Character recognition device

Abstract

To provide a character recognition device that accurately acquires a recognition target item entered within a predetermined range of a form.SOLUTION: Information acquisition units 22, 31 of a character recognition device 10 perform image recognition on the entire images of a reference form and a processing target form to acquire box coordinates and character strings. A first guide setting unit 23 sets a box of the reference form in which characters are recognized in the box as a reference guide box, and a first range setting unit 25 sets an entry range of a recognition target item of the reference form as a reference entry range. A second guide setting unit 32 sets, as a processing target guide box, a box in which a character string the same as the character string of the reference guide box is described and in which a ratio between box aspect ratios is within a predetermined range as compared with the reference guide box, from among boxes of the processing target form in which the characters are recognized in the boxes. A second range setting unit 35 sets a range of the processing target form corresponding to the reference entry range as a processing target entry range based on a relative relationship between both the guide boxes. An item acquisition unit 36 performs character recognition on the processing target entry range.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a character recognition device that acquires recognition target items that are filled in within a predetermined range on a form that is made up of frame lines and characters. [Background technology]

[0002] In recent years, a technology has been utilized that converts character strings written on forms into text by character recognition. In this technology, the character string portion of the image of the form to be processed is identified, and character recognition is performed on the image obtained by extracting the character string portion. Patent Document 1 discloses a technology that recognizes an intersection area between a first group rectangle, which is a collection of frames arranged in a first direction, and a second group rectangle, which is a collection of frames arranged in a second direction, as an entry box from an image of the form, and extracts an image within the entry box to perform character recognition. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2023-36833 A Summary of the Invention [Problem to be solved by the invention]

[0004] However, when extracting character strings, if it is not possible to extract only the relevant character strings without mixing in frame information other than the character strings or character strings outside the required range as noise, this can have a significant impact on character recognition accuracy, which is an issue.

[0005] Furthermore, even if the form is in a standard format, each item of the imaged form to be processed is not necessarily located in the same position as the original form. This is because many of the forms to be processed are imaged using a scanner or the like, and the margin settings and scale are often different from the original form during the imaging process, and the imaged form to be processed is itself a replica created by copying the original form, etc. For this reason, it is not practical to automatically extract the parts of the image of the form to be processed that are located at the same coordinate positions as the original form.

[0006] Also, as disclosed in Patent Document 1, for example, a process for recognizing frame lines of an image of a target form may be performed, and an attempt may be made to identify the written portion of the target item based on the frame. However, it is not possible to accurately determine which frame corresponds to the written portion of the target item. Furthermore, due to the processing characteristics of frame recognition and character recognition, the same frame and the same character string may not always be recognized for the target form, and some frame lines may not be recognized or characters may be misrecognized depending on the imaging state. Therefore, a frame corresponding to the written portion of the target item may not always be detected.

[0007] An object of the present invention is to provide a character recognition device that can accurately acquire items to be recognized that are entered within a predetermined area of ​​a form. [Means for solving the problem]

[0008] The character recognition device according to the present invention includes a first information acquisition unit, a first guide setting unit, a first range setting unit, a second information acquisition unit, a second guide setting unit, a second range setting unit, and an item acquisition unit. Here, the original document in which the entry box for entering any character is left blank is defined as the reference document, and the document in which any character is entered in the entry box is defined as the document to be processed, and the ratio of the width and height of the box is defined as the box aspect ratio.

[0009] The first information acquisition unit performs image recognition to recognize frame lines and characters over the entire image of the reference form, and acquires frame coordinates and character strings. The first guide setting unit sets a frame within which characters are recognized in the image recognition of the reference form as a reference guide frame. The first range setting unit sets a range for filling in the items to be recognized on the reference form as a reference entry range.

[0010] The second information acquisition unit performs image recognition on the entire image of the target form to acquire frame coordinates and character strings within the frame. If there is only one frame among the frames in which characters are recognized in the image recognition of the target form, in which the same character string as the character string of the reference guide frame is written and whose frame aspect ratio is within a predetermined range compared to the reference guide frame, the second guide setting unit sets the frame as the target guide frame. The second range setting unit specifies the range of the target form corresponding to the reference entry range based on the relative relationship between the reference guide frame and the target guide frame, and sets the range as the target entry range. The item acquisition unit performs character recognition on the portion of the target entry range in the image of the target form to acquire the items to be recognized.

