JP4443133B2 - Program for realizing slip processing apparatus and slip processing method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image input system that determines whether an image such as a slip is normal or abnormal and determines whether an image is correctly input by image processing.
[0002]
[Prior art]
It is required to store the image as data in order to review the processed slip and the like in the future. However, when inputting an image image, there are two types: an abnormality caused by noise, slipping and tearing of the slip, and an abnormality in which a back image is input even though a front image should be input. Both are problematic because the image of the slip is not correctly input. In particular, the method for determining the back image may be, in other words, a method for determining the front image or a method for determining either the front or back. Conventional methods for determining the back image include a method using a bar code printed at a predetermined position (for example, refer to Patent Document 1), or simply counting the number of pixels on the entire surface of the paper sheet. There is a method of determining whether or not the back surface is based on a preset threshold (see, for example, Patent Document 2).
[Patent Document 1]
Japanese Patent Laid-Open No. 9-50482 [Patent Document 2]
JP-A-4-223657 [Problems to be Solved by the Invention]
The front / back determination method using the bar code cannot cope with many kinds of actual slips, that is, atypical slips. In other words, in a distribution system, for example, it is necessary to input images of various non-standard slips at each company, and the front and back images are very diverse. Furthermore, a method that does not depend on the direction or inclination of the slip is desired. The method of Patent Document 2 counts the number of black pixels, and therefore can be applied only to the white back surface as shown in FIG. Further, it is determined whether or not the pixel density is uniform throughout, and the back surface can be determined when it is uniform, but it can be applied only to the determination when it is uniformly colored.
[0003]
[Means for Solving the Problems]
In order to solve the above problems, the outline of the invention disclosed in the present application will be described as follows.
Processing including an image acquisition unit that acquires an image of a slip including a form, a recording unit for the slip image, a determination unit that determines whether the slip image is abnormal, and a display unit that displays a result of the determination unit The slip processing device, wherein the determination unit performs the determination based on a distribution of pixel density for each partial area in the slip image.
[0004]
DETAILED DESCRIPTION OF THE INVENTION
FIG. 1 is a system configuration diagram of the present invention. In FIG. 1, an input unit 109 is a device for inputting images such as slips and forms, and is an image scanner such as an OCR or an image input system, or an image image input device such as a digital camera. The input control unit 102 is a device that controls the image input. The display unit 110 is a device that displays images and their recognition results, and the display control unit 103 is a device that controls them. The storage device 106 is a device that stores images. The display unit may be a voice output or the like, and may be a notification means that an abnormal slip has been detected by the user. The stacker 111 is a stacker used when a paper-like material such as a slip is ejected outside the system. A plurality of stackers are usually prepared, and can be classified into two types, for example, normal slips and abnormal slips. The stacker control unit 104 controls the operation of the stacker described above. The abnormal form determination unit 105 can determine the front and back of the slip, or can determine an abnormal slip due to tears, noise, or the like.
[0005]
Hereinafter, a slip or a form is an image that is input to an OCR or an image input system, and is a term indicating a general paper-like form such as a slip, a form, or an application form.
Next, the slip will be explained. FIG. 4 is an example of a table image of a slip. The image is a multi-color dropout image. Dropout is a word commonly used in the field of OCR. For example, if it is a red border, the border is removed by the red dropout, and the pattern to be read (in this case, a character) Only the remaining image becomes. In recent years, the design of slips and forms has been diversified, and the colors used are also varied. Such a voucher needs to be dropped out of multiple colors instead of a single color. FIG. 4 shows an image with multiple color dropouts. The characters 402, 407, 405, 404, etc. are left, and the barcode 406 is also left. A punch hole 403 is also left. If the dropout of multiple colors is not perfect, some ruled lines 408 may remain. FIG. 5 is an example of the back image 501. This is an example of the case where black 502 of carbon has a relatively large area. FIG. 6 is also an example of the back image 601 and is an example of a case where the black area due to the carbon 602 is small. FIG. 7 is also an example of the back image 701, which is usually white as in the back image 1001 shown in FIG. 10, but in the example of FIG. 7, noise 702 due to black carbon remains in a relatively large area. Incidentally, the edges 401, 501, 601, 701, 1001 of the slip do not appear in the actual image as black lines. In other words, this is a convenient line for presentation as a diagram. It is possible to assume that the actual image has no line at the edge of these slips, or the background is black as shown in FIG. In the present application, in particular, for the purpose of determining the front and back of the image as disclosed in FIGS. 5 to 7, it is possible to process the image with or without the background.
