JP2008165339A - Business form identification unit and business form identification program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve identification efficiency of a business form. <P>SOLUTION: From a business form image data input from an input unit 102, a read unit 103 reads a pattern described in each particular point of the business form, as a read pattern of each particular point. For each business form section, a calculator 104 calculates a distance value between the read pattern read by the read unit 103 and a registration pattern of each particular point stored in a storage 101. A totaling unit 105 totals the distance values of the respective particular points calculated by the calculator 104 for each business form section. Based on the distance value totaled for each business form section by the totaling unit 105, a decision section 106 in a processing unit 152 decides which business form section the business form image data input by the input unit 102 pertains to. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a form identification device and a form identification program. The present invention particularly relates to a form identification method in an OCR (optical character recognition) system.

Conventional OCR systems (OCR systems including FAX OCR) receive feature information (patterns) of ruled lines of images received by FAX or read by a scanner, and form IDs (identifiers) represented by marks and numbers Is a system for recognizing handwritten characters described in a specific field and converting the recognized characters into data. In this conventional system, as a procedure for recognizing a form, after a ruled line pattern is verified, a form ID is verified and a form is specified. In a system that processes a large amount of forms, it is common to automatically identify a form in this way and distribute it to a workflow according to each form (for example, see Patent Documents 1 to 5).
JP-A-9-231291 JP-A-10-11531 Japanese Patent Laid-Open No. 11-259597 JP 2004-164673 A JP 2003-263608 A

  In the conventional system, first, the registered data and the ruled line pattern are collated. If the collation fails, error data is output at that time and the processing is stopped. Only when the matching of the ruled line pattern is successful, the registered data and the form ID are verified. If the verification fails, the error data is similarly output and the processing is stopped. Only when the verification of the form ID is successful, the form can be specified. However, some of the data received by fax usually includes fading due to noise, smearing of the fax machine, paper breakage, partial loss of image due to tilt (same for data read by a scanner), and ruled lines There were many cases where an error occurred because the pattern or form ID could not be acquired. As described above, in the conventional system, it is often impossible to automatically specify the form, and manual work by the operator is required from the primary distribution to the workflow, which causes a problem that work efficiency is deteriorated.

  An object of the present invention is to improve the identification efficiency of a form.

The form identification device according to one aspect of the present invention is:
A storage unit that stores in advance in the storage device a registered pattern of each specific location as a registered pattern of each specific location for each category of the form determined by a description pattern in at least two specific locations of the form;
An input unit for inputting form image data from an input device;
A reading unit that reads a pattern described in each specific part of the form from the image data of the form input by the input unit by a processing device as a reading pattern of each specific part;
A calculation unit that calculates a distance value between a reading pattern of each specific location read by the reading unit and a registration pattern of each specific location stored by the storage unit for each section of the form,
For each section of the form, a totaling unit that counts the distance value of each specific location calculated by the calculation unit with a processing device;
A determination unit that determines in which processing unit the image data of the form input by the input unit is based on a distance value for each category of the form that is aggregated by the aggregation unit; It is characterized by that.

  The totaling unit sorts the distance values for each category of the form calculated by the calculating unit in ascending order for each specific part of the form, and according to the difference between the distance value of each rank and the distance value of the next rank Weighting is performed, and the distance value of each weighted specific portion is totaled for each category of the form.

The storage unit stores a threshold value of a distance value in advance in a storage device for each specific part of the form,
The totaling unit weights only the distance value exceeding the distance value threshold stored in the storage unit among the distance values for each category of the form sorted in ascending order for each specific part of the form. Features.

The storage unit stores a pattern of an identifier that uniquely indicates a category of a form as a registration pattern of one specific location,
The reading unit reads the pattern of the identifier written in the specific part from the image data of the form input by the input unit as the reading pattern of the one specific part,
The calculation unit calculates a distance value between the identifier pattern read by the reading unit and the identifier pattern stored by the storage unit for each category of the form.

The storage unit stores a ruled line pattern as a registration pattern of one specific location,
The reading unit reads a ruled line pattern written in the specific part from the image data of the form input by the input unit as the one specific part reading pattern,
The calculation unit calculates a distance value between a ruled line pattern read by the reading unit and a ruled line pattern stored by the storage unit for each category of the form.

The storage unit stores a pattern of a specific character string as a registration pattern of one specific location,
The reading unit reads a pattern of a specific character string written in the specific part from the image data of the form input by the input unit as the reading pattern of the one specific part,
The calculation unit calculates a distance value between the pattern of the specific character string read by the reading unit and the pattern of the specific character string stored by the storage unit for each category of the form.

When the reading unit can only read a partial pattern of the specific character string written in the specific part from the image data of the form input by the input unit as the reading pattern of the one specific part, the specific part Guess the rest of the specific character string written in the place, correct the pattern of the read specific character string,
The calculation unit calculates a distance value between the specific character string pattern corrected by the reading unit and the specific character string pattern stored by the storage unit for each category of the form.

The form identification device further includes:
Recognition in which the processing device performs optical character recognition of the image data of the form input by the input unit based on the classification of the form determined by the determination unit and the registered pattern of each specific location stored by the storage unit And
And an output unit that outputs a result of optical character recognition performed by the recognition unit from an output device.