[0011] In this way, in the present invention, frame recognition and character recognition are performed on the entire images of both the reference form and the form to be processed. This makes it possible to accurately grasp the relative relationship between the two forms by using a guide frame that associates information on character strings and frame coordinates as a set, rather than just the positional relationship based on frame coordinate information. In addition, by not treating as a guide frame any frame whose information does not satisfy certain conditions when setting the guide frame for the form to be processed, it is possible to suppress a decrease in item acquisition accuracy caused by some frame lines not being recognized or characters being erroneously recognized. [Brief description of the drawings]

[0012] [Figure 1] 1 is a functional block diagram of a character recognition device according to an embodiment. [Diagram 2] FIG. 13 is a diagram illustrating an example of an image of a reference form according to an embodiment. [Diagram 3]11 is a diagram illustrating an example of an image of a form to be processed according to an embodiment; [Figure 4] 3 is a diagram illustrating an example of information acquired from the reference form in FIG. 2 by image recognition. [Diagram 5] 11 is a diagram illustrating information of a reference guide frame according to an embodiment; [Figure 6] 11 is a diagram illustrating an example of information on a reference guide candidate frame according to an embodiment; [Figure 7] 3 is a diagram in which a portion of the reference form in FIG. 2 that is set as a reference entry range is shaded. FIG. [Figure 8] FIG. 13 is a diagram illustrating information of a reference entry range according to an embodiment. [Figure 9] 4 is a diagram illustrating an example of information acquired by image recognition from the form to be processed in FIG. 3. [Figure 10] 6 is a diagram in which a determination result when a processing target guide frame is set is added to the information in FIG. 5. [Figure 11] 11 is a diagram illustrating an example of information of a processing target guide frame according to an embodiment; [Figure 12] FIG. 6 is a diagram illustrating an example of information of the reference guide frame after it has been changed from the state shown in FIG. 5 in response to a determination result at the time of setting the processing target guide frame. [Figure 13] FIG. 7 is a diagram illustrating an example of information of the reference guide candidate frame after it has been changed from the state shown in FIG. 6 in response to a determination result at the time of setting the processing target guide frame. [Figure 14] This figure shows the information of the reference guide candidate frame in Figure 13, plus information on the reference guide frame that is closest to the reference guide candidate frame, and information on a "theoretical frame" that corresponds to the reference guide candidate frame on the document to be processed. [Figure 15] 15 is a diagram showing a schematic diagram of the positional relationship between the "theoretical frame" and the "actual frame" on the processing target form of FIG. 14. [Figure 16] 12 is a diagram illustrating an example of information of the processing target guide frame after it has been changed from the state shown in FIG. 11 in response to a promotion determination result by a promotion determination unit. [Figure 17]13 is a diagram illustrating an example of information of the reference guide frame after it has been changed from the state shown in FIG. 12 in response to a promotion determination result by a promotion determination unit. [Figure 18] 17A and 17B are diagrams illustrating an example of the frame coordinates of the processing target guide frame converted into the size of the reference form from the state shown in FIG. 16, the distance between the center point and the reference guide frame, the median value of the distance, and the exclusion determination result. [Figure 19] 17 is a diagram illustrating an example of information of the processing target guide frame after it has been changed from the state shown in FIG. 16 in response to a result of exclusion determination by an exclusion determination unit. [Figure 20] 18 is a diagram illustrating an example of information of the reference guide frame after it has been changed from the state shown in FIG. 17 in response to a result of exclusion determination by an exclusion determination section. FIG. [Figure 21] This figure shows the information of the specific pattern and the reference guide frame in addition to the information of FIG. 8. [Figure 22] FIG. 13 is a diagram illustrating a case where the reference entry range does not belong to a frame. [Figure 23] 13 is a diagram illustrating an example of information on a processing target entry range according to an embodiment; FIG. [Figure 24] 1 is a flowchart showing a procedure of a process executed by a character recognition device according to an embodiment to handle information of a standard form. [Diagram 25] 1 is a flowchart showing a procedure of a process executed by a character recognition device according to an embodiment to handle information of a form to be processed. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] A character recognition device according to an embodiment of the present invention will be described below with reference to the drawings. The character recognition device 10 shown in Fig. 1 is a device for acquiring items filled in a predetermined area of ​​a standard form consisting of a frame line and characters by character recognition. The above items are defined as "recognition target items".

