JPS63293690A - Character recognizing system - Google Patents

Abstract

PURPOSE:To improve the operability and the generality by displaying a character pattern of the image information concerned on a display device from character frame information of a rejected character. CONSTITUTION:In a process that a recognizing operation is executed with respect to a series of each image file, and a recognizing information file is generated, a controller 21 stores whether a reject exists or not. After a series of operations is ended, the controller 21 searches for the one in which the top recognizing code of the recognizing information file generated the reject is a reject code determined in advance, calculates a character pattern information area in the corresponding image information file information from its character frame information, and displays it on a monitor 24a. By looking at the displayed character pattern in such a way, an operator executes a correction from a keyboard 24b. Said operation is repeated until the reject is all corrected.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a character recognition system, and particularly to unreadable (unrecognizable) processing.

(Prior Art) Conventionally, as this type of reading device, for example, [Automatic Recognition of Characters and Graphics for Computer Manually] (January 1, 1972)
5) There is something disclosed in the Institute of Electrical Engineers of Japan, P47.54-55.160. Its basic configuration consists of a scanning unit that obtains image information from the object to be read, such as a form, in the form of an electrical signal, a recognition unit that recognizes characters from the image information converted to electrical signals, and a recognition unit that recognizes the recognized character code. It consists of an interface section for directly transferring data to a higher-level control section such as a computer, or storing it in a file memory such as a FDD. If the reading operation is performed normally, character codes resulting from recognition are transferred or stored one after another. When a reading rejection occurs, one of the following three methods is used for glue jet processing. In the first method, the reading result of the rejected character is replaced with a special code and the reading operation continues.
The line in which the rejection occurred is marked, and in the second method, the form in which the rejection occurred is discharged to a stacker separate from normally read forms after the reading operation is completed. The third method is to display the character pattern including the rejected character and the recognition result on the monitor at the time when a rejection occurs or after the reading operation of the line where the rejection occurred and to stop the recognition operation, and then the operator This is to key in the correct character.

Note that the target is usually determined based on the results of several trials rather than one trial of the recognition operation. The signal used for recognition is a "1" and "0" signal that is finally binarized by a certain threshold value from an analog signal that has been photoelectrically converted in a scanning unit. This is because the character pattern that has been written changes and is often readable. Therefore, in the device configuration, when a character that cannot be read is generated, the data of that character is manually scanned through the scanning section again and the threshold value is changed to obtain a binary character pattern, and the other is that the multi-value data pattern is stored in advance. Some devices have a built-in memory for scanning, A/D convert the analog signal obtained by scanning, and store it in the memory, thereby obtaining a retry signal from the memory. The latter configuration usually requires a large amount of memory (for one line or one form, etc.)
Since signals can be obtained by electronically scanning the memory, it has the advantage of being able to read at high speed compared to mechanical scanning.

(Problems to be Solved by the Invention) However, reading devices using any one of the three methods described above have the following problems.

(1) In the first method, in which a reject line is marked in some way, the operator later looks at the marked line in the form and corrects the reading result, but this requires a very heavy workload.

(2) The second method, in which forms containing rejected characters are ejected into a separate stacker from normally read forms, is often adopted because an operator does not need to be present during the reading operation. ing.

However, when the rejected form is run through the reader again,
It is necessary to dedicate a reading device to perform the same work as in the first method. An even bigger problem is that the objects to be read are limited to forms that can be separated into individual sheets, such as forms.

(3) The third method, in which an operator keys in a character code from a character pattern displayed on a monitor when a reject occurs, requires an operator to be present during the reading operation because the occurrence of a reject cannot be predicted.

SUMMARY OF THE INVENTION The present invention aims to provide a character recognition system that reduces the burden on the operator in jet processing as described above, removes restrictions on the written paper that is to be read, and has excellent operability and versatility. .

(Means for Solving the Problems) In order to solve the above-mentioned problems, the character recognition system of the present invention includes: a scanning means for scanning an object to be read to obtain image information; a memory device for storing the image information; 1. A reading device that recognizes characters on a reading target based on image information read from the memory device and obtains recognition information of a recognition result to be stored in the memory device, and controls the scanning means and the memory device; , and a control device that determines whether the characters are unreadable based on the recognition information and outputs image information corresponding to the unreadable characters.

(Operation) The technical means of the present invention operates as follows. The control device sequentially stores image information obtained by scanning an object to be read, such as a form or paper, in a memory device using a scanning device (scanner). When a predetermined amount of image information, for example, one sheet of image information, has been stored, the recognition device recognizes characters based on this image information. The recognition information of the recognition result for one sheet of paper is stored in the memory device by the control device, and based on the recognition information, it is determined whether or not the paper is unreadable (rejected). For example, the recognition information includes the character code of the recognition result as well as the degree of similarity and character frame information (indicating at least the position), and those whose degree of similarity is equal to or less than a predetermined threshold are determined to be rejected. Therefore, by displaying the character pattern of the image information that corresponds to the character frame information of the rejected characters determined in this way on the display device, the operator can modify the rejected characters all at once. Therefore, the burden on the operator can be reduced. Further, since the image information is stored in the memory device and then the rejection process is performed based on the recognition result based on the stored image information, there is no restriction on the form of the object to be read.