The form identification program according to one aspect of the present invention is
A form identification program that uses a storage device that pre-stores a description pattern in each specific part of a form as a registration pattern for each specific part for each category of the form determined by a description pattern in at least two specific parts of the form,
Input processing to input the image data of the form from the input device;
A reading process in which a processing device reads a pattern described in each specific part of the form from the image data of the form input by the input process as a reading pattern of each specific part;
A calculation process for calculating a distance value between a reading pattern of each specific location read by the reading process and a registered pattern of each specific location stored in the storage device for each section of the form,
For each form category, a totaling process for summing up the distance values of each specific location calculated by the calculation process with a processing device;
A determination process for determining in which processing unit the image data of the form input by the input process belongs to a processing unit based on the distance value for each of the divisions of the form totaled by the tabulation process. It is made to perform.

  According to one aspect of the present invention, in the form identification apparatus, the calculation unit includes, for each section of the form, a reading pattern of each specific place read by the reading unit and a registration pattern of each specific place stored by the storage unit. The totaling unit calculates the distance value of each specific part calculated by the calculation unit for each category of the form, and the determination unit calculates the distance value of each form calculated by the totaling unit. Based on this distance value, the classification efficiency of the form is improved by determining which category of form the image data of the form input by the input unit belongs to.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

Embodiment 1 FIG.
The form identification apparatus according to the present embodiment refers to a registered form format and simultaneously evaluates a plurality of identification conditions in order to determine the type (classification) of the form. For example, by performing a ruled line pattern test, a form ID test, and a specific character string test (whether there is a specific character string) in parallel, the three identification conditions can be evaluated simultaneously. The form identification device then determines the type of form by integrating the reliability obtained for each identification condition.

  As described above, in the present embodiment, the form identification apparatus evaluates a plurality of identification conditions at the same time, so that even when there is an error in the evaluation of one identification condition, it is possible to determine a stable form. Therefore, the form identification efficiency is improved.

  Hereinafter, the configuration of the form identification apparatus according to the present embodiment will be described.

  FIG. 1 is a block diagram showing a configuration of a form identification apparatus 100 according to the present embodiment.

  In FIG. 1, the form identification apparatus 100 includes a storage unit 101, an input unit 102, a reading unit 103, a calculation unit 104, a totaling unit 105, a determination unit 106, a recognition unit 107, and an output unit 108. In addition, the form identification device 100 includes hardware devices such as a storage device 151, a processing device 152, an input device 153, and an output device 154 (or these hardware devices are connected to the form identification device 100). The hardware device is used by each unit of the form identification device 100. For example, the processing device 152 is used for performing calculation, processing, reading, writing, and the like of data and information in each unit of the form identification device 100. The storage device 151 is used to store the data and information. The input device 153 is used to input the data and information, and the output device 154 is used to output the data and information.

  The storage unit 101 stores, in advance in the storage device 151, a pattern described in each specific place of the form as a registered pattern of each specific place for each category of the form. The category of the form is determined by the description pattern in at least two specific locations of the form.

  Here, how the category of the form is determined will be described using an example shown in FIG.

  In FIG. 2, there are four forms of forms (A) to (D). In this example, the category of the form is determined by a ruled line pattern written near the center of the form, a form ID pattern written near the upper left of the form, and a specific character string pattern written near the upper right of the form. The form ID is an example of an identifier that uniquely indicates a form classification, and is represented by a three-digit number in this example. The specific character string is a character string predetermined for each category of the form, and in this example, the title of the form is represented as a character string.

  In the example shown in FIG. 2, the ruled line pattern of the form (A) and the ruled line pattern of the form (B) are the same, but the ruled line pattern of the form (C) is slightly different, and the ruled line of the form (D) The pattern is much different. The specific character strings of the forms (A), (C), and (D) are the same, but the specific character strings of the form (B) are different. Specifically, the specific character string of the forms (A), (C), and (D) is “Order Form”, and the specific character string of the form (B) is “Invoice”. Therefore, in this example, it can be said that the form classification is determined by a combination of a ruled line pattern and a specific character string pattern.

  In the example shown in FIG. 2, the forms (A) to (D) are all classified into different forms, and therefore the form IDs of the forms (A) to (D) are naturally different. Specifically, the form ID of the form (A) is “123”, the form ID of the form (B) is “128”, the form ID of the form (C) is “390”, and the form ID of the form (D) is “ 456 ". If the form ID is written on the form in this way, not only the combination of the ruled line pattern and the specific character string pattern but also the form ID pattern can be verified together, further improving the identification efficiency of the form. To do.

  In the following, in each form, the part where the ruled line is written (near the center of the form) is “ruled line written part”, the part where the form ID is written (near the upper left of the form) is “ID written part”, a specific character string A part where “” is written (near the upper right corner of the form) is called a “character string entry part”. The ruled line description location, the ID description location, and the character string description location are all examples of specific locations of the form.