[0014] As shown in FIG. 2, the frames that make up a form include a heading frame and an entry frame. A heading frame is a frame in which headings and notes, such as "Name" and "Date of Birth", are written in advance. An entry frame is a frame that has a blank space within the frame in which any character can be written. For example, the contents of the heading frame "Name" are to be written in the blank entry box to the right of it. The contents of the heading frame "Date of Birth" are to be written in the blank space of the heading frame "Date of Birth". In this way, the heading frame and the entry frame in which the contents are written may be separate frames or they may be the same frame.

[0015] Here, an original form with blank entry boxes as shown in Fig. 2 is defined as a "reference form." A form with any characters entered in the entry boxes as shown in Fig. 3 is defined as a "form to be processed." An image 81 of a reference form 80 shown in Fig. 2 and an image 91 of a form to be processed 90 shown in Fig. 3 are images of each form captured by the scanner 15 (see Fig. 1). The frame of the form to be processed 90 in Fig. 3 is larger in size and shifted to the right compared to the frame of the reference form 80 in Fig. 2.

[0016] Character recognition device 10 is configured to include a computer equipped with a CPU, RAM, ROM, an input / output interface, etc. (not shown), and includes various functional units described below. The functions of the various functional units are realized by software processing in which the CPU executes a character recognition program previously stored in the ROM while utilizing the temporary storage function of the RAM.

[0017] (function) The character recognition device 10 has, as its main functional units, an image acquisition unit 21, a first information acquisition unit 22, a first guide setting unit 23, a guide candidate setting unit 24, a first range setting unit 25, a second information acquisition unit 31, a second guide setting unit 32, a promotion determination unit 33, an exclusion determination unit 34, a second range setting unit 35, and an item acquisition unit 36.

[0018] The image acquisition unit 21 acquires image data of a reference form 80 and a target form 90 from the scanner 15, the external storage device 16, or the like.

[0019] The first information acquisition unit 22 performs image recognition to recognize frame lines and characters over the entire image 81 of the reference form 80, and acquires frame coordinates and character strings. The acquired information is stored in the database of the character recognition device 10 in the form of a set of frame coordinates and character strings. As shown in FIG. 4, the database stores frame coordinates, character strings, and frame identification codes (hereinafter FIDs) "S-0" to "S-26" as information on the reference form 80. In this embodiment, the right direction when the upper left corner of the image 81 is taken as the starting point is represented by X, and the downward direction is represented by Y. The length of the frame in the X direction is represented by width W, and the length in the Y direction is represented by height H. The ratio of the width W to the height H of the frame is defined as the frame aspect ratio W / H.

[0020] The first guide setting unit 23 sets a frame in which characters are recognized in the image recognition of the reference form 80 as a "reference guide frame". Basically, a heading frame is set as the reference guide frame. As shown in FIG. 5, in the database of the character recognition device 10, in addition to frame coordinates and the like as information on the reference guide frame, a match type when matching a character string in the frame with a character string on the processing target form 90 is stored. The match types include "perfect match", "prefix match", "reverse match", and "partial match". In FIG. 5, a small number of reference guide frames are shown for the sake of simplicity, but in reality, all frames in which characters are recognized are set as reference guide frames.

[0021] The guide candidate setting unit 24 sets the frames other than the reference guide frame as the reference guide candidate frames among the frames recognized by the image recognition of the reference document 80. As shown in Fig. 6, the frame coordinates and FID are stored as information on the reference guide candidate frames in the database of the character recognition device 10. In Fig. 6, the number of reference guide candidate frames is shown as small to simplify the explanation, but in reality, all frames that were not set as the reference guide frame are set as the reference guide candidate frames.

[0022] The first range setting unit 25 sets a range for filling in the recognition target items in the reference form 80 as a reference entry range. As shown in Fig. 7, the reference entry range may be set such that the outer edge of the range exactly overlaps the entry box (i.e., the range exactly fits in the entry box), such as the range corresponding to the header box "Name", or such that at least a part of the outer edge of the range does not overlap the entry box (i.e., the range is smaller than the entry box), such as the range corresponding to the header box "Date of Birth". The first range setting unit 25 sets the reference entry range appropriately based on input information from the operator of the character recognition device 10.

[0023] As shown in Fig. 8, the database of the character recognition device 10 stores information on the reference entry range, such as frame coordinates, as well as information on the associated frame. The associated frame is the frame to which the reference entry range belongs. "The reference entry range belongs to the frame" means that the reference entry range exactly overlaps with the frame in the height direction, that is, the height of the reference entry range is set to the full height of the frame.