(Example) The present invention will be described in detail below with reference to the accompanying drawings.

FIG. 1 is a block diagram of a character recognition system showing a first embodiment of the present invention. The system of this embodiment includes a recognition device 10
and a workstation 20. The workstation 20 consists of a control device 21, a scanner 22, a file memory device 24, and a human power device 24 connected thereto. The human power device 24 includes a monitor 24a as a display section and a keyboard 24b as an operation section.

Next, the operation will be explained.

First, before scanning an object to be read such as a form with the scanner 30 to obtain image information, an identification code is input using the keyboard 24b of the input device 24. The identification code specifies the area in the image information to be read that is to be recognized. This identification code usually has an identification number written in a specific area at the beginning of the form, and an area for recognition for each identification number is defined in advance by the recognition device 10. Therefore,
By recognizing the identification number, the recognition device GTO can determine the recognition target area, and because it does not recognize unnecessary areas, recognition speed is improved, and when designing forms, characters can be freely placed in areas other than the recognition target area. , the degree of freedom in design can be increased because figures can be described.

In this embodiment, the target is not only the form on which this identification number is written, but also includes printed matter such as books, so the recognition target area is defined for each sheet of paper or page, and the area mentioned above is is changed, keyboard 2
4b, you can enter it manually (key in) using a 3-digit code. However, if there are various papers with identification numbers attached, it is difficult to key in the identification code each time, so in this case, for example, key in the number 000, and if this 000 is keyed in. In some cases, the recognition area is defined by the identification number written on the paper, and in other cases, the recognition area is defined by the keyed-in code.

In this way, first, an identification code is input using the keyboard 24b, then the paper to be read is scanned by the scanner 22, and the output image information is converted into 8-bit image information by the A/D converter built into the scanner 22. / pixel resolution. This image information is stored in the file memory device 23 with a header identification code added to each sheet of paper. After a series of image information is sequentially stored in the file memory device 23 in this way, the control device 21 and the recognition device 10 are electrically connected. As a result, a series of image information is transferred from the file memory device 23 to the recognition device IO via the control device 21. In this embodiment, since the recognition device IO has a built-in temporary storage memory (not shown) that stores image information for one sheet of paper, the image information of one sheet of paper is the unit of transfer. Therefore, when the transfer to the temporary memory is completed, the recognition device IO executes the recognition operation, and when the recognition operation is completed, all the recognition result information is sent to the control device 2 with respect to the image information in the temporary memory. Transfer to. The control device 21 stores recognition result information (recognition information) in the file memory device 23, reads out the next image information transfer from the file memory device 23 again, transfers it to the recognition device 10, and repeats the same operation. Here, the above-mentioned recognition result information is the result of recognizing the recognition target characters, symbols, etc. written on the paper, and refers to the number of rejects, character code, and character frame information. The number of rejected items, the number of rejected items, and the character code are the codes of the candidates from 2nd place to 10th place among the first place candidates (including the reject code) in the recognition results. An explanatory diagram of character frame information is shown in FIG. 2. Character frame information is information detected during the recognition process, and as shown in the figure, in the circumscribed rectangle (circumscribed frame) of each character or symbol,
The address on the temporary memory corresponding to the top left vertex P (
) and the address length (ΔX, ΔY) on the temporary storage memory corresponding to the length of the side.

FIG. 3 ((), (0) shows the basic configuration of each sheet or page regarding the storage format in the file memory device 40. FIG. 3(A) shows the format of the image information file. Here, The serial number is a serial number set for each file, and is provided for confirmation when reading image information sequentially from the file memory device 4υ.The identification code is the 3-digit key-in code described above, followed by an 8-bit code. Image information expressed in density follows.The same figure (0
) is the format of the recognition information file corresponding to the image information file in (a) of the same figure, the serial number is the same as (a), and the identification code is usually the same as (a), but is 000.
When specified in (a), the code that is the result of recognizing the identification number written on the paper side is entered, and the number of glue jets in the same file is also entered. After that, the character codes and character frame information of the recognition results from 1st place to 10th place are included, and the recognition results are listed in order below. In addition, in this embodiment, a fixed length format is used for the file length because it is easy to process.Therefore, since the file length corresponding to the maximum paper size is prepared, valid data is determined from the identification code. .

In the process of performing a recognition operation on each series of image files and creating a recognition information file, the control unit 21 stores the presence or absence of a reject and, if there is a reject, the file name. , Finally, after all the recognition information files have been created, the same file may be read and examined again. After the series of operations described above is completed, if a reject has occurred, the operator corrects it using the same procedure as the third method described in the conventional example. That is, the control device 21 searches for a recognition information file in which a rejection has occurred whose first recognition code is a predetermined reject code, and extracts character pattern information in the corresponding image information file information from the character frame information. The area is calculated and displayed on the monitor 24. When displaying, most normal monitors can only handle binary data, so in that case, 8-bit data can be binarized using an appropriate threshold value and displayed. Looking at the character pattern thus displayed, the operator makes corrections on the keyboard 24b.
Start from Repeat the above operations until all rejects are corrected.

