embodiment,
The ROM 103, the RAM 104, the HDD 10
5. The FD 106, the CD-ROM 108, and the like correspond to an information storage medium, in which programs and data necessary for various operations are stored as software. For example, a control program for causing the CPU 101 to execute various processing operations includes the FD 106 and the CD-ROM 1
08 is written in advance. Such software is pre-installed in the HDD 105, and the RA is activated when the computer system 12 is started.
Read the CPU101 is copied to M 104.

As described above, the CPU 101 reads an appropriate program and executes various processing operations, whereby various functions are realized as various means. As such various means, the OCR system 1 of the present embodiment
Reference numeral 1 denotes an image extraction unit 2, a multi-valued image memory 3, a character cutout unit 4, and a field coordinate memory 1 as shown in FIG.
4. Standard character height memory 15 which is a character height storage means, missing detector 16, first and second reference position correctors 17, 18, character normalizer 5, standard pattern memory 6, character recognizer 7,
It has a result selection unit 19 and the like logically.

The image extracting section 2 is adapted to execute the CPU 101 according to a control program set in the ROM 103.
Controls the operation of the image scanner 13 to optically capture the characters to be recognized printed on the form, perform photoelectric conversion, and output digitized multilevel image data. The multi-valued image memory 3
Is equivalent to a storage area of a rewritable data storage medium such as the RAM 104 or the HDD 105, and temporarily stores multi-valued image data read from a form by the image extraction unit 2.

The character extracting section 4 is provided in the ROM 10
3 according to the control program set in the CPU 1.
01 performs a predetermined processing operation, thereby binarizing the multi-valued image data stored in the multi-valued image memory 3 and cutting out the data in units of characters, and the height and width of each of the image data in units of characters. Is detected, and the character coordinates and the character height for one field (for example, one line) are calculated. The field coordinate memory 14 is also provided in the RAM 10
4 and a storage area of a rewritable data storage medium such as the HDD 105, and temporarily stores various data such as the height of image data detected by the character cutout unit 4.

The standard character height memory 15 also includes the RAM
It corresponds to a storage area of a rewritable data storage medium such as 104 or the HDD 105. Here, a prescribed value of the height of the standard pattern is stored in advance as a character height storage means so as to be freely updated. The missing detector 16 is provided in the ROM 103
CPU 10 corresponding to the control program set in
1 performs a predetermined processing operation, a missing determining means for determining whether there is a missing in the vertical direction of the image data cut out in character units, a direction detecting means for detecting the direction of the missing detected presence, It functions as height detecting means for detecting the height of the missing part whose presence has been detected.

That is, the missing portion detecting section 16 compares the character height stored in the field coordinate memory 14 with a predetermined character height set in the standard character height memory 15 in advance, and determines whether a character is missing. Is detected, and if there is no missing part, various data are directly output to the character normalization unit 5 as in the conventional case. However, if there is a missing portion, the direction, height and amount of the missing portion are detected based on the character coordinates stored in the field coordinate memory 14, and the first and second reference position correcting sections 17 are detected in accordance with the missing direction. , 18 are output.

More specifically, the missing detecting section 16 determines the presence or absence of a missing in a field unit and detects the direction and height of the missing only when a missing character is partially detected in the field. I do. In this case, when an upper drop is detected, various data is output to the first reference position corrector 17, and when a lower drop is detected, various data is output to the second reference position corrector 18. . However, when all the characters in the field are missing, only the height is detected without detecting the direction of the missing, and various data are output to both the first and second reference position correction units 17 and 18. I do.

The first and second reference position correctors 17, 1
Reference numeral 8 denotes a position of the image data with respect to the image area corresponding to the direction and the height of the detected missing portion when the CPU 101 executes a predetermined processing operation in accordance with the control program set in the ROM 103. It functions as a position correcting means for correcting in the vertical direction. That is, the first and second reference position correction units 17 and 18 displace the reference position in the up and down direction with respect to the input image data in accordance with the height of the missing part. FIG.
Since the reference position is the position of the upper edge portion of the field as shown in (1), if the coordinates are changed so that the position goes up and down, the center position of the image data also moves up and down.