[0024] The second information acquisition unit 31 performs image recognition to recognize frame lines and characters on the entire image 91 of the target form 90, and acquires the frame coordinates and the character string within the frame. The acquired information is stored in the database of the character recognition device 10 as a set of the frame coordinates and the character string. As shown in Fig. 9, the database stores the frame coordinates, the character string, and FIDs "T-0" to "T-23" as information on the target form 90.

[0025] If, among the frames in which characters are recognized during image recognition of the document 90 to be processed, there is only one frame that satisfies the following two conditions (a1) and (a2), the second guide setting unit 32 sets that frame as the "guide frame to be processed."

[0026] (a1) The same character string as the character string in the reference guide frame is written. In other words, the character string recognized in the frame matches the character string in the reference guide frame when compared with the specified match type (see Figure 5).

[0027] (a2) The ratio of the frame aspect ratios W / H compared to the reference guide frame is within a predetermined range. In this embodiment, the ratio of the frame aspect ratio W / H of the reference guide frame to the frame aspect ratio W / H of the processing target guide frame is within the range of 0.9 to 1.1.

[0028] In this embodiment, the frame shown in Fig. 11 is set as the processing target guide frame based on the judgment result shown in Fig. 10. The FID of the processing target guide frame is set to be the same as the FID of the reference guide frame corresponding to that frame. The reference guide frame that is not adopted at this time is excluded from the reference guide frame as shown in Fig. 12, and is added (demoted) to the reference guide candidate frame as shown in Fig. 13.

[0029] The promotion determination unit 33 determines whether or not there is a frame that satisfies a predetermined condition among the actual frames recognized by image recognition of the processing target document 90, and promotes the frame that satisfies the condition to a processing target guide frame, and promotes the corresponding reference guide candidate frame to a reference guide frame. This series of processes, that is, the promotion determination, is performed in the following steps (b1) to (b3).

[0030] (b1) Based on the relative relationship between the reference guide frame and the processing target guide frame, a "theoretical frame" equivalent to the reference guide candidate frame is calculated on the processing target document 90. This calculation is performed in the following steps (b1-1) to (b1-4). In the following, X, Y, W, and H of the reference guide frame are X1, Y1, W1, and H1, X and Y of the reference guide candidate frame are X2 and Y2, and X, Y, W, and H of the processing target guide frame are X3, Y3, W3, and H3.

[0031] (b1-1) The scale Sx in the X direction and the scale Sy in the Y direction of the reference document 80 and the document to be processed 90 are calculated from the formulas (1) and (2). Sx = {Maximum value of (X3 + W3) - Minimum value of X3} / {Maximum value of (X1 + W1) - Minimum value of X1} (1) Sy = {maximum of (Y3+H3) - minimum of Y3} / {maximum of (Y1+H1) - minimum of Y1} (2)

[0032] (b1-2) The reference guide frame closest to the reference guide candidate frame is selected (see FIG. 14). Here, the determination of whether or not there is a reference guide frame whose height direction distance from the reference guide candidate frame is greater than a predetermined value is defined as a first distance determination. The "closest reference guide frame" is the reference guide frame closest to the reference guide candidate frame in the "height direction" if the first distance determination is positive, and is the reference guide frame closest to the reference guide candidate frame in the "width direction" if the first distance determination is negative.

[0033] (b1-3) The X-direction distance ΔX and the Y-direction distance ΔY between the reference guide candidate frame and the "closest reference guide frame" are calculated from equations (3) and (4). ΔX = X2 - X1 (3) ΔY = Y2 - Y1 (4)

[0034] (b1-4) The coordinates of the "theoretical frame" are calculated from the coordinates (X3a, Y3a, W3a, H3a) of the processing target guide frame having the same FID as the "closest reference guide frame" using formulas (5) to (8) (see FIG. 14). X, Y, W, and H of the "theoretical frame" are set to X4, Y4, W4, and H4. X4 = X3a + (ΔX × Sx) (5) Y4 = Y3a + (ΔY × Sy) (6) W4 = W3a × Sx (7) H4 = H3a × Sy (8)

[0035] (b2) In relation to the "theoretical frame," it is determined whether or not there is a frame that satisfies a predetermined matching condition among the "actual frame" recognized by image recognition of the document 90 to be processed (see FIG. 14). The above matching conditions are the following four conditions (c1) to (c4). In the following, W and H of the "actual frame" are assumed to be W5 and H5. Also, as shown in FIG. 15, the area of ​​the overlapping portion between the "theoretical frame" Ft and the "actual frame" Fa is assumed to be S1, and the area of ​​the rectangular virtual frame Fv that circumscribes the "theoretical frame" Ft and the "actual frame" Fa is assumed to be S2. (c1) The ratio of W5 / H5 to W4 / H4 is within the range of 0.9 to 1.1. (c2) W5 / W4 is within the range of 0.95 to 1.05. (c3) H5 / H4 is within the range of 0.95 to 1.05. (c4) (S1 / S2) × 100 is 80 or more.