  In the example illustrated in FIG. 2, the storage unit 101 stores, for each of the forms (A) to (D), a pattern described in a ruled line description location (that is, a ruled line pattern) as a registration pattern of the ruled line description location. 151. Also, the pattern described in the ID description location (that is, the form ID pattern) is stored in the storage device 151 as a registration pattern of the ID description location. In addition, the pattern described in the character string description location (that is, the pattern of the specific character string) is stored in the storage device 151 as a registered pattern of the character string description location. Hereinafter, the registration pattern stored in the storage device 151 is also referred to as “registration data”, “master information”, or “correct answer data”.

  The input unit 102 inputs form image data from the input device 153. For example, the input unit 102 receives form image data from a FAX or reads it from a scanner. In the example illustrated in FIG. 2, the input unit 102 receives image data of a certain form X from the FAX.

  The reading unit 103 reads, from the image data of the form input by the input unit 102, the pattern described in each specific place of the form with the processing device 152 as a read pattern of each specific place. In the example illustrated in FIG. 2, the reading unit 103 reads a ruled line pattern written in the ruled line description part of the form X from the image data of the form X as a ruled line description part reading pattern. Further, as the ID pattern reading pattern, the form ID pattern written in the ID description part of the form X is read from the image data of the form X. Also, as a reading pattern of the character string description location, a pattern of a specific character string written in the character string description location of the form X is read from the image data of the form X. As a result, the reading patterns of the ruled line description location, ID description location, and character string description location of the form X are obtained.

  The calculation unit 104 calculates a distance value between the reading pattern of each specific location read by the reading unit 103 and the registered pattern of each specific location stored in the storage unit 101 by the processing device 152 for each category of the form. In the example illustrated in FIG. 2, the calculation unit 104 calculates the distance value between the reading pattern of the ruled line description location of the form X and the registered pattern of each ruled line description location of the forms (A) to (D). Further, the calculation unit 104 calculates a distance value between the reading pattern of the ID description location of the form X and the registered pattern of the ID description location of each of the forms (A) to (D). In addition, the calculation unit 104 calculates a distance value between the reading pattern of the character string description portion of the form X and the registered patterns of the character string description portions of the forms (A) to (D). As a result, distance values (3 × 4 = 12 distances) between the read pattern of the form X and the registered patterns of the forms (A) to (D) for each of the ruled line description position, the ID description position, and the character string description position. Value). The obtained distance value may be stored in the storage device 151. Hereinafter, the distance value stored in the storage device 151 is also referred to as “accuracy data”.

  The totaling unit 105 uses the processing device 152 to total the distance values of the specific parts calculated by the calculation unit 104 for each category of the form. Specifically, the totaling unit 105 first sorts the distance values for each category of the form calculated by the calculation unit 104 in ascending order for each specific part of the form (when rearranging in descending order, in the subsequent processing) Think in reverse order). Then, the distance value of each rank is weighted according to the difference from the distance value of the next rank. After that, the totaling unit 105 sums up the distance values of the specific portions that have been weighted for each category of the form. In the example illustrated in FIG. 2, the totaling unit 105 rearranges the four distance values in ascending order for each of the ruled line description location, the ID description location, and the character string description location. Then, for each of the ruled line description location, the ID description location, and the character string description location, the first distance value is weighted according to the difference between the second distance value and the third distance value. Weighting is performed according to the difference from the distance value, and weighting is performed according to the difference from the fourth distance value to the third distance value. Thereafter, the totaling unit 105 adds up the three distance values of the weighted ruled line description location, ID description location, and character string description location for each of the forms (A) to (D). As a result, distance values between the form X and the forms (A) to (D) are obtained. The weighting method will be described later.

  Here, the memory | storage part 101 may memorize | store the threshold value of a distance value in the memory | storage device 151 beforehand for every specific location of a form. Then, the totaling unit 105 weights only the distance values exceeding the distance value threshold stored in the storage unit 101 among the distance values for each category of the form sorted in ascending order for each specific part of the form. May be. In the example of FIG. 2, the storage unit 101 stores a distance value threshold value in advance in the storage device 151 for each of the ruled line description location, the ID description location, and the character string description location. Then, the totaling unit 105 weights only the distance values from the first place to the fourth place that exceed the threshold of the distance value for each of the ruled line description position, the ID description position, and the character string description position. Since the counting unit 105 does not count the distance values that are not weighted, the processing efficiency is improved.

  Based on the distance value for each category of the form totaled by the totalization unit 105, the determination unit 106 determines in which classification the form image data input by the input unit 102 belongs to. To do. Here, assuming that the smaller the distance value between the forms, the closer the form pattern is, in the example of FIG. 2, in the example of FIG. 2, the determination unit 106 assigns the form classification with the smallest distance value from the form X to the form ( A) to (D) are selected and determined as the category of the form X.

  The recognizing unit 107 processes the OCR of the image data of the form input by the input unit 102 based on the form classification determined by the determining unit 106 and the registered pattern of each specific location stored by the storage unit 101. 152. In the example illustrated in FIG. 2, the recognition unit 107 includes a ruled line description location, an ID description location, and a character string description location corresponding to the category of the form selected from the forms (A) to (D) by the determination unit 106. The registered pattern is read from the storage device 151, and OCR is performed after recognizing each pattern of ruled line, form ID, and specific character string included in the image data of the form X. Thereby, the identification efficiency of a form improves.

