How to identify computer and document types

Abstract

To efficiently and highly accurately identify a type of a document using an image feature and language feature.SOLUTION: A computer stores document definition information for managing a document definition including image features and language features in a plurality of document types, acquires an image of a target document, generates a processed image by executing image processing on the image of the target document, acquires the image feature from the processed image, calculates the first confidence degree indicating the similarity between the image feature acquired from the processed image and the image feature of the document type for each of the plurality of document types, selects a similar document type on the basis of the first confidence degree, acquires the language feature from the processed image, calculates the second confidence degree indicating the similarity between the language feature acquired from the processed image and the language feature of the similar document type, and selects a candidate document type from the similar document type on the basis of the first confidence degree and second confidence degree.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present invention relates to an apparatus and method for identifying the type of an inputted form. [Background technology]

[0002] There is a known method of automating business processes by reading and checking attributes from documents using Optical Character Recognition (OCR) technology. Conventionally, to obtain necessary attributes from documents using OCR technology, form definitions that define the form format for each form type are used. The form definition defines the position and attributes of the character string to be read.

[0003] In recent years, there has been an increasing need to use OCR for work involving a wide variety of forms. In such cases, before using OCR, it is necessary to identify the type of input form and specify the form definition to be used. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2002-245403 A Summary of the Invention [Problem to be solved by the invention]

[0005] Known methods for identifying the type of form include a method that uses the image features of the form and a method that uses the language features of the form. The image feature method identifies the type of form by using image features such as the size, color tone, and layout of the form image. The language feature method identifies the type of form by using character strings included in the form.

[0006] Here, the problems with the image feature method and the language feature method will be described. Figures 10 and 11 are diagrams for explaining the problems with the conventional techniques.

[0007] The two forms shown in Fig. 10 have similar layouts and cannot be distinguished when image features are used. The two forms shown in Fig. 11 have many common character strings, so there is a possibility that erroneous identification results will be output when language features are used. The technology described in Patent Document 1 is known to address such issues.

[0008] Patent document 1 states that "The form dictionary stores information on corresponding points, which are points that represent the characteristics of registered forms, and information on character portions as features different from corresponding points. Corresponding points between the form to be processed and the registered forms are detected (step S9). The corresponding point information for these corresponding points is referenced from the form dictionary to calculate the degree of dissimilarity between the two forms (step S11). If there are not multiple registered forms whose degree of dissimilarity is close enough to a predetermined degree (N in step S9), the type of the form to be processed is identified based on the magnitude of the dissimilarity (step S13). If there are multiple registered forms (Y in step S9), the type of form is identified by reference to the features of the registered form that are different from the corresponding points from the form dictionary (steps S22 to S12)."

[0009] In Patent Document 1, the results of document type identification using image features and language features are independent, and the problems associated with the two identification methods described above cannot be avoided.

[0010] An object of the present invention is to provide an apparatus and method for identifying the type of a document efficiently and with high accuracy by using image features and language features. [Means for solving the problem]

[0011] A representative example of the invention disclosed in the present application is as follows: That is, a computer including an arithmetic unit, a storage device connected to the arithmetic unit, and an interface connected to the arithmetic unit, the storage device stores form definition information for managing form definitions including image features and language features of a plurality of form types, the arithmetic unit acquires an image of a target form via the interface, generates a processed image by executing image processing on the image of the target form, acquires image features from the processed image, calculates, for each of the plurality of form types, a first certainty factor indicating a similarity between the image features acquired from the processed image and the image features of the form type, selects a similar form type based on the first certainty factor, acquires language features from the processed image, calculates a second certainty factor indicating a similarity between the language features acquired from the processed image and the language features of the similar form type, selects a candidate form type from the similar form types based on the first certainty factor and the second certainty factor, and presents information about the candidate form type. the image processing is an image rotation process, and the arithmetic device generates a plurality of the processed images by performing rotation processes at different rotation angles on the image of the target form, acquires image features of each of the plurality of processed images, calculates the first certainty factor, generates data in which the similar form type, the rotation angle, and the first certainty factor are associated with each other, acquires language features from the processed image generated by the rotation process at the rotation angle corresponding to the data, and calculates the second certainty factor. . Effect of the Invention