[0036] (b3) If a frame that satisfies the matching condition exists, that frame is set as a processing target guide frame (see FIG. 16), and the reference guide candidate frame corresponding to that frame is set as a reference guide frame (see FIG. 17).

[0037] The exclusion determination unit 34 performs outlier detection on the guide frames that have been set up to this point (see FIG. 16 and FIG. 17), and excludes the guide frames that correspond to the outliers. This series of processes, that is, the exclusion determination, is performed in the following steps (d1) to (d5).

[0038] (d1) The scales Sx and Sy are recalculated from the above equations (1) and (2).

[0039] (d2) The frame coordinates of the guide frame to be processed are converted into the size (X3b, Y3b, W3b, H3b) of the reference form 80 by the formulas (9) to (12) (see FIG. 18). X3b = X3 / Sx (9) Y3b = Y3 / Sy (10) W3b = W3 / Sx (11) H3b = H3 / Sy (12)

[0040] (d3) The distance Dc between the center points of corresponding guide frames (that is, the distance between the center point of the processing target guide frame of the standard document size and the center point of the standard guide frame) is calculated (see FIG. 18).

[0041] (d4) It is determined whether the distance Dc between the center points is outside the determination range of the formula (13). The median of the distance Dc between the center points is Dcm, and the median absolute deviation is MedAD. Judgment range = Dcm ± (MedAD × 1.4826) × 6 (13)

[0042] (d5) As shown in Fig. 18, if the judgment in (d4) above is negative (i.e., if the center-to-center distance Dc is within the judgment range), the judgment result is OK, and the corresponding reference guide frame and processing target guide frame are maintained. On the other hand, if the judgment in (d4) above is positive (i.e., if the center-to-center distance Dc is outside the judgment range), the judgment result is NG, and the corresponding reference guide frame and processing target guide frame are excluded (see Figs. 19 and 20).

[0043] The second range setting unit 35 identifies a range on the target document 90 that corresponds to the reference entry range based on the relative relationship between the reference guide frame and the target guide frame, and sets the range as the target entry range. This series of processes is performed in the following steps (e1) to (e5).

[0044] (e1) The scales Sx and Sy are recalculated from the above equations (1) and (2).

[0045] (e2) A method of identifying the range on the target document 90 that corresponds to the reference entry range, i.e., a specific pattern is selected. If the determination of whether or not both of the following conditions (e2-1) and (e2-2) are satisfied is defined as a quota determination, then if the quota determination is affirmative (if both conditions (e2-1) and (e2-2) are satisfied), then specific pattern A is selected, and if the quota determination is negative, then specific pattern B is selected (see FIG. 21). (e2-1) The reference entry range belongs to the frame. That is, in Fig. 21, "Does it belong to the frame?" is "True." "Does it not belong to the frame" refers to a case where it is written outside the frame, such as the form name "Hospitalization and Surgery Certificate" in Fig. 2, or the height of the reference entry range shown with a shaded area is not set to the full height of the frame, as shown in Fig. 22. (e2-2) There is a guide frame to be processed that corresponds to the belonging frame. That is, there is a guide frame to be processed that has the same FID as the above-mentioned belonging frame in FIG.

[0046] (e3) When the specific pattern A is selected, the range corresponding to the reference entry range is calculated on the processing target document 90 based on the relative relationship between the belonging frame and the processing target guide frame corresponding to the belonging frame, and the range is set as the processing target entry range. Specifically, the coordinates of the processing target entry range are calculated from the formulas (14) to (17) (see FIG. 23). In the following, X, Y, W, and H of the reference entry range are X6, Y6, W6, and H6, X, Y, W, and H of the belonging frame of the reference entry range are X7, Y7, W7, and H7, X, Y, W, and H of the processing target guide frame corresponding to the belonging frame are X3c, Y3c, W3c, and H3c, and X, Y, W, and H of the processing target entry range are X8, Y8, W8, and H8. X8=X3c+(X6-X7)×(W3c / W7)...(14) Y8=Y3c+(Y6-Y7)×(H3c / H7)...(15) W8 = W3p × W6 / W7 (16) H8 = H3p × H6 / H7 (17)

[0047] (e4) When the specific pattern B is selected, a range equivalent to the reference entry range is calculated on the processing target document 90 based on the relative relationship between the reference guide frame and the processing target guide frame, and the calculated range is set as the processing target entry range. Specifically, the coordinates of the processing target entry range are obtained by the following steps (e4-1) to (e4-4) (see FIG. 23).