[0006]
FIG. 2 shows an example of a system flow when the input slip is determined as an abnormal slip. The abnormal slip determination 201 will be described in detail later. If the slip of the slip is large, it is a torn slip, or a slip in which an image is not normally taken due to a torn small piece of paper, This is a process of determining that there is. Further, as will be described later, it is possible to set a reject when the parameters in the front / back determination are within a certain range, and such a reject slip is one of the abnormal slips. Here, the abnormal slip determination includes a case where the back image is taken when the front image should be taken. Of course, some systems may aim to take a back image. When the determination branching unit 202 of the abnormal slip determination 201 does not determine that it is an abnormal slip (“N” in the figure), the flow flows to the slip end determination 209 until the slip input determination 209 eliminates the input slip. The processing from the abnormal slip determination 201 is repeated.
[0007]
When the abnormal slip determination 201 determines that the slip is abnormal (“Y” in the figure), the count processing 203 records the number of abnormal slips. If the number of abnormal slips reaches a predetermined number set in advance, the system stop process 205 is repeated, and if not, the top process is repeated. The prescribed number is arbitrary, and when set to zero, the system is immediately stopped when an abnormal slip occurs. If a certain number of sheets is allowed, the number may be set in advance. In this case, to stop the system means to stop the image input of the slip. Specifically, canceling the image input may be a state in which the system is simply not moved, that is, a state in which the slip remains in the scanner, or the slip may be returned or sent depending on necessity. You may do it. After the system is stopped by the system stop process 205, for example, the display unit 110 is used to display a warning to the user that the system has been stopped due to an abnormal slip. This warning may be a sound such as a buzzer or sound, or information on light such as a scanner or an LED or a flash inside the scanner. Further, after the warning display 206, the user can select whether or not to issue a true system stop request. A true system stop request is a process for discharging a slip remaining inside a scanner or the like. In this case, the abnormal slip is discharged by the abnormal slip discharge 208. With respect to slips that have already been ejected, a list of slip numbers or the like can be displayed on the display unit 110 or the like. The place where the slip is discharged may be the same place as a normal slip, or may be another place for distinction. The place here is called a stacker in the case of OCR. Usually, there are a plurality of stackers, and the stacker to be discharged may be changed according to the degree of abnormality. The user stop request 207 may not be present. In this case, the abnormal slip discharge 208 is immediately after the warning display 206. This series of processing is repeated until there are no more slips. A counter may be provided according to the type of abnormal slip, and the prescribed number may be divided. For example, the number of reject slips may be relatively large, and a certain number of sheets may be discharged to another stacker. However, an abnormal slip such as a tear may need to be immediately stopped.
[0008]
FIG. 3 is an example of a system flow associated with front / back slip determination. Although the processing is almost the same as that in FIG. 2, there is a difference in that the abnormality determination processing is the front / back determination 301 processing. Although details will be described later, the front / back determination 301 may be front / back determination or back determination. Of course, the front or back may be determined at the same time. When the front / back determination 301 determines that the back side, the count process 303 is performed. This system is a system in which a front image is to be input. In contrast, in a system in which a back image is to be input, it goes without saying that the back determination 302 is a front determination. In the following, the above idea applies in all cases. After the count process 303, a slip discharge process 304 is performed. This is the same process as the abnormal slip discharge 208. The slip discharge process 304 may be omitted only by the count process 303. This is because, depending on the system, only the number of sheets may be counted. These series of processes are repeated until there are no slips. When the count is added by the count process 303, the system stop process 308 is performed and the warning display 309 is performed at the end of taking all the slip images. These processes are the same as the warning display 206 and the system stop process 205, respectively. These warning display and system stop processing may be performed in reverse order.