The character normalizing section 5 is also provided with the ROM
The CP corresponding to the control program set in 103
U101 executes a predetermined processing operation to remove minute irregularities of the image data, and then refers to the character height and character width stored in the field coordinate memory 14 to determine whether the center position of the image data is a standard pattern. Is moved so as to match the center position of the image area of, and a recognition pattern is generated.

However, as described above, since the reference position of the image data whose upper and lower portions are missing is displaced in the vertical direction corresponding to the direction and height of the missing portion, the image data of the standard pattern is lost as described above. The center position aligned with the center position of the area is also displaced in accordance with the lack, and as a result, a recognition pattern with the upper and lower parts missing is also generated at the original position.

The character recognizing section 7 is also provided with the ROM 1
03 according to the control program set in the CPU 03
The 101 performs a predetermined processing operation to sequentially match the various standard pattern memories 6 with the recognition patterns normalized by the character normalization unit 5, and recognizes the characters of the standard pattern having the highest matching accuracy as a recognition result. Output.

However, as described above, the recognition pattern for which the missing direction is not determined is matched with the standard pattern while being displaced both upward and downward. In this case, the character recognizing unit 7 uses the result selection means. As a result, two matching results are compared to select one having a higher matching accuracy.

The various means as described above are realized by using hardware such as the image scanner 13 and the RAM 104 as necessary.
04 corresponding to the software written in
This is realized by the operation of U101.

Such software, for example, controls the operation of the image scanner 13 to read multi-level image data from a form and store it in a predetermined area of the RAM 104 or the like. Is binarized and cut out in character units, the height and width of each of the image data are extracted to calculate the character coordinates and character height for one field, and such various data are stored in the RAM 104 or the like. Storing in a predetermined area, comparing the character height detected as described above with a predetermined value set in advance, and detecting the presence or absence of a missing in a field unit; If a part of the image data is detected, the direction and height of the missing image data are detected, and the reference position is set to one of the vertical directions. If a drop is detected in the entire field, only the height of the drop in the field is detected and the reference position of all image data is displaced both in the vertical direction. Removing the irregularities, referring to the character height and character width, aligning the center position of the image data with the center position of the image area of the standard pattern, and generating a recognition pattern. It is described as a control program for causing the CPU 101 and the like to execute processing operations such as sequential matching, outputting a character of a standard pattern having the highest matching accuracy as a recognition result, and the like.

The software of such a control program includes various types of data such as a standard pattern of characters as a recognition result and a prescribed value of the height in the form of a separately updatable data file. It is possible to add.

The character recognition method by the OCR system 11 according to the present embodiment in the above configuration will be described below. First, a user sets a form to read a character on the image scanner 13 and sets the computer system 12
Is manually operated to activate a character recognition program.

Then, the CP of the computer system 12
U101 controls the operation of the image scanner 13 so that the image extracting unit 2 optically reads the characters printed on the form, photoelectrically converts the characters, and digitizes the multi-level image as shown in FIG. Data is generated and stored in the multi-valued image memory 3 (step S1).

Next, the character cutout unit 4 performs binarization on the multi-valued image data by setting the black pixel portion of the character to 1 and the black portion not existing in the background to 0 (step S2). ), The image data is cut out in character units (step S3). Further, the character height and character width are detected (step S4), and these various data are stored in the field coordinate memory 14 (step S5). At this time, as shown in FIG. 3, the coordinates of the uppermost end and the lowermost end are detected for all the characters in the field with reference to the image area of the form, and the character height is calculated as a difference between them.

Next, the missing portion detecting section 16 reads out the specified value of the character height of the standard pattern from the standard character height memory 15.
(Step S6). In the standard character height memory 15, the character height of the standard pattern is set in advance in correspondence with the characters to be read from the form and the characters of the rubber stamp. Is set for each character group.