[0048] (e4-1) The reference guide frame closest to the reference entry range is selected as the reference guide frame (see FIG. 21). Here, the determination as to whether or not there is a reference guide frame whose height direction distance from the reference entry range is greater than a predetermined value is defined as the second distance determination. The "reference guide frame closest to the reference entry range" is the reference guide frame closest to the reference entry range in the "height direction" if the second distance determination is affirmative, and is the reference guide frame closest to the reference entry range in the "width direction" if the second distance determination is negative.

[0049] (e4-2) The X-direction distance ΔX and the Y-direction distance ΔY between the reference entry range and the “reference guide frame closest to the reference entry range” are calculated from equations (18) and (19). ΔX = X6 - X1 (18) ΔY = Y6 - Y1 (19)

[0050] (e4-3) The coordinates (X9, Y9, W9, H9) of the range corresponding to the reference entry range on the document 90 to be processed are calculated from equations (20) to (23) based on the coordinates (X3d, Y3d, W3d, H3d) of the guide frame to be processed which has the same FID as the "reference guide frame closest to the reference entry range". X9 = X3d + (ΔX × Sx) (20) Y9 = Y3d + (ΔY × Sy) (21) W9 = W3d × Sx (22) H9 = H3d × Sy (23)

[0051] (e4-4) The range calculated in (e4-3) above is set as the entry range to be processed. That is, X8=X9, Y8=Y9, W8=W9, and H8=H9 are set.

[0052] The item acquiring unit 36 ​​performs character recognition on the part of the image 91 of the form 90 to be processed that is within the range of the entry to be processed, and acquires the items to be recognized (see FIG. 23).

[0053] (Processing Procedure) Next, the procedure of the process executed by the character recognition device 10 will be described with reference to Fig. 24 and Fig. 25. The process includes a process shown in Fig. 24 that handles information on a reference form 80, and a process shown in Fig. 25 that handles information on a target form 90.

[0054] 24 when a predetermined execution instruction is given for information processing of the reference form 80. First, in step 100, the image acquisition unit 21 acquires image data of the reference form 80 from the scanner 15 or the external storage device 16. In the following, each step is abbreviated as "S".

[0055] Next, in S110, the first information acquisition unit 22 performs image recognition to recognize frame lines and characters in the entire image 81 of the reference form 80, and acquires frame coordinates and character strings. The acquired information, including the frame coordinates and the character string, is stored in the database of the character recognition device 10 as a set.

[0056] Next, in S120, the first guide setting unit 23 sets a frame in which characters are recognized in the image recognition of the reference form 80 as a "reference guide frame."

[0057] Next, in S130, the guide candidate setting unit 24 sets frames other than the reference guide frame among frames recognized by image recognition of the reference form 80 as reference guide candidate frames.

[0058] Next, in S140, the first range setting unit 25 sets, as a reference entry range, a range for entering the recognition target items in the reference form 80. After S140, the series of processes in FIG.

[0059] 25 when a predetermined execution instruction is given for information processing of the target form 90. First, in step 200, the image acquisition unit 21 acquires image data of the target form 90 from the scanner 15, the external storage device 16, or the like.

[0060] Next, in S210, the second information acquisition unit 31 performs image recognition to recognize frame lines and characters on the entire image 91 of the processing target form 90, and acquires the frame coordinates and the character string within the frame.

[0061] Next, in S220, if there is only one frame among the frames in which characters are recognized during image recognition of the document 90 to be processed that satisfies the two conditions (a1) and (a2), the second guide setting unit 32 sets that frame as the "guide frame to be processed."

[0062] Next, in S230, the promotion determination unit 33 determines whether or not there is a frame that satisfies the specified matching condition among the ``actual frame'' recognized by image recognition of the document 90 to be processed, and promotes the frame that satisfies the condition to a guide frame to be processed, and also promotes the reference guide candidate frame corresponding to that frame to a reference guide frame.

