JPH0252328A - Image information reader - Google Patents

Abstract

PURPOSE:To extract and read necessary data efficiently, readily, and precisely by employing a microfilm in which image information containing data and form are imprinted. CONSTITUTION:A part of the transmissive light of a film 32 is divided by a half mirror 62, and the divided light is image formed on an image sensor 66 through an image forming lens 64. First, a discriminating device 68 reads an only data code 16, and then discriminates data necessary to identify the position, type, etc., of data. Consequently, the position (address), type of character, size, font, pitch, etc., of the data are discriminated to be stored in a data code memory 70. An analog image output of the image sensor 66, that is, the image containing form 10 and data is processed, and only necessary data is segmented in a data segmenting circuit 76. Since the only necessary data is extracted from the image, necessary data is readily identified out of the forms and data contained in the image and speedily and precisely read.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to an image information reading device that extracts and reads necessary data from a microfilm on which a predetermined form and data are recorded.

(Technical Background of the Invention) Information recorded in various forms, fixed asset registers, slips, customer management cards, hospital patient cards, etc. is stored as data in a computer, and can be stored in a specified manner when necessary. It is widely practiced to output data from a computer to a form with data entry fields and record and save this data on microfilm. In the past, this microfilm image was projected onto a league, read it with the human eye, and input it into a keyboard.

For this reason, reading data from a large amount of microfilm was difficult.

On the other hand, it is also possible to read the data recorded on microfilm using an OCR (optical character reader) and input it manually into a computer as a code, but the data area,
It is necessary to distinguish between forms and ruled lines, as well as font size, font pitch, font type (kanji or numbers, etc.), font type (Mincho, Gothic, etc.), etc.

For this reason, there have been problems such as a slow speed when reading characters and converting them into codes, and a decrease in reading accuracy.

Note that if the format of the microfilm is fixed, it is possible to read only a predetermined position on the microfilm using OCR, but there is a problem that it cannot be used when the format of the microfilm changes. .

(Object of the Invention) The present invention was made in view of the above circumstances, and it is an object of the present invention to efficiently, easily and accurately extract necessary data from a microfilm on which image information including data and forms is imprinted. An object of the present invention is to provide an image information reading device that enables image information reading.

(Structure of the Invention) According to the present invention, the object is to provide an image information reading device for extracting and reading data from a microfilm on which an image of data is recorded together with a predetermined form. an identification data reading means for reading identification data for reading, a discriminator for determining corresponding data using the read identification data, a data specifying means for specifying necessary data, and a data specifying means for specifying the necessary data. This is achieved by an image information reading device characterized by comprising a data cutting circuit that cuts out data using the output of the discriminator.

(Operation) The identification data reading means reads the identification data from the microfilm on which a predetermined form, data, and identification data for reading this data are recorded, and the discriminator reads the data position, character type, Find dimensions, pitch, font, etc. When necessary data is specified, the data necessary for reading this data is read out from this discriminator, and the data cutting circuit selects and cuts out this data part.

(Example) Fig. 1 is a diagram showing a microfilm image of a customer management card, which is an embodiment of the present invention, and Fig. 2 is a schematic diagram of an image information recording device that records the image of this card on microfilm. Figure 3 is a flowchart of its operation, Figure 4 is a schematic diagram of an image information reading device that reads images on this microfilm, and Figure 5
The figure is a flowchart of the operation.

An image of the card as shown in FIG. 1 is recorded on the microfilm. In this figure, the form (filling form) 10 has a large number of columns separated by vertical and horizontal partition lines, with odd numbered rows (
Lines 1, 3, 5, and 7) have item nature 12 in which the item name is written.

The even-numbered rows have an entry property of 14 having entry columns in which data corresponding to each item above them is entered. Also, data code 16 for identifying each data and indicating its position, character size, pitch, font type, etc.
are recorded in a predetermined positional relationship with corresponding data. In this embodiment, this data code 16 is written on the left side of each data, for example, by bar code representation. Also, in a predetermined position such as the margin on the right side of each entry property 14, a check code 18 for checking the data for each entry property 14 is written using a bar code display method. In FIG. 1, the data entered are only numbers and katakana, but of course names, addresses, etc. may also be written in kanji.