The missing detector 16 detects the presence or absence of missing image data, and as shown in FIG. 6, the processing is divided into three types. First, the presence / absence of a vertical drop is determined for all character unit image data of one field stored in the field coordinate memory 14 (step S7). If they are uniformly detected (step S8), only the height of the missing portion is calculated without detecting the direction of the missing portion (step S9). In this case, various data such as the height of the image data is output to both the first and second reference position correction units 17 and 18, whereby the reference position of the image data is both up and down corresponding to the height of the lack. The correction is made (step S10). Creating two types of recognition patterns from the image data when the direction of the missing is not determined in this way is referred to as a first process here.

If a missing part is detected from some character-based image data of one field (step S
11), the height of the missing portion and the direction of the missing portion are calculated for the image data in which the missing portion is detected (step S12,
S13), the reference position is corrected to one of upper and lower directions corresponding to this direction (steps S14 to S16). Creating the recognition pattern corresponding to the direction of the missing in this manner is referred to as a second process herein.

Further, if no omission is detected from the whole one field (step S11), a recognition pattern is generated in the same manner as in the related art without correcting the reference position. This is referred to herein as a third process.

Here, the algorithm of processing such as detection of missing data will be described. First, the detection of the lack (algorithm 1) is performed, as shown in FIG. 3, by using the coordinates of the upper end and the lower end of the character-based image data cut out by the character cutout unit 4.
(H1, L1) are detected, and the character height (h)
1) is compared with the prescribed value (kh) of the character height of the standard pattern. In this case, the absolute value of "h1-kh" is defined as the character height difference. If the absolute value is within the allowable range, it is determined that there is no loss, and if it is outside the allowable range, it is determined that there is a loss. At this time, the CPU 101 operating according to the appropriate program functions as a data correction unit, and the average value of the character height of the image data determined to be no missing and the specified value of the character height of the standard pattern is calculated. The value is updated as a high specified value (step S15).

However, if the updated value is out of the allowable range with respect to the initially specified value, the updating of the specified value is stopped. As a result, variations in character size due to printing that occur in each optical system, each form, and each field can be absorbed. It is preferable that the allowable range described above is calculated in advance by an experiment.

Next, the direction of the missing part detected from the image data of a part of the field is determined (algorithm 2).
The calculation of the height (algorithm 3) will be described below with reference to FIG. First, image data determined to have no loss is considered as a reference character, and the coordinate values of the upper end and the lower end are used as reference coordinates.

The difference between the coordinates H3 and H1 at the upper end between the reference character and the image data to be processed is calculated as an absolute value. Similarly, the absolute value of the difference between the lower ends L3 and L1 is calculated. Is determined to be a missing upper end (pattern 1) when the upper end is missing, and conversely, it is determined to be a missing lower end (pattern 2) if the larger one is the lower end. If the comparison difference is within the allowable range,
It is determined that there is no missing part (pattern 3) or the same number of missing parts (pattern 4), and the reference position correction is not executed. The height of the gap is the absolute value of the difference between the coordinates of the upper end and the lower end.

In the correction of the reference position (algorithm 4), if the upper end is missing, the first reference position correction unit 17 determines the height of the missing portion from the numerical value of the reference position indicating the beginning of the image data. Subtract half the number. In this case, the numerical values of the coordinates of the upper end and the lower end of the image data are reduced by half of the lack, and the image data is increased by only half of the lack of the coordinates.

However, since the actual image data on the bit map is not displaced, when the image data is read from the bit map based on the coordinates as described above, the height of the image data is relatively half that of the missing data. Will only descend. Accordingly, the coordinates of the center position are calculated by averaging the coordinates of the upper end and the lower end which have been raised as described above, and the character normalizing unit 5 converts the image data on the bit map so that the coordinates match the center position of the image area. When the image data is moved, the image data is arranged at an original position where no image data is missing.