[0063] Next, in S240, the exclusion determination unit 34 performs outlier detection on the guide frames that have been set so far, and excludes any guide frames that fall under the outliers.

[0064] Next, in S250, the second range setting unit 35 identifies a range on the target document 90 that corresponds to the reference entry range based on the relative relationship between the reference guide frame and the target guide frame, and sets the range as the target entry range.

[0065] Next, in S260, the item acquiring unit 36 ​​performs character recognition on the part of the entry range to be processed in the image 91 of the form to be processed 90, and acquires the recognition target item. After S260, the series of processes in FIG.

[0066] (effect) As described above, the character recognition device 10 includes a first information acquisition unit 22, a first guide setting unit 23, a first range setting unit 25, a second information acquisition unit 31, a second guide setting unit 32, a second range setting unit 35, and an item acquisition unit 36. The first information acquisition unit 22 performs image recognition to recognize frame lines and characters in the entire image 81 of the reference form 80, and acquires frame coordinates and character strings. The first guide setting unit 23 sets a frame in which characters are recognized in the image recognition of the reference form 80 as a reference guide frame. The first range setting unit 25 sets a range in the reference form 80 for filling in items to be recognized as a reference entry range.

[0067] The second information acquisition unit 31 performs image recognition on the entire image 91 of the target form 90, and acquires the frame coordinates and the character string within the frame. If there is only one frame in which characters are recognized in the image recognition of the target form 90, in which the same character string as the character string of the reference guide frame is written and the ratio of the frame aspect ratios is within a predetermined range compared to the reference guide frame, the second guide setting unit 32 sets the frame as the target guide frame. The second range setting unit 35 specifies the range of the target form 90 corresponding to the reference entry range based on the relative relationship between the reference guide frame and the target guide frame, and sets the range as the target entry range. The item acquisition unit 36 ​​performs character recognition on the part of the target entry range in the image 91 of the target form 90, and acquires the recognition target items.

[0068] In this embodiment, frame recognition and character recognition are performed on the entire images of both the reference form 80 and the target form 90. This makes it possible to accurately grasp the relative relationship between the two forms by using a guide frame that associates information on character strings and frame coordinates as a set, rather than just the positional relationship based on frame coordinate information. In addition, by performing a demotion judgment so that frames whose information does not satisfy certain conditions are not treated as guide frames when setting the guide frame for the target form 90, it is possible to suppress a decrease in item acquisition accuracy caused by some frame lines not being recognized or characters being erroneously recognized.

[0069] In this embodiment, the character recognition device 10 further includes a guide candidate setting unit 24 and a promotion determination unit 33. The guide candidate setting unit 24 sets, as a reference guide candidate frame, a frame other than the reference guide frame among the frames recognized by the image recognition of the reference form 80. The promotion determination unit 33 calculates a theoretical frame corresponding to the reference guide candidate frame on the processing target form 90 based on the relative relationship between the reference guide frame and the processing target guide frame, and sets, as the processing target guide frame, a frame that satisfies a predetermined matching condition among the actual frames recognized by the image recognition of the processing target form 90 in relation to the theoretical frame, and sets the reference guide candidate frame corresponding to the frame as the reference guide frame. This allows more guide frames to be set, and the accuracy of the relative position calculation based on the relative relationship between the two forms can be improved.

[0070] In this embodiment, when the second distance determination is positive, the second range setting unit 35 specifies the processing target entry range based on the relative relationship between the reference guide frame that is closest to the reference entry range in the height direction and the processing target guide frame that corresponds to the reference guide frame. When the second distance determination is negative, the second range setting unit 35 specifies the processing target entry range based on the relative relationship between the reference guide frame that is closest to the reference entry range in the width direction and the processing target guide frame that corresponds to the reference guide frame. This makes it possible to minimize calculation errors when calculating the relative positions based on the relative relationship between the two documents (i.e., calculating the difference in frame coordinates).

[0071] Furthermore, in this embodiment, when the belonging frame determination is affirmative, the second range setting unit 35 calculates a range on the processing target document 90 that corresponds to the reference entry range based on the relative relationship between the belonging frame and the processing target guide frame corresponding to the belonging frame, and sets the range as the processing target entry range. When the belonging frame determination is negative, the second range setting unit 35 calculates a range on the processing target document 90 that corresponds to the reference entry range based on the relative relationship between the reference guide frame and the processing target guide frame, and sets the range as the processing target entry range. This makes it possible to automatically create coordinate information for the corresponding processing target entry range from information on the previously set reference entry range, even if the frame recognition state of the processing target document 90 is insufficient.

