JPS6355684A - Optical character reader - Google Patents

Abstract

PURPOSE:To efficiently read a slip that includes an illustration part by attaching prescribed first and second marks to determine the range of the illustration part in the slip. CONSTITUTION:For a slip 1-1 in which a character part and an illustration part are mixed with each other, the vertical boundary between the character part and the illustration part is attached with the first mark, and the second mark whose length is equal to the width of an illustration part is attached above the illustration part. These marks are detected by a mark detection sensor 1-10, and the control of the transfer of the slip is executed in accordance with the result of the said detection, and at the same time, the outputting of an image data stored in a buffer part 1-6 is changed over to a character recognition part 1-8 and to an image processing part 1-9.

Description

DETAILED DESCRIPTION OF THE INVENTION (Industrial Application Field) The present invention relates to an optical character reading device (hereinafter referred to as OCR), and particularly relates to reading control of text and illustration portions on a form.

(Prior Art) FIG. 5 is a schematic diagram showing the configuration of a reading section of a conventional OCR. The form 5-2 stacked in the hopper 5-1 is conveyed along the guide 5-4 by the form running roller 5-3. Then, the lamp 5-5 illuminates the form 5-2, and the reflected light is input to the photoelectric converter 5-7 through the lens 5-6''f:, where it is converted from an optical signal to an optical signal.

FIG. 6 is a block diagram showing an example of the circuit configuration of a conventional OCR including a photoelectric conversion section. The sensor section 6-1 includes a photoelectric conversion section 5-7 shown in FIG. 5, and here the lens 5-6 shown in FIG.
Photoelectrically converts the original gold input through the A/D conversion circuit 6
-2. The A/D conversion circuit 6-2 is connected to the sensor section 6-
The analog signal from 1 is converted into a distal signal and output, and the output is stored in the buffer 6-3 as image data. The data stored in the buffer is extracted character by character by the character recognition section 6-4 and recognized.

By the way, recently, printed materials such as papers and articles are often converted into a database using OCR, and as shown in Figure 7, not only the text parts (7-1, 7-3) but also advertisements,
It is becoming more common to read forms that have an illustration section (7-2) such as an illustration.

Conventionally, when reading a form as shown in FIG. 7, first the format information of the text section 7-1 indicated by diagonal lines is registered, and then the format information of the text section 7-3 is registered, excluding the illustration section 7-2. The formant information was registered and read according to the formant information.

Here, the format information generally refers to the size of the form, the length of the form, the position of the reading line, the type of font used in one line, the type of font used,
This refers to information such as line pitch, etc., and this information is registered, and when reading, document transport control and data storage control in the buffer are performed based on this information. Enter the boundary mark between the text section and the illustration section at the right end (or left end) of the form, as shown in the following.
A method has also been proposed in which switching between character recognition and image processing is controlled by detecting this mark. This method detects the mark 8-1 to detect the character part (7 in Figure 7).
-1) is read, the character is recognized, and the illustration part (corresponding to 7-2 in Figure 7) is read as an image by the detection of the mark 8-2, and then the character is recognized by the detection of the mark 8-3. It reads the character part (corresponding to 7-3 in FIG. 7) and performs character recognition.

(Problems to be Solved by the Invention) However, for documents with illustrations, it is necessary to define the i-reading for each document to be read, which is complicated.

It is also possible to treat the illustration part as text and omit the illustration part after reading the entire document, but this correction work is also time-consuming.

In addition, in the method shown in Figure 8, in which a boundary mark between the text and illustration sections is written on the right edge (or left edge) of the form, this mark is detected, and the reading of the text and illustration sections is switched and controlled. It has been impossible to control a form such as shown in FIGS. 9 and 10, in which the inside of a horizontally written form is divided into two stages and has a horizontal half or a partial illustration part in the horizontal direction.

On the other hand, since there are surprisingly many pages with this type of layout in technical magazines, etc., there has been a desire for efficient reading of such pages.

The present invention has been made in order to solve the problems of the prior art as described above, and it is possible to simplify the creation of format information and efficiently separate the reading between the illustration part and the text part. The purpose is to provide OCR that can be used.

(Means for Solving the Problems) The present invention includes a reading unit that transports a form, irradiates text and illustrations on the form, and converts the reflected light into an electrical signal, and a reading unit that converts the reflected light into an electrical signal. , an A/D conversion section that performs A/D conversion, a buffer section that temporarily stores the output of the A/D conversion section as image data, and a character recognition section that performs character recognition on the character part of the image data of the Pana section. and an image processing unit that performs image processing on the image data related to the illustration part of the image data in the buffer part. A first mark is attached to the direction boundary, and a second mark having a length equal to the width of the illustrated section is attached to the upper part of the illustration section, and mark detection means for detecting the first and second marks is provided.