If the lower end is missing, the second reference position correction unit 18 adds a half value of the height of the missing to the coordinates of the reference position of the image data, and the image data is shifted to the image area. It should be arranged at the original position without any missing parts. As described above, the reference position of the image data for which the missing direction is not determined is corrected by both the first and second reference position correction units 17 and 18. The value 0 is set in the memory content of the image data in which the missing part has been compensated by the correction of the reference position as described above.

Next, the character normalization unit 5 performs normalization for removing minute irregularities in the same manner as in the prior art, and calculates the coordinates of the center position by averaging the coordinates of the upper and lower ends of the image data. This is matched with the center position of the image area of the standard pattern to generate a recognition pattern (step S17).
At this time, since the coordinates of the upper end and the lower end of the image data are calculated from the reference position, the first and second reference position correctors 1
If the reference position is displaced in steps 7 and 18, the image data is aligned with the image area at the original position without any loss. Note that, as described above, since the reference position is displaced both up and down, recognition patterns displaced both up and down are generated for the image data in which the missing is detected in the entire field and the direction is not determined as described above.

Then, the character recognition section 7 reads out a large number of standard patterns from the standard pattern memory 6 in order and matches them with the recognition pattern (step S18), and outputs a character of the standard pattern having the highest matching accuracy as a recognition result. (Step S19). However, with respect to image data in which two types of recognition patterns are generated by omitting the entire field, a standard pattern is matched with the two types of recognition patterns, and one having the highest accuracy is selected. In this case, since the missing direction is determined for the first image data in that field, it is possible to specify the missing direction for the subsequent image data and omit one matching. As described above, the OCR system 11 according to the present embodiment determines the presence / absence of a vertical drop in image data in units of characters, detects the direction and height of the detected drop, and detects the detected drop. The position of the image data with respect to the image area is corrected in the vertical direction according to the direction and height of the image, so that the image data of characters with missing upper and lower parts can be arranged in the image area as if there was no missing , Characters with missing upper and lower portions can be recognized with good accuracy.

Further, since the presence or absence of a missing portion is determined by comparing the height of the image data in character units with a prescribed value prepared in advance, the presence or absence of the missing portion can be detected satisfactorily by simple processing. Furthermore, since the specified value of the height is corrected based on the height of the image data determined not to be missing, even when the height of the image data is entirely changed due to an error in printing or reading, Correspondingly, characters with missing upper and lower portions can be detected relatively favorably.

Further, image data for which the missing direction is not known is displaced both up and down, and the two matching results are compared to select one, so that a character for which the missing direction is not known can be well recognized. . In that case, even if the entire field is missing and the direction is not known, the direction can be determined if the first character is recognized, so that for subsequent characters, two pattern generations and matching can be omitted as one. .

The present invention is not limited to the above-described embodiment, but allows various modifications without departing from the scope of the invention. For example, in the above-described embodiment, an example has been described in which the coordinates of the reference position are moved up and down in the direction of the missing portion by half the value of the height of the missing portion in order to arrange the image data whose upper and lower portions are missing at the original position of the image area. However, if the upper part is missing, subtract the height of the missing part from the coordinates of the upper end of the image data, and if the lower part is missing, add the height of the missing part to the coordinates of the lower end. Other things are possible.

Further, in the above-described embodiment, an example has been described in which various units of the OCR system 11 are realized by the operation of the CPU 101 according to a control program stored as software in the RAM 104 or the like. But,
It is also possible to form each of these various means as unique hardware, and it is also possible to store a part of the means as software in the RAM 104 or the like and form a part as hardware.

In the above embodiment, when the computer system 12 is started, the software stored in the HDD 105 is copied to the RAM 104, and the software stored in the RAM 104 is copied to the CPU 101.
However, such software can be used by the CPU 101 while the software is stored in the HDD 105, or can be fixedly written in the ROM 103 in advance. Further, software is written in the FD 106 or the CD-ROM 108 which is an information storage medium which can be handled alone, and the RA is read from the FD 106 or the like.
Although software can be installed in the M104 or the like, the CPU 101 can directly read the software from the FD 106 or the like and execute the processing operation without performing such installation.