[0009]
Details of the front / back determination are shown in FIG. First, the slip image is divided into meshes of an arbitrary size (1200). In other words, it divides equally vertically and horizontally. Each mesh is, for example, a square of 16 pixels vertically and 16 pixels horizontally. Of course, these sizes vary depending on the size and resolution of the image. Hereinafter, the case of this mesh size will be described. For all meshes in the image, the number of pixels in the mesh is counted for each mesh (1201). The minimum is 0 and the maximum is 256. As described above, the back image has many black or white meshes, and the front image has many intermediate numerical meshes such as characters, barcodes, punch holes, and the like. The front / back determination is performed by dividing the number of pixels in the mesh into three types in this way (1202) and determining how many meshes are in the image. Dividing the number of pixels into three types means, for example, dividing into 0 to 16, 17 to 240, and 241 to 256. The white mesh frequently appears in the back image, the intermediate mesh frequently appearing in the front image, and the black mesh frequently appearing in the back image, respectively. These types of meshes are counted and, of course, normalized by the area of the entire image or the total number of meshes. If there are a number of back meshes within a specified range, it can be determined that the back image, otherwise the front image. For example, since black carbon as shown in FIG. 6 counts black meshes, FIG. 6 enables accurate back determination. On the other hand, for the noise as shown in FIG. 7, the intermediate mesh is counted and may be erroneously determined in the table, and further detailed determination is required. Therefore, in the intermediate mesh, the number of noises is counted, and if it is large, it is determined as a noise mesh (1203) and excluded from the front / back determination, or if the number of noise meshes is large, the slip itself can be rejected as an abnormal slip. In the noise determination method described above, isolated point noise (one pixel in which eight neighboring pixels are all white pixels) is registered as a basic pattern for determination. Furthermore, depending on the size of the slip image, it may correspond to noise even for several pixels, and since it is a value that changes according to the size of the slip image, the above-mentioned noise is not necessarily isolated point noise, It may be a cluster of several pixels. Incidentally, the mesh may of course cover the entire image as described above, but the edge of the slip is often unstable due to inclination, tearing, distortion, and a black background. In many cases, the mesh for such unstable images is not effective for the judgment. Therefore, it is practical to limit the sampling of the mesh to the vicinity of the center of the image or to exclude the periphery of the image. Is effective.
[0010]
So far, in the front / back determination of the present application, the method of dividing into meshes has been disclosed, but the method of dividing into meshes is an example, and the distribution of black pixels (or white pixels) for a certain area in the image is shown. It only needs to be able to analyze. For example, a method may be used in which a plurality of predetermined sampling points measure the pixel density taking into account pixels around the point and analyze their distribution.
[0011]
For example, in the case where there is only a back image as shown in FIG. 5 where there is a pure white back image or a large area of black carbon, a method of simply counting the number of black pixels and determining by any number is effective. However, according to the configuration of the present application, compared to the above method, for example, the conventional method can determine only a uniform back image, but various back images in which carbon black is partially present. Can also be determined.
[0012]
The flow of FIG. 9 on the assumption that it is operated as an actual image input system will be described. The slip take-in process 900, which is not described in FIGS. 2 and 3, is included in the same manner, and is a process for actually taking a slip into the apparatus when inputting the slip image. Here, for example, as shown in FIG. 4, processing of a slip having a barcode on the surface is assumed. Barcode determination 901 is determination processing for detecting a barcode from a front image of a slip, and the method described in JP-A-2001-43308 can be used. If the barcode determination 901 determines that the barcode is present (902), the table image is registered in the system by the image registration 906 because it is a table image. If it is determined that there is no barcode, a front / back determination 903 is performed. The front / back determination 903 is the content of the present invention described later. If the front / back determination 903 determines that the front side is the front side (904), the image registration 906 is executed. If it is determined to be the reverse side, a slip discharge process 905 is executed. The slip discharge process 905 is as described above, and discharges the slip to another stacker. These series of processes are repeated until the slip is completed (907). These series of processes are not necessarily required and may be completed only once. In other words, a system is assumed in which a slip image is input manually by a stand type scanner, a flatbed scanner, a digital camera, a copier, or the like.