  The output unit 108 outputs the result of the OCR performed by the recognition unit 107 from the output device 154. In the example illustrated in FIG. 2, the output unit 108 outputs the OCR result of the form X. Hereinafter, the OCR result of the form X is also referred to as “identification result data”.

  Here, the above-described weighting method will be described using the example shown in FIG.

(1) Case in which the weight of the verification result of the form ID is strong For example, it is assumed that the form is classified into forms (A) and (C), and the form (C) is the form X. In this case, the form (C) has a very similar ruled line pattern to the form (A), and the specific character string is the same, so the accuracy of the ruled line pattern test and the test for the presence of the specific character string is low. . In comparison, the form ID “390” of the form (C) and the form ID “123” of the form (A) are less likely to be misrecognized, so the accuracy of the form ID verification is high. Therefore, the totaling unit 105 reduces the weighting (or does not perform weighting) for the distance values of the ruled line description part and the character string description part, and increases the weighting for the distance value of the ID description part. Good. Thereby, since the distance value of the ID description location is strongly reflected in the distance value that is subsequently counted by the counting unit 105, the identification efficiency of the form is improved.

(2) Case in which the weight of the ruled line pattern verification result is high For example, it is assumed that the forms are classified into forms (A) and (D), and the form (D) is the form X. In this case, since the form (D) has the same specific character string as the form (A), the accuracy of the test for the presence of the specific character string is low. The form ID “456” of the form (D) and the form ID “123” of the form (A) are less likely to be mistakenly recognized, and the form ID test has a certain degree of accuracy. This increases the accuracy of the ruled line pattern test. Therefore, the totaling unit 105 weakens (or does not perform weighting) the distance value of the character string description location, performs some weighting on the distance value of the ID description location, and determines the distance of the ruled line description location. What is necessary is just to make weighting strong about a value. Thereby, since the distance value of the ruled line description portion is strongly reflected in the distance value that is subsequently counted by the counting unit 105, the identification efficiency of the form is improved.

(3) Case in which the weight of the test result for the presence / absence of a specific character string is increased. For example, it is assumed that the forms are classified into forms (A) and (B), and the form (B) is the form X. In this case, the form (B) has the same ruled line pattern as the form (A), and the form ID “128” of the form (B) and the form ID “123” of the form (A) are also the first and second digits from the left. The third digit “8” and “3” are likely to be erroneously recognized. Therefore, the accuracy of ruled line pattern verification and form ID verification is low. Compared to this, the specific character string “Invoice” of the form (B) and the specific character string “Order Form” of the form (A) are different, so the accuracy of the test for the presence or absence of the specific character string is increased. Therefore, the totaling unit 105 reduces the weighting (or does not perform weighting) for the distance values of the ruled line description location and the ID description location, and increases the weighting for the distance value of the character string description location. Good. Thereby, since the distance value of the character string description portion is strongly reflected in the distance value that is subsequently counted by the counting unit 105, the identification efficiency of the form is improved. In particular, ruled lines are vulnerable to image degradation (noise, blurring, etc.) at the time of FAX and may be interrupted and may not be recognized. However, a character string with a large font such as a title is highly likely to be recognized even if it is blurred. A special effect can be obtained by the presence or absence test.

  As described above, in the present embodiment, the totaling unit 105 weights the distance value for each section of the form for each specific place of the form, and each weighted specific place for each section of the form. Sum the distance values. That is, the totaling unit 105 integrates the obtained reliability from the accuracy of each test result. As a result, it is possible to determine the type of form more reliably and output the OCR recognition result than before. For example, even if an image is partially degraded and cannot be recognized as a form by ruled line information alone, according to the present embodiment, there is a possibility that it can be recognized as a form from the form ID test result and the specific character string test result. There is.

  FIG. 3 is a diagram illustrating an example of hardware resources of the form identification device 100.

  In FIG. 3, a form identification apparatus 100 includes a system unit (not shown), a display device 901 having a display screen of a CRT (Cathode / Ray / Tube) or LCD (Liquid Crystal Display), a keyboard 902 (K / B), a mouse 903, Hardware resources such as an FDD 904 (Flexible Disk Drive), a CDD 905 (Compact Disk Drive), a printer device 906, and a scanner device 907 are provided, and these are connected by cables and signal lines. The system unit is a computer, and is connected to a facsimile machine (FAX) (not shown) with a cable, and is connected to a network such as a LAN (local area network).

  The form identification apparatus 100 includes a CPU 911 (Central Processing Unit) (also referred to as “arithmetic unit”, “microprocessor”, “microcomputer”, or “processor”) that executes a program. The CPU 911 is an example of the processing device 152. The CPU 911 includes a ROM 913 (Read / Only / Memory), a RAM 914 (Random / Access / Memory), a communication board 915, a display device 901, a keyboard 902, a mouse 903, an FDD904, a CDD905, a printer device 906, and a scanner device via a bus 912. 907 is connected to the magnetic disk device 920 and controls these hardware devices. Instead of the magnetic disk device 920, a storage medium such as an optical disk device or a memory card reader / writer may be used.