The card shown in FIG. 1 is recorded on microfilm by the recording device shown in FIG. In this Figure 2, 2
0 is a laser light source such as an argon laser, 22 is an optical modulator, 24 is a rotating polygon mirror, 26 is a lens, 28 is a galvanometer mirror, 30 is a lens, and 32 is a microfiche film. The light beam emitted from the light source 20 is modulated on and off by the optical modulator 22, and then modulated by the rotating polygon mirror 24.
is scanned in the main scanning direction (horizontal deflection direction) by the reflecting surface. The light beam reflected by the rotating polygon mirror 24 passes through the lens 26 and enters the galvanometer mirror 28 . This galvanometer mirror 28 deflects the light beam in the sub-scanning direction (vertical deflection direction) by the -scanning line interval during one scanning by the rotating polygon mirror 24. As a result, the light beam passing through the lens 30 and converging on the film 32 is
Two-dimensional scanning (raster scanning) will be performed on 2.

The type of form 10 and the data to be recorded on the film 32 are specified using the keyboard 34.

Of course, the data may be read from a memory (not shown) built into the computer (hereinafter referred to as CPU) 36 itself or from another CPU, but it may also be input directly from the keyboard 34.

A code specifying a form 10 is input into a form memory 38 which stores a variety of different form 10 patterns, and the corresponding form 10 is selected. The form video signal of the corresponding form 10 is then outputted by the form generator 40.

Part of the data specified on the keyboard 34 is transferred to a buffer memory 42 at a predetermined timing, and then converted into a data/video signal by a character generator 44. Also, a data code/video signal for displaying the identification data code 16 (see Figure 1) necessary for reading this data, and a check code for displaying the check code 18 (see Figure 1). - The video signal is output from the code video signal generator 46. Data code 16 includes data position, character type, font type, character size,
As mentioned above, it indicates the character pitch, etc. These form video signals, data video signals, data code video signals and check code
The video signals are added by an adder circuit 48, and a driver 50 controls on/off the optical modulator 22 based on the added signal. Note that the rotating polygon mirror 24, galvanometer mirror 28, and form/character/code generators 40.44.4
6 are controlled by the CPU 36, and each generator 40, 44 .
The video signal from 46 is directed to summing circuit 48 to control optical modulator 22 . As a result, as shown in FIG. 1, on the film 32, predetermined data is recorded in the predetermined data entry fields of the form 10 in a distributed manner together with the data code 16.
A check code 18 is also recorded at a predetermined position.

The above image recording process is the procedure shown in FIG. That is, when the keyboard 34 specifies a form 10 (step 100), the corresponding form 10 is read out from the form memory 38 (step 102). Further, the data designated by the keyboard 34 is input from the keyboard 34 itself or read from the memory of the CPU 36, and the type of data is input from the keyboard 34 (step 104). Since the address at which data is recorded changes depending on the form 10, the address to be recorded is determined by the form 10 to be used (step 106).
. On the other hand, the CPU 36 outputs data codes and check codes for identifying characters such as the recording data area, character type, font, size, pitch, etc. (step 10).
8). Then, corresponding data and forms or data codes 16 and check codes 18 are recorded at corresponding locations (addresses) on the film 32 while being dispersed on the image (step 110).

Next, an image information reading device for extracting and reading necessary data from the image information recorded on the microfilm 32 in this manner will be described with reference to FIG. 4.5.

In FIG. 4, 52 is a light source, 54 is a condenser lens, and the light from the light source 52 is guided to the microfiche film 32 by the condenser lens 54. film 32
The transmitted light is guided to a screen 60 by a projection lens 56 and an imaging lens 58, where it can be visually observed. Further, a part of the light transmitted through the film 32 is divided by a half mirror 62, and this divided light is passed through an imaging lens 64.
An image is formed on the image sensor 66 by this.