[0012] According to the present invention, a computer can efficiently and accurately identify the type of a document by using image features and language features. Other problems, configurations, and effects will become apparent from the following description of the embodiments. [Brief description of the drawings]

[0013] [Figure 1] FIG. 1 illustrates an example of a system configuration according to a first embodiment. [Diagram 2] FIG. 4 is a diagram illustrating an example of a data structure of form definition information according to the first embodiment. [Diagram 3] FIG. 2 is a diagram for explaining an outline of a process executed by a computer according to the first embodiment. [Figure 4] 11 is a flowchart illustrating an example of a document type discrimination process using image features executed by a computer according to the first embodiment. [Figure 5A] FIG. 13 is a diagram illustrating an example of a data structure of intermediate information generated by the computer of the first embodiment. [Figure 5B] FIG. 13 is a diagram illustrating an example of a data structure of similar document type information generated by the computer according to the first embodiment. [Figure 6] 11 is a flowchart illustrating an example of a document type discrimination process using language features executed by the computer according to the first embodiment. [Figure 7] FIG. 11 is a diagram showing an example of a data structure of certainty information generated by the computer of the first embodiment. [Figure 8] 11 is a flowchart illustrating an example of a candidate form type selection process executed by the computer according to the first embodiment. [Figure 9] FIG. 11 is a diagram showing an example of a screen presented to a user by the computer of the first embodiment. [Figure 10] FIG. 1 is a diagram for explaining a problem with the conventional technology. [Figure 11] FIG. 1 is a diagram for explaining a problem with the conventional technology. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. However, the present invention is not to be interpreted as being limited to the description of the embodiment shown below. It is easily understood by those skilled in the art that the specific configuration can be changed without departing from the idea or purpose of the present invention. In the configuration of the invention described below, the same or similar configurations or functions are given the same reference numerals, and duplicated explanations are omitted. The notations "first", "second", "third", etc. in this specification are given to identify components, and do not necessarily limit the number or order. The position, size, shape, range, etc. of each configuration shown in the drawings, etc. may not represent the actual position, size, shape, range, etc. in order to facilitate understanding of the invention. Therefore, the present invention is not limited to the position, size, shape, range, etc. disclosed in the drawings, etc. EXAMPLES

[0015] FIG. 1 is a diagram illustrating an example of a system configuration according to the first embodiment.

[0016] The system includes a computer 10 that identifies the type of a form, and a scanner device 20 that acquires an image of the form. The computer 10 and the scanner device 20 are connected to each other via a network 40 such as a LAN (Local Area Network).

[0017] A user inputs a form 30 into the scanner device 20. The scanner device 20 generates an image of the form 30 (form image 124) and transmits it to the computer 10. The computer 10 uses the form image 124 to identify the form type of the form 30 and presents information about the form type of the form 30 to the user.

[0018] The computer 10 includes a calculation unit 101, a storage device 102, a communication interface 103, an input / output interface 104, an input device 105, a display device 106, and an external storage device 107. The calculation unit 101, the storage device 102, the communication interface 103, and the input / output interface 104 are connected to each other via an internal bus. The input device 105, the display device 106, and the external storage device 107 are connected to the input / output interface 104.

[0019] The arithmetic device 101 is a processor, a GPU, or the like, and executes a program stored in the storage device 102. The arithmetic device 101 executes processing according to the program, thereby operating as a functional unit (module) having a specific function. In the following explanation, when processing or the like is explained with a functional unit as the subject, it indicates that the arithmetic device 101 is executing a program that realizes the functional unit. The functional units realized by the arithmetic device 101 will be described later.