FIG. 4 shows another format example of the recognition information file. The difference from the one in Figure 3 (0) is that each
The degree of similarity (sometimes referred to as feature distance) of the 0th recognition result is added. Further, in this case, the recognition device 10 may or may not generate a reject. Therefore,
The number of rejects is not particularly transferred from the recognition device IO.

The presence or absence of rejection is determined by keying in a threshold value for the degree of similarity from the operator or by using a threshold value predetermined by the control device 21. The $ control device 21 sequentially checks the similarity of each character in the recognition information file and displays the characters on the display 24a as in the case described above as characters with a similarity below the threshold value.
It is the same for the operator to make corrections.

The advantage of using a recognition information file with this type of structure is that it can be used depending on the situation, such as when giving priority to character code information conversion even if the judgment is a little lenient, or when making it stricter to prevent errors and correct them by reject processing. It is something that can be separated.

Note that in the structure of either the recognition information file shown in FIG. 3 ([11) or FIG. This is because it is often used in text processing (determining characters based on subsequent character relationships, word concepts, etc.), which is often performed when a character recognition device and an automatic translation device are combined. In addition, monitor 24 during reject processing
The recognition candidate results are also displayed on the monitor 24a when character code key-in is performed for the character pattern displayed on screen a, a number is added to each candidate, and correction can be made by keying in the number, thereby making it easier for the operator. This is because the operation can be extremely simplified.

FIG. 5 is a block diagram of a character recognition system showing a second embodiment of the present invention. The file memory device 23 of the first embodiment shown in FIG. 1 is assumed to be, for example, a magnetic disk device with a fixed disk, but in the second embodiment, a large disk with a removable disk such as an optical disk device is assumed. A capacitive storage device is used as the file memory device 23 of the workstation 20, and the recognition device IO is provided with a similar file memory device 11. As a result, workstation 2
Even if the control device 21 and the recognition device 10 are separated, data can be exchanged by writing information on the disk and then setting it on the other side.

Also, for optical disk devices, a 5% inch diameter disk has a capacity of approximately 100 MB per side, and a 12 inch disk has a capacity of approximately 1.2
Since it has a G Byte user area, it can store image information for approximately 100 sheets of B5 paper on a 5% inch disk, and 400 sheets on a 12 inch disk.

As described above, according to this embodiment, rejected characters can be corrected all at once, and the operator only needs to respond when making corrections, and the processing time can be shortened since only the rejected characters are affected. This greatly reduces the burden on the operator. In addition, since the correction at the time of rejection is mainly performed by the operator, in the conventional system, the recognition device is occupied for a long time, but in this embodiment, the recognition device can be opened during that time, so the recognition device is used only for the recognition operation. I can concentrate on it. Therefore, it is also possible to connect a plurality of workstation systems to the recognition device and operate the recognition device efficiently.

Note that although the file memory device in this embodiment is described as storing image information and recognition result information on the same disk, it is easy to make correspondence between image information and recognition result information on two disks even if the disks are different. It is clear that similar operations can be performed if the configuration is such that it can read data from two disks.

(Effects of the Invention) As described above in detail, according to the present invention, it is possible to reduce the burden on the operator in reject processing, and to perform recognition processing efficiently without being limited by the format of the object to be read. Can be done.

[Brief explanation of drawings]

Fig. 1 is a configuration diagram of the first embodiment of the present invention, Fig. 2 is an explanatory diagram of character frame information, and Fig. 3 <4> and (0) are example formats of an image information file and a recognition information file, respectively. FIG. 4 is a diagram showing another format example of a recognition information file, and FIG. 5 is a configuration diagram of a second embodiment of the present invention. 10--Recognition device, ll, '2:l-File, memory device, 21--Control device, 22--Scanner, 2
4--Human power device, 24 a--Monitor, 24b-
··keyboard.

Claims (4)

[Claims] (1) A scanning device that scans an object to be read to obtain image information, a memory device that stores the image information, and a memory that recognizes characters on the object to be read based on the image information read from the memory device. a reading device that obtains recognition information of recognition results to be stored in the device; and a reading device that controls the scanning means and the memory device, and determines whether or not it is unreadable based on the recognition information to generate an image corresponding to an unreadable character. A character recognition system comprising: a control device that outputs information. (2) The image information includes an identification code indicating an area on the reading target, and the recognition information includes information on a character frame that is a circumscribing frame of a character resulting from recognition. Character recognition system. (3) The character recognition system according to claim 1 or 2, wherein the memory device is provided in the recognition device and the control device, respectively, and is a disk device with a removable disk. (4) The character recognition system according to claim 2 or 3, wherein the reading information further includes a degree of similarity corresponding to the candidate character code of the recognition result.