[0072] [Other embodiments] In other embodiments, some or all of the functional units of the information device serving as the character recognition device may be provided in another device, i.e., the character recognition device may be configured by a plurality of devices. In other embodiments, the functions of the functional units of the character recognition device may be realized by hardware processing using dedicated electronic circuits, or may be realized by a combination of software processing and hardware processing.

[0073] Although not specifically illustrated, the present invention can be implemented with various modifications without departing from the spirit and scope of the present invention. [Explanation of symbols]

[0074] 10 Character recognition device 21 Image acquisition unit 22 1st Information Acquisition Department 23 First guide setting section 24 Guide candidate setting section 25 First range setting section 31 2nd Information Acquisition Department 32 Second guide setting section 33 Promotion Judgment Department 34 Exclusion judgment part 35 Second range setting section 36 Item acquisition part 80 Standard Forms 90 Documents to be processed

Claims

1. A character recognition device that acquires recognition target items that are filled in within a predetermined range on a form that is composed of frame lines and characters, comprising: If an original document in which a writing box for writing arbitrary characters is left blank is defined as a reference document, and a document in which arbitrary characters are written in the writing box is defined as a processing target document, and the ratio of the width and height of the box is defined as the box aspect ratio, then: a first information acquisition unit that performs image recognition for recognizing frame lines and characters over the entire image of the reference form and acquires frame coordinates and character strings; a first guide setting unit that sets a frame in which characters are recognized in image recognition of the reference form as a reference guide frame; a first range setting unit that sets a range for entering the recognition target items in the reference form as a reference entry range; a second information acquisition unit that performs image recognition on the entire image of the target form and acquires frame coordinates and character strings within the frame; a second guide setting unit that sets a frame as a processing target guide frame when there is only one frame among the frames in which characters are recognized in the image recognition of the processing target form, the frame containing the same character string as the character string of the reference guide frame and the ratio of the frame aspect ratios of the reference guide frame being within a predetermined range; A second range setting unit that specifies a range of the target document corresponding to the reference entry range based on a relative relationship between the reference guide frame and the target guide frame, and sets the range as a target entry range; an item acquisition unit that performs character recognition on the part of the entry range to be processed in the image of the form to be processed and acquires the recognition target items; A character recognition device comprising:

2. The character recognition device includes: A guide candidate setting unit that sets a frame other than the reference guide frame as a reference guide candidate frame among frames recognized by image recognition of the reference form; a promotion determination unit that calculates a theoretical frame corresponding to the reference guide candidate frame on the processing target document based on the relative relationship between the reference guide frame and the processing target guide frame, and sets a frame that satisfies a predetermined matching condition among the actual frames recognized by image recognition of the processing target document in relation to the theoretical frame as the processing target guide frame, and sets the reference guide candidate frame corresponding to the calculated frame as the reference guide frame; The character recognition device of claim 1 further comprising:

3. If the determination as to whether or not there is a reference guide frame whose distance in the height direction from the reference entry range is greater than a predetermined value is defined as a distance determination, The second range setting unit is If the distance determination is positive, the processing target entry range is identified based on a relative relationship between the reference guide frame that is closest to the reference entry range in a height direction and the processing target guide frame that corresponds to the reference guide frame; 2. The character recognition device according to claim 1, wherein when the distance determination is negative, the processing target entry range is identified based on a relative relationship between the reference guide frame that is closest to the reference entry range in the width direction and the processing target guide frame that corresponds to the reference guide frame.

4. When the height of the reference entry range is set to the full height of the frame of the reference form, it is defined as belonging to the frame, and the frame to which it belongs is defined as an belonging frame. When the determination of whether or not the reference entry range belongs to a frame and the processing target guide frame corresponding to the belonging frame is defined as an belonging frame determination, The second range setting unit is If the determination of the belonging frame is affirmative, a range equivalent to the reference entry range is calculated on the processing target form based on a relative relationship between the belonging frame and the processing target guide frame corresponding to the belonging frame, and the calculated range is set as the processing target entry range; 2. The character recognition device according to claim 1, wherein when the belonging frame determination is denied, a range equivalent to the reference entry range on the document to be processed is calculated based on the relative relationship between the reference guide frame and the processing target guide frame, and the calculated range is set as the processing target entry range.

Images