[0020] The storage device 102 is a memory or the like, and stores the programs executed by the arithmetic device 101 and information used by the programs. The storage device 102 is also used as a work area.

[0021] The communication interface 103 communicates with an external device. The input / output interface 104 connects to the external device. The input device 105 is a keyboard, a mouse, a touch panel, or the like. The display device 106 is a display, or the like. The external storage device 107 is, for example, a HDD (Hard Disk Drive), or the like.

[0022] The storage device 102 stores a form recognition program 120, a form management program 121, setting information 122, and form definition information 123. The storage device 102 also stores a form image 124 received from the scanner device 20, and stores a recognition result 125 of the form image 124.

[0023] The setting information 122 is information for managing the processing contents of the functional units, the threshold values ​​used in the processing, the information to be output, etc. The form definition information 123 is information for managing the form definition. The data structure of the form definition information 123 will be described with reference to FIG. 2.

[0024] The arithmetic device 101 executes the form recognition program 120 to function as a form image input unit 110, a form recognition processing unit 111, and a recognition result output unit 112. The form image input unit 110 accepts input of a form image 124 and stores it in the storage device 102. The form recognition processing unit 111 recognizes the form type of the form 30 corresponding to the form image 124. The recognition result output unit 112 generates display information for displaying the processing result of the form recognition processing unit 111.

[0025] The calculation device 101 executes a form management program 121 to function as a data registration unit 113, a data management unit 114, a screen display unit 115, and a setting unit 116. The data registration unit 113 accepts information related to form definitions and stores it in form definition information 123. The data management unit 114 manages setting information 122 and form definition information 123. The screen display unit 115 displays a screen on the display device 106. The setting unit 116 accepts various setting information and stores it in the setting information 122.

[0026] Regarding each functional unit of the computer 10, multiple functional units may be integrated into one functional unit, or one functional unit may be divided into multiple functional units for each function.

[0027] The programs and information stored in the storage device 102 may be stored in the external storage device 107. In this case, the arithmetic device 101 reads out the programs and information from the external storage device 107 and loads them into the storage device 102.

[0028] 1 is merely an example, and is not limited to this. There may be two or more scanner devices 20. Also, a computer system made up of multiple computers may have the same functions as the computer 10.

[0029] Note that the calculator 10 may have an OCR function, or may be connected to a device having an OCR function.

[0030] FIG. 2 is a diagram illustrating an example of a data structure of the form definition information 123 according to the first embodiment.

[0031] The form definition information 123 stores a form definition 200 for each form type. A form ID and a form type name are assigned to the form definition 200 as identification information for the form type. The form definition 200 also includes an image feature dictionary 201, a language feature dictionary 202, and attribute information 203.

[0032] The image feature dictionary 201 is a dictionary that stores image features of form types corresponding to the form definition 200. In the present invention, hash values ​​such as aHash, pHash, dHash, and wHash are used as image features. Note that the present invention is not limited to the types and calculation methods of image features.

[0033] The language feature dictionary 202 is a dictionary that stores language features of form types corresponding to the form definition 200. The language feature dictionary 202 stores one or more pieces of keyword data each of which has an attribute (character string), an attribute type, and the position of an attribute in a form as one piece of data.

[0034] The attribute information 203 stores information about items to be read from the form. The attribute information 203 stores one or more item data consisting of an attribute type and an attribute reading position in the form. The attribute information 203 of the form definition 200 with a form ID of "1" includes item data of "item 1" and "item 2".

[0035] FIG. 3 is a diagram for explaining an outline of the process executed by the computer 10 of the first embodiment.

[0036] The form image input unit 110 receives input of a form image 124 generated from a form 30 (step S100).

[0037] The form recognition processing unit 111 executes a form type discrimination process using image features (step S200). In this process, the form recognition processing unit 111 acquires image features of the form 30 (step S210), calculates a confidence level indicating a similarity with the image features of each form type (step S220), and selects a similar form type based on the confidence level (step S230).