  The RAM 914 is an example of a volatile memory. The storage media of the ROM 913, the FDD 904, the CDD 905, and the magnetic disk device 920 are an example of a nonvolatile memory. These are examples of the storage device 151. A communication board 915 or a FAX connected to the communication board 915, a keyboard 902, a mouse 903, a scanner device 907, an FDD 904, a CDD 905, and the like are examples of the input device 153. The communication board 915, the display device 901, the printer device 906, and the like are examples of the output device 154.

  The communication board 915 is connected to a FAX, a LAN, or the like. The communication board 915 is not limited to a LAN, but is the Internet, or an IP-VPN (Internet, Protocol, Private, Network), a wide area LAN, a WAN (Wide Area Network) such as an ATM (Asynchronous, Transfer, Mode) network, or the like. It does not matter if it is connected to.

  The magnetic disk device 920 stores an operating system 921 (OS), a window system 922, a program group 923, and a file group 924. The programs in the program group 923 are executed by the CPU 911, the operating system 921, and the window system 922. The program group 923 stores programs for executing functions described as “˜unit” and “˜means” in the description of the present embodiment. The program is read and executed by the CPU 911. The file group 924 includes data and information described as “˜data”, “˜information”, “˜ID (IDentifier)”, “˜flag”, “˜result” in the description of this embodiment. Signal values, variable values, and parameters are stored as items of “˜file”, “˜database”, and “˜table”. The “˜file”, “˜database”, and “˜table” are stored in a storage medium such as a disk or a memory. Data, information, signal values, variable values, and parameters stored in a storage medium such as a disk or memory are read out to the main memory or cache memory by the CPU 911 via a read / write circuit, and extracted, searched, referenced, compared, and calculated. Used for processing (operation) of the CPU 911 such as calculation / control / output / printing / display. Data, information, signal values, variable values, and parameters are temporarily stored in the main memory, cache memory, and buffer memory during processing of the CPU 911 such as extraction, search, reference, comparison, calculation, control, output, printing, and display. Is remembered.

  In the block diagrams and flowcharts described in the description of this embodiment, the arrows indicate mainly input and output of data and signals. Data and signals are stored in a memory such as a RAM 914, a flexible disk (FD) of the FDD 904, and a CDD 905. Recording is performed on a recording medium such as a compact disk (CD), a magnetic disk of the magnetic disk device 920, other optical disks, a mini disk (MD), and a DVD (Digital Versatile Disc). Data and signals are transmitted by a bus 912, a signal line, a cable, and other transmission media.

  In the description of the present embodiment, what is described as “to part” and “to means” may be “to circuit”, “to device”, and “to device”, and “to step”. , “˜step”, “˜procedure”, and “˜treatment”. That is, what is described as “˜unit” and “˜means” may be realized by firmware stored in the ROM 913. Alternatively, it may be realized only by software, or only by hardware such as an element, a device, a board, and wiring, or a combination of software and hardware, and further by a combination of firmware. Firmware and software are stored as programs in a recording medium such as a magnetic disk, flexible disk, optical disk, compact disk, minidisk, or DVD. This program is read by the CPU 911 and executed by the CPU 911. In other words, the program causes the computer to function as “to part” and “to means” described in the description of the present embodiment. Alternatively, the procedure or method of “˜unit” and “˜means” described in the description of the present embodiment is executed by a computer.

  Hereinafter, the operation of the form identification apparatus 100 will be described. In the following, in order to make the description more specific, it is assumed that the form identification device 100 is realized by the computer and hardware resources illustrated in FIG. In addition, for the classification of the form, the example shown in FIG. 2 is used.

  FIG. 4 is a flowchart showing the operation of the form identification device 100. The flow shown in the flowchart of FIG. 4 corresponds to a processing procedure of a program (form identification program) executed on a computer that implements the form identification device 100. FIG. 5 is a flowchart showing this processing procedure.

  4 and 5, the storage unit 101 uses a magnetic disk device as registration data 408 in advance as registration data 408 for each of the forms (A) to (D). It is assumed that the data is stored in 920 (an example of the storage device 151). Further, it is assumed that the storage unit 101 stores in advance in the magnetic disk device 920 all the threshold values of the distance values of the ruled line description location, the ID description location, and the character string description location as “−100”. Here, it is assumed that the closer the distance value of the two patterns is to “0”, the closer the two patterns are.

  The input unit 102 receives the image data of the form X from a FAX (an example of the input device 153) or takes in from the scanner device 907 (an example of the input device 153) (step S401). The input unit 102 converts the image data into a file, and distributes it to each process of ruled line pattern verification, form ID verification, and specific character string verification (step S402).

  The reading unit 103 extracts a reading pattern of a ruled line description portion of the form X from the image data of the form X by the CPU 911 (an example of the processing device 152). Then, the calculation unit 104 verifies the reading pattern of the ruled line description location of the form X, and as a result, the CPU 911 calculates the distance value between the read pattern and the registered pattern of each ruled line description location of the forms (A) to (D). (Step S403a). The calculation unit 104 stores the calculated distance value in the magnetic disk device 920 as the accuracy data 404a (step S404a). FIG. 6A shows an example of the accuracy data 404a. In this example, the distance value between the reading pattern of the ruled line description portion of the form X and the registration pattern of the ruled line description portion of the forms (A) and (B) is “−20”, and the ruled line description portion of the form (C) is registered. The distance value from the pattern is “−30”, and the distance value from the registered pattern of the ruled line description portion of the form (D) is “−60”.