That is, when the various means of the character recognition apparatus of the present invention are realized by software, the software only needs to be in a state where the CPU 101 can read and execute the corresponding operation. It is also possible to form a control program for realizing the various means as described above by combining a plurality of software. In this case, the information storage medium as a single product includes the character recognition device of the present invention. It is only necessary to store only the minimum necessary software for realization.

For example, a CD-ROM is stored in a computer system 12 on which an existing operating system is mounted.
In the case where application software is provided by an information storage medium such as the ROM 108, software for implementing various units of the character recognition device of the present invention is implemented by a combination of application software and an operating system, and thus depends on the operating system. Part of the software can be omitted from the application software of the information storage medium.

The method of supplying the software described on the information storage medium to the CPU 101 is not limited to loading the information storage medium directly into the computer system 12. For example, it is also possible to store the software as described above in an information storage medium of a host computer, connect the host computer to a terminal computer via a communication network, and supply the software from the host computer to the terminal computer by data communication. is there.

[0063] When as described above, the terminal computer itself
It is also possible to execute a stand-alone processing operation with the software downloaded to the personal information storage medium, but it is also possible to execute the processing operation by real-time data communication with the host computer without downloading the software. is there. In this case, the entire system in which the host computer and the terminal computer are connected by a communication network corresponds to the character recognition device of the present invention.

[0064]

Since the present invention is configured as described above, it has the following effects.

According to the first aspect of the present invention, a standard pattern of a character to be recognized is set in advance, characters of a predetermined height marked on a form are optically read, and the read image data is converted into a character unit. The center position is aligned with the center position of the image area of the standard pattern, and the image data with the aligned center position is sequentially matched with many standard patterns as a recognition pattern, and the recognition result of the character of the standard pattern with the highest matching accuracy is obtained. In a character recognition device that outputs as: a missing part determining unit that determines whether there is a missing part in the vertical direction of the image data cut out in character units, a direction detecting unit that detects the direction of the missing part whose presence is detected, Height detecting means for detecting the height of the detected drop, and image data corresponding to the direction and height of the detected drop. Correcting the position with respect to di-region in the vertical direction, the image data direction of missing is not found in upper and lower
And position correcting means Ru is displaced towards the two matching results
And the result selection means for comparing and selecting one of them, the image data of the character with the upper and lower portions missing can be arranged in the image area in the same manner as the state without the missing portion. Characters with missing characters can be recognized with good accuracy, and characters with unknown directions
Can be recognized well.

According to a second aspect of the present invention, there is provided the character recognition apparatus according to the first aspect, further comprising a character height storage unit in which a prescribed value of a character height is set in advance, By comparing the height of the image data with the specified value to determine the presence or absence of a missing portion, the presence or absence of the missing portion can be satisfactorily detected by simple processing.

According to a third aspect of the present invention, there is provided the character recognition apparatus according to the second aspect, further comprising a data correcting means for correcting a specified value of the height based on the height of the image data determined to have no loss. As a result, even if the height of the image data fluctuates overall due to printing or reading errors, it is possible to relatively well detect characters with missing upper and lower portions in response to this. it can.

[0068]

According to a fourth aspect of the present invention, a standard pattern of a character to be recognized is set in advance, characters of a predetermined height marked on a form are optically read, and the read image data is converted into a character unit. The center position is aligned with the center position of the image area of the standard pattern, and the image data with the aligned center position is sequentially matched with many standard patterns as a recognition pattern, and the recognition result of the character of the standard pattern with the highest matching accuracy is obtained. In the character recognition method that is output as, the presence / absence of a vertical drop of the image data cut out in character units is determined, the direction of the missing where the presence is detected is detected, and the presence of the missing is detected. Detects the height and corrects the position of the image data with respect to the image area in the vertical direction according to the direction and height of the detected dropout. By doing so, the image data of the characters with the upper and lower parts missing can be arranged in the image area in the same way as the state without the missing, so that the characters with the upper and lower parts missing can be recognized with good accuracy. it can.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating a character recognition apparatus according to an embodiment of the present invention;
It is a schematic diagram which shows the logical structure of CR system.