[0013]
In an image input system that collects a front image, a mistake that mistakenly determines that the back image is a front surface is the most problematic. One method is to set a reject when it is difficult to determine. For example, when the distribution of the number of black meshes, intermediate meshes, and white meshes described above is taken for each of the front image and the back image, it has a distribution that is highly likely to be both the front and back sides. Is rejected (see FIG. 13). A method of positively determining rejects or front images in the previous stage and determining all remaining images as back images is preferable. This is because the slip image samples are finite, and even if the rejection parameters as described above are set from these images, it is inevitable that erroneous determination will occur. This is because the front image has a relatively stable number of intermediate meshes, whereas the back image has an unstable distribution due to the type of black carbon or noise. In other words, if parameters are tuned with a front image with a stable distribution and all images that deviate from it are determined to be back, either an unknown front image or an unknown back image comes in as input. Since the image can be determined to be the back image, a warning is possible, and the user can check the information sequentially and send it to the system as it is if it is on the front, and flip it over if it is the back and input the front image again. If this is the other way around, the table image to be registered is not input, which is a serious problem. Therefore, the above method is very effective. FIG. 13 shows a specific example. If the intermediate mesh and the black mesh are taken as axes, the distribution of the remaining white mesh is not necessary. This is because the sum of the three types of meshes is invariant, and two of them should be used as parameters. An area “back” represents the distribution of the back slip, and an area “table” represents the distribution of the front slip. In addition to the area where “front” and “back” overlap, the upper right area is also set as the reject area. As described above, since it is an unstable region, it can be said that it is correct to set it as a reject region.
[0014]
FIG. 8 is an example of a front image that is desired to be excluded as abnormal except for front / back determination. As shown in FIG. 8, such a white band 801 is generated when a small paper piece with a torn slip remains in the scanner for an image with a black background. Even if such a table image is registered, for example, an inquiry from the user cannot be answered, so it is necessary to detect the occurrence of such a white band 801 as an abnormality. In the case of an image with a black background, the white band 801 has penetrated to the black background, and according to the configuration of the present application, the detection is performed by looking at the distribution of the positional relationship between the black mesh and the white mesh in detail. Easy. Specifically, a white band can be easily detected by detecting the presence of a white mesh having a certain width in the black background distribution (1204).
FIG. 11 shows another example of an abnormal table image that is an image in which the edge of a slip is cut. If it is a black background, it is easy to detect the ends 1101 and 1105 of the slip, and it is also easy to detect their length, that is, the width 1102 of the slip. If the black background is used, it is easy to detect the portion 1103 where the slip is cut and the length 1104 thereof. Similarly in the vertical direction, detection of 1106 to 1110 is easy. When the length of 1104 for 1102 and 1107 for 1109 is larger than a predetermined value, the edge of the slip is It can be determined that it is torn.
[0015]
There is no problem if the back determination slip discharged to another stacker is actually back, but there is also a slip that is back determined even though it is a front. Even if such a voucher is input again to the system again, there is a high possibility that it will be discharged to another stacker. Therefore, in the actual system, it is effective to input the image by specifying explicitly that the back judgment slip discharged to such another stacker is a front image after aligning the front and back with human eyes. It is. That is, it is necessary to be able to switch the mode without front / back determination according to a user request. Specifically, in FIG. 3, the flow skips 301, 302, 303, 304, 307, 308, and 309. A key or the like for this instruction may be provided in the apparatus. By adding such processing, the accuracy of detecting an abnormal image can be increased.
In order to make the front / back determination more accurate, keywords such as logos and characters, and color information are positively registered, and the information may be used. In general, slips have a pair of images on the front and back, so that information on the pair should be registered. Such information includes the above-described logo, keyword, and color information, and may be arrangement information for the above-described black, white, and intermediate mesh images. If these pieces of information are stored in the storage device in association with the front and back information in advance, they can be used for abnormality determination, and the determination accuracy can be improved.
[0016]
This front / back determination and abnormal slip determination are image processing, and can be mounted on a copier, FAX, ATM, or the like. Since it is possible to determine an abnormal slip or a back slip, a user merit is generated, and thus it is possible to configure a fee collection mechanism. For example, it is a method of collecting a fee for each slip. This need not be taken from the end user and may be received from the purchaser of the device. Alternatively, an additional fee may be charged as an option when selling the device.