  In parallel with the processing described above, the reading unit 103 extracts the reading pattern of the ID description location of the form X from the image data of the form X by performing OCR of the ID description location of the form X by the CPU 911. Then, the calculation unit 104 verifies the reading pattern of the ID description portion of the form X, and as a result, the CPU 911 calculates the distance value between the reading pattern and the registered pattern of each ID description portion of the forms (A) to (D). (Step S403b). The calculation unit 104 stores the calculated distance value in the magnetic disk device 920 as the accuracy data 404b (step S404b). FIG. 6B shows an example of the accuracy data 404b. In this example, the distance value between the reading pattern of the ID description location of the form X and the registration pattern of the ID description location of the form (A) is “−50”, and the distance value between the registration pattern of the ID description location of the form (B). Is “−75”, the distance value from the registered pattern of the ID description location of the form (C) is “−5”, and the distance value from the registration pattern of the ID description location of the form (D) is “−110”. Yes.

  In parallel with the above processing, the reading unit 103 extracts the reading pattern of the character string description portion of the form X from the image data of the form X by performing the OCR of the character string description portion of the form X by the CPU 911. And the calculation part 104 verifies the reading pattern of the character string description location of the form X, and, as a result, the distance value of the said reading pattern and the registration pattern of each character string description location of the form (A)-(D) Is calculated by the CPU 911 (step S403c). The calculation unit 104 stores the calculated distance value in the magnetic disk device 920 as the accuracy data 404c (step S404c). FIG. 6C shows an example of the accuracy data 404c. In this example, the distance value between the read pattern of the character string description portion of the form X and the registered pattern of the character string description portion of the forms (A), (C), and (D) is “−20”, respectively, and the form (B). The distance value from the registered pattern of the character string description location is “−45”.

  The totaling unit 105 rearranges the four distance values in ascending order for each of the ruled line description location, the ID description location, and the character string description location. Then, the counting unit 105 refers to the distance value threshold values stored in the magnetic disk device 920 for each of the ruled line description location, the ID description location, and the character string description location, and calculates the distance values from the first place to the fourth place. Among them, the distance value exceeding the threshold value is determined. For each of the ruled line description location, the ID description location, and the character string description location, the counting unit 105 sets only the distance value determined to exceed the threshold as the second distance value relative to the first distance value. Weighting according to the difference, weighting according to the difference between the second distance value and the third distance value, weighting according to the difference between the third distance value and the fourth distance value Do. Thereafter, the totaling unit 105 sums up the three distance values of the ruled line description location, the ID description location, and the character string description location for each of the forms (A) to (D) (step S405).

  The accuracy data 404a to 404c illustrated in (a) to (c) of FIG. 6 are rearranged in ascending order by the counting unit 105 and weighted, respectively, and are shown in (a) to (c) of FIG. . In this example, when the difference between the distance value of one rank and the distance value of the next rank is 0 to 9, the weight is “5” (times), and when it is 10 to 19, the weight is “4” (times). ), 20 to 29, the weight is "3" (times), 30 to 39, the weight is "2" (times), and 40 to 49, the weight is "1" (times). Shall be set. The smaller the weight coefficient, the stronger the weight, and the larger the weight coefficient, the weaker the weight.

  In FIG. 7A, the distance values of the accuracy data 404a are arranged in the form (A), (B), (C), (D) in the original arrangement order. The difference between the distance value of the first form (A) and (B) and the distance value of the third form (C) is 10 (since there are two first places, the next rank of the first place is 3). The weight for the distance values of the forms (A) and (B) is “4”. Similarly, the weight for the distance value of the form (C) is “2”. Since there is no fourth rank, no weighting is required for the distance value of the form (D).

  In FIG. 7B, the distance values of the accuracy data 404b are rearranged in the order of the forms (C), (A), (B), and (D). Since the difference between the distance value of the first form (C) and the distance value of the second form (A) is 45, the weight for the distance value of the form (C) is “1”. Similarly, the weights for the distance values of the forms (A) and (B) are “3” and “2” in order. Since there is no fourth rank, no weighting is required for the distance value of the form (D). Further, since the threshold value of the distance value is “−100”, the weight value is omitted for the distance value “−110” of the form (D) that does not exceed the threshold value even if the fifth rank or lower rank exists. The

  In FIG. 7C, the distance values of the accuracy data 404c are rearranged in the order of forms (A), (C), (D), and (B). The difference between the distance value of the first form (A), (C), (D) and the distance value of the fourth form (B) is 25 (since there are three first places, The weight for the distance values of the forms (A), (C), and (D) is “3”. Since there is no fourth rank, no weighting is required for the distance value of the form (D).

  The accuracy data 404a to 404c illustrated in (a) to (c) of FIG. 7 are weighted by the totaling unit 105, respectively, and are shown in (a) to (c) of FIG.