FIG. 2 is a block diagram showing a physical structure of the OCR system.

FIG. 3 is a schematic diagram showing a standard recognition pattern.

FIG. 4 is a schematic diagram showing a recognition pattern in which each part is missing together with a standard recognition pattern.

FIG. 5 is a flowchart showing the first half of a character recognition method by the OCR system.

FIG. 6 is a flowchart showing the second half of the character recognition method by the OCR system.

FIG. 7 is a schematic diagram showing an OCR system that is a conventional character recognition device.

FIG. 8 is a schematic diagram showing a standard pattern and a recognition pattern.

[Explanation of symbols]

2 Image extraction unit 3 Multi-valued image memory 4 Character extraction unit 5 Character normalization unit 6 Standard pattern memory 7 Character recognition unit 11 OCR system which is a character recognition device 12 Computer system 13 Image scanner 14 Field coordinate memory 15 Character height storage means Certain standard character height memory 16 Missing detecting section corresponding to various means such as missing detecting means 17 First reference position correcting section corresponding to position correcting means 18 Second reference position correcting section corresponding to position correcting means 101 Computer Certain CPU 102 Bus line 103 ROM 104 as information storage medium RAM 105 as information storage medium HDD 106 as information storage medium FD 107 FDD 108 as information storage medium CD-ROM 109 as information storage medium CD drive 110 Keyboard 111 Mau S 112 Display 113 Communication I / F

──────────────────────────────────────────────────続 き Continued on front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G06K 9/62 610

Claims (4)

(57) [Claims] 1. A standard pattern of a character to be recognized is set in advance, characters of a predetermined height marked on a form are optically read, and the read image data is cut out in character units to determine a center position. A character recognition device that matches the center position of the image area of the standard pattern, sequentially matches the image data with the aligned center position as a recognition pattern with a number of standard patterns, and outputs a character of the standard pattern with the highest matching accuracy as a recognition result. A drop determining means for determining the presence or absence of a drop in the vertical direction of the image data cut out in character units; a direction detecting means for detecting a direction of the drop in which the presence is detected; and a height of the drop in which the presence is detected. Height detecting means for detecting the height of the image data, and the position of the image data with respect to the image area in accordance with the detected direction and height of the defect. And a position correcting means for displacing image data for which the missing direction is not known in both directions, and a result selecting means for comparing two matching results and selecting one of them. Character recognition device. 2. A character height storage means in which a prescribed value of a character height is set in advance, and a missing determining means compares the height of image data in character units with a prescribed value to determine the presence or absence of missing data. The character recognition device according to claim 1, wherein the character recognition is performed. 3. The character recognition device according to claim 2, further comprising data correction means for correcting a specified value based on the height of the image data determined to be missing. 4. A standard pattern of a character to be recognized is set in advance, characters of a predetermined height marked on a form are optically read, and the read image data is cut out in character units to determine a center position. The center position of the image area of the standard pattern is aligned, the image data with the aligned center position is sequentially matched as a recognition pattern with a number of standard patterns, and the characters of the standard pattern with the highest matching accuracy are output as the recognition result. In the character recognition method, the presence / absence of a missing part in the vertical direction of the image data cut out in character units is determined, the direction of the missing part detected is detected, and the height of the missing part detected is detected. The position of the image data with respect to the image area is corrected vertically according to the detected direction and height of the missing part, and the direction of the missing part is determined. Character recognition method without image data is displaced vertically both, characterized by being adapted to select one by comparing the two matching results.