[0038] The document recognition processing unit 111 executes a document type discrimination process using the language features of similar document types (step S300). In this process, the document recognition processing unit 111 extracts a character string (language feature) from a position specified by the keyword data (step S310), calculates a certainty indicating the similarity with the language features of the similar document type (step S320), and outputs certainty information 700 (see FIG. 7) (step S330).

[0039] The form recognition processing unit 111 selects a candidate form type from among the similar form types based on the confidence information 700 (step S400).

[0040] Fig. 4 is a flowchart for explaining an example of a form type discrimination process using image features executed by the computer 10 of the first embodiment. Fig. 5A is a diagram showing an example of a data structure of intermediate information generated by the computer 10 of the first embodiment. Fig. 5B is a diagram showing an example of a data structure of similar form type information generated by the computer 10 of the first embodiment.

[0041] The form recognition processing unit 111 acquires the image feature dictionary 201 for all form types from the form definition information 123 (step S211).

[0042] The form recognition processing unit 111 acquires the form image 124 from the storage device 102 (step S212).

[0043] The form recognition processing unit 111 executes image processing on the form image 124 (step S213).

[0044] Specifically, the form recognition processing unit 111 executes a rotation process to rotate the form image 124 by any angle. In this embodiment, four rotation processes of 0 degrees, 90 degrees, 180 degrees, and 270 degrees are executed. As a result, four images are generated from one form image 124. In the following description, an image generated by the rotation process is referred to as a processed image.

[0045] The form recognition processing unit 111 acquires image features from each processed image (step S214). For example, the form definition information 123 calculates pHash from the processed image and stores it as an image feature in the storage device 102. The method of calculating pHash is a known technique, so a detailed description will be omitted.

[0046] The form recognition processing unit 111 starts loop processing of the form type (step S221). Specifically, the form recognition processing unit 111 selects one form type from the form types. At this time, the form recognition processing unit 111 adds an entry to the intermediate information 500.

[0047] An entry includes an ID 501, a form type 502, and a confidence level 503. One entry is generated for one form type. The ID 501 is a field that stores the form ID included in the form definition 200 that corresponds to the form type. The form type 502 is a field that stores the form type name. The confidence level 503 is a group of fields that stores confidence levels that indicate the similarity between the image features of the form 30 and the image features of the form type. The confidence level 503 includes the same number of fields as the processed images.

[0048] The form recognition processing unit 111 sets values ​​for the added entry's ID 501 and form type 502. At this point, each field of the confidence level 503 is blank.

[0049] The form recognition processing unit 111 starts loop processing of the processed image (step S222). Specifically, the form recognition processing unit 111 selects one processed image from among a plurality of processed images.

[0050] The form recognition processing unit 111 calculates a confidence level indicating the similarity between the image feature of the selected form type and the image feature of the selected processed image (step S223). If the image feature is pHash, the form recognition processing unit 111 calculates the similarity between the pHashes as the confidence level. The form recognition processing unit 111 searches for an entry corresponding to the form image 124 in the intermediate information 500, and sets the confidence level in the field corresponding to the processed image in the confidence level 503 of the entry.

[0051] The form recognition processing unit 111 determines whether or not the processing has been completed for all the processing images (step S224).

[0052] If the processing has not been completed for all the processed images, the form recognition processing unit 111 continues the loop processing of the processed images and returns to step S222.

[0053] When the processing has been completed for all the processed images, the form recognition processing unit 111 ends the loop processing of the processed images, and determines whether the processing has been completed for all the form types (step S225).

[0054] If the process has not been completed for all the document types, the document recognition processing unit 111 continues the loop process for the document types and returns to step S221.

[0055] When the process is completed for all the document types, the document recognition processing unit 111 ends the loop process of the document types and selects a similar document type based on the confidence factor (step S231). After that, the document recognition processing unit 111 ends the document type discrimination process.