  In FIG. 7A, the distance values of the forms (A) and (B) of the accuracy data 404a are “−20” × “4” = “− 80”, and the distance value of the form (C) is “− 30 ”ד 2 ”=“ − 60 ”. In FIG. 7B, the distance value of the form (C) of the accuracy data 404b is “−5” × “1” = “− 5”, and the distance value of the form (A) is “−50”. × “3” = “− 150”, and the distance value of the form (B) is “−75” × “2” = “− 150”. In FIG. 7C, the distance values of the forms (A), (C), and (D) of the accuracy data 404c are “−20” × “3” = “− 60”.

  FIG. 8D illustrates the accuracy data 404a to 404c illustrated in FIGS. 8A to 8C that are aggregated by the aggregation unit 105. FIG.

  Of the forms (A) to (D), only the forms (A) and (C) are weighted for all three distance values of the ruled line description location, the ID description location, and the character string description location. It is. The total of the three distance values is “−290” for the form (A) and “−125” for the form (C).

  The determination unit 106 selects the category of the document having the smallest distance from the document X from among the documents (A) to (D), and determines it as the category of the document X by the CPU 911 (step S406). When the determination unit 106 fails to determine the category of the form X, the determination unit 106 stores the error data 407 in the magnetic disk device 920 or outputs it from the display device 901 (an example of the output device 154) to the screen, and then The process ends (step S407). In the example illustrated in FIG. 8D, the determination unit 106 determines that the category of the form X is a form (C). At this time, for example, when the threshold value for determination is set to “−100”, the determination unit 106 outputs the error data 407 because the distance value of the form (C) is “−125”. The process ends. For example, when the condition that the difference between the first and second distance values must be 200 or more is set, the difference between the distance value of the form (C) and the distance value of the form (A) is 165. Therefore, the determination unit 106 outputs error data 407 and ends the process.

  The recognizing unit 107 reads the registration data 408 corresponding to the category of the form selected from the forms (A) to (D) by the determining unit 106 from the magnetic disk device 920, and the ruled lines included in the image data of the form X The image data of the form X is analyzed with reference to the respective patterns of the form ID and the specific character string, and then the OCR of the image data of the form X is performed by the CPU 911 (step S408).

  The output unit 108 outputs the OCR result of the form X from the display device 901 to the screen as recognition result data 409 (step S409).

  As described above, in the present embodiment, when the form identification device 100 receives an image of a form, a plurality of identification conditions such as ruled line pattern verification, form ID verification, and whether or not there is a specific character are simultaneously applied. By evaluating, it becomes possible to recognize the form reliably.

Embodiment 2. FIG.
In the present embodiment, differences from the first embodiment will be mainly described.

  The configuration of the form identification apparatus 100 in the present embodiment is the same as that shown in FIG.

  In the present embodiment, the reading unit 103 uses a part of the specific character string written in the specific part from the image data of the form input by the input unit 102 as the reading pattern of the specific part where the specific character string is written. If only the above pattern can be read, the remainder of the specific character string written in the specific portion is estimated, and the pattern of the read specific character string is corrected. The calculation unit 104 calculates a distance value between the specific character string pattern corrected by the reading unit 103 and the specific character string pattern stored in the storage unit 101 for each category of the form. For example, the form shown in FIG. 9 includes three specific character strings “FAX order form”, “Tokyo center”, and “entry example”. The image data of this form is input by the input unit 102. In this case, it is assumed that the reading unit 103 can only read “FAX order form” as “F_X order_order” (“_” indicates a blank). In this case, the reading unit 103 searches the registered pattern of the character string description location stored in the storage unit 101, and extracts the “FAX order form” that is the specific character string closest to “F_X order_order”.

  FIG. 10 is a flowchart showing the operation of the form identification device 100 when the form identification device 100 performs a test for the presence or absence of a specific character string (step S403c in FIGS. 4 and 5).

  First, the reading unit 103 extracts a characteristic character string (in the example of FIG. 9, “FAX order form”, “Tokyo center”, “entry example”) (step S1001), and a highly accurate character string (FIG. 9). In this example, OCR is performed by selecting “FAX order form”) having a large font size (step S1002).

  When the character string is read in step S1002, the calculation unit 104 collates the character string with the registration data 408, and generates accuracy data 404c (step S1003).

  If a part of the character string cannot be read in step S1002, the reading unit 103 guesses the correct character string by the fuzzy search (if it is recognized as “F_X Note_Note”, the fuzzy search is used to make the “FAX order form”. (Step S1004). Then, the calculation unit 104 collates the character string with the registration data 408, and generates accuracy data 404c (step S1005).

  As described above, in the present embodiment, the form identification device 100 uses the character string ambiguous search technology, so that it is possible to determine whether a specific character string is included by simple character string matching. It is possible to perform column data collation.

  As mentioned above, although embodiment of this invention was described, you may implement combining 2 or more embodiment among these. Alternatively, one of these embodiments may be partially implemented. Or you may implement combining two or more embodiment among these partially.

1 is a block diagram illustrating a configuration of a form identification device according to Embodiment 1. FIG. 3 is a conceptual diagram illustrating an example of a form classification in Embodiment 1. FIG. 3 is a diagram illustrating an example of hardware resources of the form identification apparatus according to Embodiment 1. FIG. 3 is a flowchart showing the operation of the form identification device according to the first embodiment. 4 is a flowchart illustrating a processing procedure of a form identification program according to the first embodiment. 4 is a table showing an example of accuracy data in the first embodiment. 4 is a table showing an example of accuracy data in the first embodiment. 4 is a table showing an example of accuracy data in the first embodiment. 10 is a conceptual diagram illustrating an example of a form classification in Embodiment 2. FIG. 10 is a flowchart illustrating a part of the operation of the form identification device according to the second embodiment.