[0017]
【The invention's effect】
As described above, according to the present invention, it is possible to stably determine the character type of a character in a document or slip, and to output a highly reliable character recognition result. Become. In particular, according to this method, not only on the uniformly white back surface, but also on various back surfaces where carbon black is partially present, and without depending on the inclination and direction of the paper sheets, Easy to apply.
[0018]
[Brief description of the drawings]
FIG. 1 is a system configuration diagram of the present invention.
FIG. 2 is a system flow associated with abnormal slip determination.
FIG. 3 is a system flow associated with front / back slip determination.
FIG. 4 shows an example of a front image (No. 1).
FIG. 5 shows an example of a back image (part 1).
FIG. 6 is an example of a back image (part 2).
FIG. 7 illustrates an example of a back image (part 3).
FIG. 8 shows an example (part 2) of a front image.
FIG. 9 is a front / back determination system flow with barcode recognition.
FIG. 10 is an example of a back image (part 4).
FIG. 11 shows an example of a front image (No. 3).
FIG. 12 shows details of front / back determination.
FIG. 13 shows an example of parameters for slip front / back determination.
[Explanation of symbols]
101: Central control unit 102: Input control unit 103: Display control unit 104: Stacker control unit 105: Abnormal form determination unit 106: Storage device 109: Input unit 110: Display unit 111: Stacker 201: Abnormal slip determination 202: “Abnormal "?" 203: Count processing 204: "Are the specified number reached?" Determination 205: System stop processing 206: Warning display 207: "User stop request?" Determination 208: Abnormal slip discharge 209: "Slip end?" Judgment 302: “Back?” Judgment 303: Counting process 304: Voucher discharge process 305: “Voucher end?” Judgment 307: “Counting?” Judgment 308: System stop process 309: Warning display 401, 501, 601, 701, 1001: Edges 402, 403, 404, 405, 406, 407 of a slip: Parts 502, 60 constituting a table image : Black portion 702 by carbon 702: Noise 801 by carbon 801: White band noise by torn piece of paper 900: Voucher fetching process 901: Bar code determination 902: “Bar code present?” Determination 903: Front / back determination 904: “Front?” Determination 905 : Slip discharge processing 906: image registration 907: “slip end?” Determination.

Claims (10)

A slip image acquisition unit;
A recording unit for the slip image;
A determination unit for determining whether the slip image is front or back;
A display unit for displaying the result of the determination means,
The determination unit divides the slip image into a plurality of partial areas, classifies the partial areas into three types based on the number of pixels for each partial area, and based on the appearance frequency of the three types of partial areas. A slip processing apparatus that performs determination.   The slip processing apparatus according to claim 1, wherein the three types of partial areas are classified into a black mesh, an intermediate mesh, and a white mesh in descending order of the number of pixels. A recording device for recording the symbol or mark;
The slip processing apparatus according to claim 1 or 2, wherein the determination unit determines the presence or absence of the symbol or mark, and performs the determination using the determination result.   4. The slip processing apparatus according to claim 3, wherein the determination unit performs the determination after determining the presence or absence of the mark.   3. The slip according to claim 2, wherein the determination unit determines whether the noise mesh is included by counting the number of noises included in the intermediate mesh, and excludes the noise mesh from the determination. Processing equipment. A stacker for discharging the slip inserted in the image acquisition unit;
6. The slip processing apparatus according to claim 1, wherein the slip determined to be the back surface is output to the stacker different from other slips.   3. The slip processing according to claim 2, wherein the determination unit also analyzes a positional relationship of the classification results of the partial areas, and determines that the slip is an abnormal slip when a white mesh is present in a black mesh. apparatus.   8. The slip processing apparatus according to claim 1, further comprising means for outputting a warning that the back surface has been detected. Get the slip image,
The above slip image is divided into a plurality of partial areas,
Based on the number of pixels for each partial area , the partial area is classified into three types,
A program for causing a computer to execute a slip processing method that determines whether the form image is front or back based on the appearance frequency of the three types of partial areas.   10. The program according to claim 9, wherein the three types of partial areas are classified into a black mesh, an intermediate mesh, and a white mesh in descending order of the number of pixels.

Images