[0056] For example, the form recognition processing unit 111 selects a form type whose certainty is greater than a threshold value as a similar form type. The threshold value is included in the setting information 122.

[0057] At this time, the form recognition processing unit 111 generates similar form type information 510. The similar form type information 510 stores entries including ID 511, form type 512, confidence level 513, and angle 514. One entry is identified by a combination of form type and processed image. ID 511 and form type 512 are the same fields as ID 501 and form type 502. Confidence level 513 is a field that stores the confidence level stored in any field of confidence level 503. Angle 514 is a field that stores identification information for identifying the processed image. In this embodiment, the processed image is generated by performing a rotation process, so the rotation angle is used as the identification information of the processed image.

[0058] When there is a form type including two or more fields whose value of the confidence 503 is greater than the threshold value, the form recognition processing unit 111 may register only the entry in which the angle of the processed image having the largest value is set.

[0059] In this way, the computer 10 of this embodiment determines the similarity of image features while taking into account the orientation of the form 30. This improves the accuracy of form type recognition, and further improves the accuracy of subsequent OCR reading.

[0060] Fig. 6 is a flowchart for explaining an example of a document type discrimination process using language features executed by the computer 10 of the embodiment 1. Fig. 7 is a diagram showing an example of a data structure of certainty information generated by the computer 10 of the embodiment 1.

[0061] The form recognition processing unit 111 refers to the form definition information 123 based on the ID 511 of the similar form type information 510, and acquires the language feature dictionary 202 of all similar form types (step S311).

[0062] The form recognition processing unit 111 acquires the form image 124 (step S312), and executes binarization processing on the form image 124 (step S313).

[0063] The form recognition processing unit 111 executes image processing on the binarized form image 124 (step S314), and extracts character strings (language features) from each of the multiple processed images (step S315).

[0064] The form recognition processing unit 111 starts a loop process of similar form types (step S321). Specifically, the form recognition processing unit 111 selects one similar form type from the similar form types registered in the similar form type information 510. At this time, the form recognition processing unit 111 adds an entry to the certainty factor information 700.

[0065] The entry includes ID 701, form type 702, angle 703, confidence (image feature) 704, and confidence (language feature) 705. One entry is generated for each combination of form type and angle. ID 701, form type 702, and angle 703 are the same fields as ID 501, form type 502, and angle 514. Confidence (image feature) 704 is the same field as confidence 513. Confidence (language feature) 705 is a field that stores the confidence between the language feature of form 30 and the language feature of the form type.

[0066] The form recognition processing unit 111 refers to the similar form type information 510 and sets the values ​​of the entry of the selected similar form type to the added entry's ID 701, form type 702, angle 703, and confidence level (image feature) 704. At this point, confidence level (language feature) 705 is blank.

[0067] The form recognition processing unit 111 calculates a degree of certainty indicating the similarity between the language features of the selected similar form type and the language features of the processed image associated with the form type (step S322).

[0068] Specifically, the form recognition processing unit 111 refers to the similar form type information 510 and acquires a character string extracted from the processed image corresponding to the angle 514 of the entry of the selected similar form type. The form recognition processing unit 111 calculates the degree of match between the character string of the processed image acquired from the position of the keyword included in the language feature dictionary 202 and the keyword included in the language feature dictionary 202. The form recognition processing unit 111 sets the sum of the calculated degrees of match to the confidence (language feature) 705 of the entry added in step S321.

[0069] Note that the number of keywords registered in the language feature dictionary 202 for each form type may differ. In this case, since the confidence level depends on the number of keywords, it is necessary to correct the confidence level according to the number of keywords. In this embodiment, a weight is set for each keyword in advance, and the sum of the keyword match rate and the weight is treated as the confidence level. The weights of keywords that are considered important in identifying the form type are set to be large. The weights are also adjusted so that the sum of the weights is 1. Information regarding the weights of keywords may be included in the setting information 122 or in the language feature dictionary 202.