This image sensor 66 is one that moves a one-dimensional line sensor in a direction perpendicular to its length (sub-scanning direction),
A two-dimensional area sensor etc. can be used. Since the output signal of the image sensor 66 includes the entire image of the card shown in FIG. 1, first, only the data code 16 is read out from the signal, and the data for identifying the position and type of data is sent to the discriminator 68. (FIG. 5, step 200). As a result, the position (address), character type, size, font, pitch, etc. of each data are determined (step 20
2), which is stored in the memory 70; The type of data required in the form 10 is input from the keyboard 72 (which may be shared with the keyboard 34 of the recording device) (step 204), and the data code 16 for identifying this data is input from the memory 70. The data is output to a data extraction circuit 76, a preprocessing circuit 78, a feature extraction circuit 80, a standard character dictionary 84, etc., which will be described later (step 206).

The check code 18 is recorded at a predetermined position on the card or at a fixed position with respect to the form 10, and can be identified by reading the output from the predetermined position of the image sensor 66.

The analog image output of the image sensor 66, that is, the image including the form 10 and data, is input to the signal processing circuit 74 (step 208), where it is subjected to analog processing such as amplification, waveform shaping, and shading correction (step 210). . From the optical image signal processed in this manner, only necessary data portions are extracted by the data extraction circuit 76 (step 212).

When only the data to be read is extracted in this way,
The image signal of this data is image recognized according to a known image recognition process, and the data on the microfilm is converted into characters (
(including numbers) and output it to the memory of the CPU.

For example, the image recognition process is performed as follows. First, the preprocessing circuit 78 extracts the image signal character by character, moves or rotates the center of the image, enlarges or reduces the image as necessary, normalizes it, and then binarizes it into a digital signal. (step 216). At this time, defects in the characters are repaired, and if necessary, the lines are made thinner to make them easier to recognize.

After the preprocessing is performed in this way, the feature extraction circuit 8
Characteristics of image c character) are extracted with 0 (step 218
), and character recognition is further performed in the recognition circuit 82 (step 220). Various methods of feature extraction are possible, but for example, one can focus on the features around a pattern, first broadly classify it, then divide it into groups of similar characters, then finely distinguish the groups, and finally identify a single character. method is used. As for the characteristics around the pattern, for example, a four-sided code method is possible, in which long windows are provided on each of the four sides surrounding the character, and the amount of character lines that fall into these windows is coded. It is also possible to use peripheral characteristics based on the intersection distance from the frame circumscribing the kanji to the point where it collides with the character line.Classification can be done by the horizontal and vertical complexity of the 6 characters (complexity index), or by the stroke density of the characters. It is also possible to use a stroke distribution method that calculates the projection of the character pattern onto the coordinate axes and uses the characteristics of the waveform of the intensity distribution.

In this way, after the candidates are roughly classified by relatively low-order feature extraction and narrowed down to several tens to hundreds of candidates, characters are further determined using a higher-order method. For example, kanji are a combination of "hen" and "tsukuri", so
It is conceivable to use hierarchical structural analysis or hierarchical matching in which pattern matching is repeated at each stage.

In the first step, this method blurs the original 32x32 pixel character pattern to a 4x4 pixel gradation of 4 bits, and narrows down the candidates by checking the matching with a standard character dictionary 84 prepared separately. Next, in the second step, matching is performed in the same way using 8x8 pixels and 4-bit gradations, and candidates are narrowed down step by step in this way.

It is also possible to quantify the degree of deviation between the read character pattern and the standard character pattern in the dictionary using evaluation scales such as "distance" and "similarity."

In any case, when features are extracted using various techniques and character recognition is completed, post-processing is performed in the post-processing circuit 86 (step 222). This is the image sensor 66
The check code 18 read by the user is used to determine whether or not character recognition has been performed correctly. For example, one line of the data recorder line 14 of the form IO is checked.

The check code can be assigned using known methods such as providing a checksum for each byte of a character, using a check code for each item, or using several check codes in combination. If it is determined that there is an error in character identification as a result of the check using the check code, image identification processing is performed again in the re-identification circuit 88 using a more advanced image recognition method. For example, it is possible to use differential matching in which matching is achieved by emphasizing only the parts that differ from similar characters, or to use the difference in strokes from similar characters (differential stroke discrimination).

When correct character identification is performed as described above, this data is encoded and stored in the CPU memory (
Step 226). This data may be transferred to a CPU connected online or another CPU offline so that they can be used mutually.

Although the recording/reading device is explained separately according to its functions in FIG. 2.4, it goes without saying that the CPU can have some or all of these functions.