Explanation of symbols

  100 Form Identification Device, 101 Storage Unit, 102 Input Unit, 103 Reading Unit, 104 Calculation Unit, 105 Totaling Unit, 106 Judgment Unit, 107 Recognition Unit, 108 Output Unit, 151 Storage Device, 152 Processing Device, 153 Input Device, 154 Output device, 404 accuracy data, 407 error data, 408 registration data, 409 recognition result data, 901 display device, 902 keyboard, 903 mouse, 904 FDD, 905 CDD, 906 printer device, 907 scanner device, 911 CPU, 912 bus, 913 ROM, 914 RAM, 915 communication board, 920 magnetic disk device, 921 operating system, 922 window system, 923 program group, 924 file group.

Claims (9)

A storage unit that stores in advance in the storage device a registered pattern of each specific location as a registered pattern of each specific location for each category of the form determined by a description pattern in at least two specific locations of the form;
An input unit for inputting form image data from an input device;
A reading unit that reads a pattern described in each specific part of the form from the image data of the form input by the input unit by a processing device as a reading pattern of each specific part;
A calculation unit that calculates a distance value between a reading pattern of each specific location read by the reading unit and a registration pattern of each specific location stored by the storage unit for each section of the form,
For each section of the form, a totaling unit that counts the distance value of each specific location calculated by the calculation unit with a processing device;
A determination unit that determines in which processing unit the image data of the form input by the input unit is based on a distance value for each category of the form that is aggregated by the aggregation unit; A form identification device characterized by that.   The totaling unit sorts the distance values for each category of the form calculated by the calculating unit in ascending order for each specific part of the form, and according to the difference between the distance value of each rank and the distance value of the next rank 2. The form identification apparatus according to claim 1, wherein the weight values are added and the distance values of the weighted specific portions are totaled for each form classification. The storage unit stores a threshold value of a distance value in advance in a storage device for each specific part of the form,
The totaling unit weights only the distance value exceeding the distance value threshold stored in the storage unit among the distance values for each category of the form sorted in ascending order for each specific part of the form. The form identifying apparatus according to claim 2, wherein the form identifying apparatus is characterized. The storage unit stores a pattern of an identifier that uniquely indicates a category of a form as a registration pattern of one specific location,
The reading unit reads the pattern of the identifier written in the specific part from the image data of the form input by the input unit as the reading pattern of the one specific part,
The said calculation part calculates the distance value of the identifier pattern read by the said reading part, and the identifier pattern memorize | stored by the said memory | storage part for every classification | category of a form. Form identification device. The storage unit stores a ruled line pattern as a registration pattern of one specific location,
The reading unit reads a ruled line pattern written in the specific part from the image data of the form input by the input unit as the one specific part reading pattern,
The said calculation part calculates the distance value of the ruled line pattern read by the said reading part, and the ruled line pattern memorize | stored by the said memory | storage part for every division | segmentation of a form, The calculation part of Claim 1 characterized by the above-mentioned. Form identification device. The storage unit stores a pattern of a specific character string as a registration pattern of one specific location,
The reading unit reads a pattern of a specific character string written in the specific part from the image data of the form input by the input unit as the reading pattern of the one specific part,
The said calculation part calculates the distance value of the pattern of the specific character string read by the said reading part and the pattern of the specific character string memorize | stored by the said memory | storage part for every division | segmentation of a form. The form identification device according to 1. When the reading unit can only read a partial pattern of the specific character string written in the specific part from the image data of the form input by the input unit as the reading pattern of the one specific part, the specific part Guess the rest of the specific character string written in the place, correct the pattern of the read specific character string,
The calculation unit calculates a distance value between a specific character string pattern corrected by the reading unit and a specific character string pattern stored by the storage unit for each category of the form. The form identification device according to 6. The form identification device further includes:
Recognition in which the processing device performs optical character recognition of the image data of the form input by the input unit based on the classification of the form determined by the determination unit and the registered pattern of each specific location stored by the storage unit And
The form identification apparatus according to claim 1, further comprising: an output unit that outputs an optical character recognition result performed by the recognition unit from an output device. A form identification program that uses a storage device that pre-stores a description pattern in each specific part of a form as a registration pattern for each specific part for each category of the form determined by a description pattern in at least two specific parts of the form,
Input processing to input the image data of the form from the input device;
A reading process in which a processing device reads a pattern described in each specific part of the form from the image data of the form input by the input process as a reading pattern of each specific part;
A calculation process for calculating a distance value between a reading pattern of each specific location read by the reading process and a registered pattern of each specific location stored in the storage device for each section of the form,
For each form category, a totaling process for summing up the distance values of each specific location calculated by the calculation process with a processing device;
A determination process for determining in which processing unit the image data of the form input by the input process belongs to a processing unit based on the distance value for each of the divisions of the form totaled by the tabulation process. A form identification program characterized in that it is executed.

Images