[0070] The form recognition processing unit 111 determines whether or not the processing has been completed for all similar form types (step S323).

[0071] If the process has not been completed for all similar document types, the document recognition processing unit 111 continues the loop process for similar document types and returns to step S321.

[0072] When the processing has been completed for all similar document types, the document recognition processing unit 111 ends the loop processing for similar document types, and also ends the document type discrimination processing.

[0073] FIG. 8 is a flowchart illustrating an example of a candidate form type selection process executed by the computer 10 according to the first embodiment.

[0074] The form recognition processing unit 111 refers to the certainty information 700 and selects a candidate form type based on the certainty of the image feature and the certainty of the language feature (step S401).

[0075] For example, the form recognition processing unit 111 selects, as a candidate form type, a similar form type whose image feature has a certainty greater than a first threshold and whose language feature has a certainty greater than a second threshold. The first and second thresholds are included in the setting information 122.

[0076] There may be cases where entries with the same form type but different rotation angles exist in the certainty information 700. When multiple entries with different rotation angles and the same form type are selected, the form recognition processing unit 111 consolidates them into one recognition result.

[0077] The form recognition processing unit 111 determines whether or not there is a candidate form type that satisfies the selection condition in step S410 (step S402).

[0078] If there is a candidate form type that satisfies the selection condition in step S410, the form recognition processing unit 111 outputs information about the candidate form type (step S404), and then ends the candidate form type selection process.

[0079] If there is no candidate document type that satisfies the selection conditions of step S410, the document recognition processing unit 111 selects a candidate document type based on the confidence level of the image features (step S403), outputs a recognition result 125 including information about the candidate document type (step S404), and then terminates the candidate document type selection process.

[0080] For example, the form recognition processing unit 111 selects, as a candidate form type, a similar form type whose image feature has a certainty degree greater than a third threshold. The third threshold is included in the setting information 122. Note that the third threshold is smaller than the first threshold.

[0081] If there is no candidate form type that satisfies the selection condition in step S403, the form recognition processing unit 111 outputs information to the effect that there is no candidate form type.

[0082] The recognition result output unit 112 generates display information for displaying a screen based on the recognition result 125 output from the form recognition processing unit 111, and the screen display unit 115 presents the screen to the user based on the display information. Fig. 9 is a diagram showing an example of a screen presented to the user by the computer 10 of the first embodiment.

[0083] The screen 900 includes display fields 901 , 902 , 903 , an OK button 904 , a CANCEL button 905 , and a registration button 906 .

[0084] The display field 901 is a field that displays the form image 124 of the input form 30. The display field 902 is a field that displays information on the form type selected by the form recognition processing unit 111. For example, an image of a representative form defined by the form definition 200 of the candidate form type is displayed. The display field 903 is a field that displays details of the candidate form type selected by the user from the display field 902. The display field 903 displays an image of the representative form, image features of the form image 124, image features of the candidate form type, language features of the candidate form type, etc.

[0085] The user selects a candidate form type from display field 902 and checks the details in display field 903. If the selected candidate form type is adopted, the user presses an OK button 904. If a different candidate form type is to be confirmed, the user presses a CANCEL button 905. If a form type different from the candidate form type is to be registered, the user presses a register button 906. Registering a form type is a well-known technique, so details will be omitted.

[0086] In this embodiment, the threshold value for the image characteristics is described as being set in advance in the setting information 122, but is not limited to this. For example, the threshold value may be set based on the variation in image characteristics of the form type registered in the form definition information 123.

[0087] According to the present invention, the computer 10 narrows down the form types using image features, and then selects the candidate form types based on language features. Thus, the problems of the two classification methods are avoided, and the form types can be classified efficiently and with high accuracy.