In the embodiment described above, the data code 16 for identifying the data area, character type, size, pitch, etc. is placed immediately before the data, but the present invention is not limited to this.

Further, in the above embodiment, as shown in FIG.
6 is shown using a one-dimensional barcode notation method, the present invention is not limited to this, and various methods such as a two-dimensional barcode display method can be adopted. As this data code, it is desirable to use a notation method that does not require complicated pattern recognition such as letters, but a simple display of predetermined fonts such as numbers can be used because it is easy to recognize.

Further, in the above embodiment, only the data code 16 is selected and read from the image including the form and data. The form code may be read to determine the form, and the image of this form is subtracted from the read image to create an image containing only the data and data code from which the form has been removed, and the data code may be read from this image.

In the above embodiment, data codes 16 for data identification are dispersed within the image. This data code 16 can be recorded all at one place within the image. In addition, while recording the identification data of various combinations in a separately prepared memory, only the discrimination mark of the identification data of the combination used is attached in the image, and when reading the data, the discrimination data of the combination used is determined from the discrimination mark. The identification data may be read from memory.

An example of the former method in which identification data is recorded at one location within an image is as shown in FIG. 6. In this figure, reference numeral 16A indicates a bar code display of data for identification used in reading the position of data in this image, type of characters, size, pitch, font, etc. Further, the reading device in case 1 is constructed in the same manner as shown in FIG. That is, the identification data code 16A read from the image is input to a discriminator 68, which discriminates the content of the identification data code and stores the content in the memory 70.

In addition, in the case of the latter method, in which a separate memory is prepared in which various combinations of data position arrangement, character type, size, font, pitch, etc. are stored,
As shown in FIG. 7, a discrimination mark 16B is attached to the image,
The identification data of the determined combination is read out from the memory. FIG. 8 is a schematic diagram of a reading device used in this case, and in this figure, 70A is a memory in which various combinations of different identification data are stored as described above. The discriminator 68A reads the corresponding combination of identification data from the memory 70A based on the discrimination mark 16B read from the image, and sends the necessary data area etc. from the contents to the data cutting circuit 76 and the like. The other operations are the same as in the case of FIG. 4, so the description thereof will not be repeated.

In the embodiments described above, the reading means for reading the identification data 16 is constituted by the image sensor 66, but a sensor other than the image sensor may be provided. Furthermore, although the data specifying means for specifying the data to be read is formed by the keyboard 72, in the present invention, the data specifying means may be made to be specified by another device such as the CPU 36.

(Effects of the Invention) As described above, the present invention records data necessary for identifying data such as the position, character type, character size, pitch, font, etc. on microfilm when reading data. This identification data is then read to determine the location of the necessary data, the type of characters, etc., and only the part of this necessary data is cut out from the image. The position (address) where the necessary data is located can be easily identified from among the data and read at high speed and with high precision. Therefore, even if the form changes, the target data can be read easily and simply.
Moreover, it becomes possible to carry out the process efficiently.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing a microfilm image of a customer management card read by the device of the present invention, FIG. 2 is a schematic diagram of an image information recording device that records this card on microfilm, and FIG. 3 is a flowchart of its operation. FIG. 4 is a schematic diagram of an image information reading device which is an embodiment of the present invention for reading images on this microfilm, and FIG. 5 is a flowchart of its operation. Moreover, FIG. 6.7 is a diagram showing an image of another method, and FIG. 8 is a schematic diagram of another reading device. 10... Form, 16... Identification data, 32...
...Microfilm, 66...Image sensor 68 as a reading means...
Discriminator, 72...Keyboard as data designation means, 76.
...Data extraction circuit. Patent Applicant Fuji Photo Film Co., Ltd. Agent Patent Attorney Yama 1) Written by Patent Attorney Yama 1) Hiroshi Shige

Claims (1)

[Scope of Claims] An image information reading device for extracting and reading data from a microfilm on which an image of data is recorded together with a predetermined form, comprising: identification data for reading the data recorded on the microfilm; an identification data reading means for reading, a discriminator for determining corresponding data using the read identification data, and a data specifying means for specifying necessary data;
An image information reading device comprising: a data extraction circuit that extracts necessary data using the output of the discriminator.