[0088] When image features are used, there is a possibility that a form type such as that shown in Fig. 10 may be selected. However, in the present invention, further narrowing down is performed based on the language features of the form type. This makes it possible to identify the form type with high accuracy. Narrowing down using image features requires less processing load than narrowing down using language features, so it is possible to identify the form type with high accuracy while suppressing the processing load.

[0089] In addition, since the form types are narrowed down using language features after the form types are narrowed down using image features, it is possible to reduce the number of keywords to be registered in the language feature dictionary 202. This makes it possible to reduce the cost required for setting the form definition 200. In addition, the load of the calculation process of the certainty of language features can also be reduced.

[0090] The present invention is not limited to the above-mentioned embodiment, but includes various modified examples. For example, the above-mentioned embodiment describes the configuration in detail to easily explain the present invention, and the present invention is not necessarily limited to the configuration including all the described configurations. Also, it is possible to add, delete, or replace a part of the configuration of each embodiment with another configuration.

[0091] In addition, each of the above configurations, functions, processing units, processing means, etc. may be realized in part or in whole by hardware, for example, by designing them as integrated circuits. The present invention can also be realized by software program code that realizes the functions of the embodiments. In this case, a storage medium on which the program code is recorded is provided to a computer, and a processor included in the computer reads the program code stored in the storage medium. In this case, the program code itself read from the storage medium realizes the functions of the above-mentioned embodiments, and the program code itself and the storage medium storing it constitute the present invention. Examples of storage media for supplying such program code include flexible disks, CD-ROMs, DVD-ROMs, hard disks, SSDs (Solid State Drives), optical disks, magneto-optical disks, CD-Rs, magnetic tapes, non-volatile memory cards, ROMs, etc.

[0092] Furthermore, the program code for realizing the functions described in this embodiment can be implemented in a wide range of program or script languages, such as assembler, C / C++, perl, Shell, PHP, Python, Java (registered trademark), etc.

[0093] Furthermore, the program code of the software that realizes the functions of the embodiments may be distributed over a network and stored in a storage means such as a computer's hard disk or memory, or in a storage medium such as a CD-RW or CD-R, and the processor of the computer may read out and execute the program code stored in the storage means or storage medium.

[0094] In the above-mentioned embodiment, the control lines and information lines are shown as those considered necessary for the explanation, and not all the control lines and information lines are shown in the product. All the components may be connected to each other. [Explanation of symbols]

[0095] 10 Calculator 20 Scanner device 30 Reports 40 Network 101 Arithmetic equipment 102 Storage device 103 Communication Interface 104 Input / Output Interface 105 Input Device 106 Display device 107 External storage device 110 Form image input unit 111 Form recognition processing unit 112 Recognition result output unit 113 Data Registration Department 114 Data Management Department 115 Screen display section 116 Setting section 120 Document Recognition Program 121 Document Management Program 122 Setting information 123 Form definition information 124 Form images 125 Recognition results 200 Report Definition 201 Image Feature Dictionary 202 Language Feature Dictionary 203 Attribute information 500 Intermediate Information 510 Similar document type information 700 Confidence Information 900 screens

Claims

1. A computer including a computing device, a storage device connected to the computing device, and an interface connected to the computing device, the storage device stores form definition information for managing form definitions including image features and language features of a plurality of form types; The computing device includes: Acquiring an image of a target form via the interface; generating a processed image by performing image processing on the image of the target form; obtaining image features from the processed image; calculating a first confidence level indicating a similarity between an image feature acquired from the processed image and an image feature of the form type for each of the plurality of form types; Selecting a similar document type based on the first certainty; obtaining language features from the processed images; Calculating a second confidence level indicating a similarity between the language features acquired from the processed image and the language features of the similar document type; selecting a candidate form type from among the similar form types based on the first certainty degree and the second certainty degree; Present information regarding the candidate document type; The image processing is an image rotation process, The computing device includes: generating a plurality of the processed images by performing rotation processing at different rotation angles on the image of the target form; acquiring image features of each of the plurality of processed images and calculating the first confidence level; generating data in which the similar document type, the rotation angle, and the first certainty factor are associated with each other; A computer that acquires language features from the processed image generated by a rotation process of a rotation angle corresponding to the data, and calculates the second confidence level.

2. A computer according to claim 1, The computing device is configured to select, as the candidate form type, the similar form type for which the first certainty degree is greater than a first threshold value and the second certainty degree is greater than a second threshold value.

3. The computer according to claim 2, The calculation device is a computer characterized in that when either the first certainty degree is less than or equal to a first threshold value and the second certainty degree is less than or equal to the second threshold value, the calculation device selects the similar document type for which the first certainty degree is greater than a third threshold value as the candidate document type.

4. A computer according to claim 1, The language feature of the form type is a keyword, the storage device stores correction information for correcting the second certainty factor in accordance with the number of the keywords defined as a language feature for each of the plurality of document types; The computing device includes: Extracting keywords from the processed image; Calculating a similarity between the extracted keywords and keywords defined as language features of the similar document type; A computer that calculates the second certainty factor using the similarity factor of the extracted keywords and the correction information.

5. A method for identifying a document type executed by a computer, comprising: the computer includes a processor, a storage device connected to the processor, and an interface connected to the processor; the storage device stores form definition information for managing form definitions including image features and language features of a plurality of form types; The method of determining the type of document is as follows: A first step in which the computing device acquires an image of a target document via the interface; a second step of the computing device performing image processing on the image of the target form to generate a processed image; a third step of the computing device acquiring image features from the processed image; a fourth step in which the arithmetic device calculates, for each of the plurality of form types, a first confidence level indicating a similarity between an image feature acquired from the processed image and an image feature of the form type; a fifth step of the calculation device selecting a similar document type based on the first certainty; a sixth step of the computing device acquiring language features from the processed image; a seventh step of calculating a second certainty factor indicating a similarity between a language feature acquired from the processed image and a language feature of the similar document type; an eighth step of the calculation device selecting a candidate form type from among the similar form types based on the first certainty degree and the second certainty degree; a ninth step in which the arithmetic device presents information about the candidate document type; The image processing is an image rotation process, The second step includes a step of generating a plurality of the processed images by executing a rotation process with different rotation angles on the image of the target form by the calculation device, the third step includes a step of the computing device acquiring image features of each of the plurality of processed images; the fourth step includes a step of the calculation device calculating the first confidence factor for each of the plurality of processed images; the fifth step includes a step of generating data in which the calculation device associates the similar document type, the rotation angle, and the first certainty degree, The sixth step includes a step of the calculation device acquiring language features from the processed image generated by a rotation process of a rotation angle corresponding to the data, The document type identification method, wherein the seventh step includes a step in which the arithmetic device calculates the second certainty factor for each of the data.

6. A method for identifying a document type as described in claim 5, comprising: The eighth step of the method for identifying a document type is characterized in that it includes a step in which the calculation device selects, as the candidate document type, the similar document type whose first certainty degree is greater than a first threshold value and whose second certainty degree is greater than a second threshold value.

7. A method for identifying a document type as described in claim 6, comprising: The eighth step of the method for identifying a document type is characterized in that it includes a step in which the calculation device selects, when either the first certainty degree is equal to or less than a first threshold value and the second certainty degree is equal to or less than the second threshold value, the similar document type for which the first certainty degree is greater than a third threshold value as the candidate document type.

8. A method for identifying a document type as described in claim 5, comprising: The language feature of the form type is a keyword, the storage device stores correction information for correcting the second certainty factor in accordance with the number of the keywords defined as a language feature for each of the plurality of document types; the sixth step includes a step of the computing device extracting keywords from the processed image; The seventh step includes: a step of calculating a similarity between the extracted keyword and a keyword defined as a language feature of the similar form type by the calculation device; a step of the calculation device calculating the second certainty factor using the similarity factor of the extracted keywords and the correction information.

Images