In accordance with the detection results of the mark detection means, form conveyance control is performed, and the output of image data stored in the buffer section is switched between a character recognition section and an image processing section.

(Function) In the present invention, in order to define the range of the illustration part in the form,
A first mark is attached to the vertical boundary with the text section, and a second mark having a length equal to the width of the illustration section is attached above the illustration section. Therefore, it is possible to reliably define the range of the illustration portion that does not occupy the entire lateral direction of the form sheet. Then, the first mark and the second mark are detected by the mark detection means, and based on the result, switching control of the reading between the character part and the illustration part is performed.
Even forms with illustrations like the one shown in the figure can be read efficiently.

(Example) An embodiment of the present invention will be described in detail below.

FIG. 1 is a block diagram of the OCR of this embodiment. Form 1-1
is conveyed by a running roller 1-2 and illuminated by a lamp 1-3. The data of the portion illuminated by the lamp 1-3 is input to the image sensor 1-4 as reflected light and photoelectrically converted. And the sensor output is A/D conversion circuit 1
-5, and is A/D converted and stored in buffers 1-6. The output of the buffer 1-6 passes through the switching circuit 1-7,
It is connected to a character recognition section 1-8 or an image processing section 1-9. On the other hand, a mark detection center 1-10 is provided at the position where the right edge of the form is, and the detection results are sent to the control unit. As shown in Figure 2, the main text (9 in Figure 9)
Mark 2-1 (mark A) is written on the right side of the form sheet corresponding to the first line (corresponding to -1). Next is the illustration section (
Corresponding to 9-2 in Figure 9) and the text next to it (corresponding to 9-2 in Figure 9)
Mark 2-2 (mark A) is written near the first line position of the line (corresponding to -3). On the other hand, in the vertical direction, a mark 2-4 (mark B) having a length equal to the width of the mark 2-4 is written at the top of the illustration section at the same line position as the mark 2-2. Furthermore, mark 2-3 (mark A) is written on the right side of the form sheet corresponding to the first line of the next text (corresponding to 9-4 in FIG. 9). It is assumed that the format information includes the paper size, line length, and font in the line.

In the OCR of this embodiment, first, the mark detection sensor 1-10, which is attached before the image sensor 1-4, detects the mark 2-1 (mark A) on the form 1-1, and the detection result is It is sent to the control unit 1-11. and control section 1-11
control the running roller 1-2 according to the control signal from the
The form 1-1 is conveyed to the mark detection position, and the mark 2-
The image data of the character line containing 1 (mark A) is input into the buffer 1-6. At this point, the buffer 1-6 is searched for mark B indicating an illustration part which is longer than the character pitch. The contents of the buffer 1-6 can be freely accessed from the control section 1-11. This mark 2
In the case of a line marked with -1 (mark A), mark B cannot be detected because all the lines are characters. Therefore, this block (corresponding to 9-1 in FIG. 9) is regarded as a character block, and the switching circuit 1-7 transfers the contents of the buffer 1-6 to the character recognition unit 1.
- Connect to 8. The character recognition unit 1-8 cuts out each character, extracts features, and performs character recognition. The following lines connect the buffer 1-6 output to the character recognition unit 1-8 and are read according to format control until mark 2-2 (mark A) is detected.

Next, when the mark detection sensor 1-10 detects the mark 2-2 (mark A), the image data of the row including the mark 2-2 (mark A) is taken into the buffer 1-6. The contents of buffer 1-6 at this time are shown in FIG. In the figure, 3 to 1 are image data corresponding to mark 2-2 (mark A).

When searching for a mark B indicating an illustration portion that is longer than the character pinch in the buffer l-6, it is found that there is a mark B having a length t with 3-2 as the starting point. Therefore, the left half is a character line, and the length is t to the right of the 3-2 position.
It is recognized that the portion 3-3 extending over is an illustration portion. Then, by the switching circuit 1-7, the left half of the contents of the buffer 1-6 is connected to the character recognition unit 1-8, and is recognized character by character, while the half of the cloth (to the right of 3-2) is connected to the character recognition unit 1-8. It is connected to the image processing section 1-9 and stores, processes, or deletes image data.

For the following lines, the above operation is repeated until mark 2-3 (mark A) is detected.

In other words, the character part (corresponding to 9-3 in Figure 9) is character recognition,
Image processing will be performed on the illustration section (corresponding to 9-2 in FIG. 9).

After that, the mark detection sensor 1-10 detects the mark 2-3 (
When the mark A) is detected, the character portion (corresponding to 9-4 in FIG. 9) is recognized as a character in the same way as when the mark 2-1 was detected as described above.

The reading of the form shown in FIG. 2 (FIG. 9) has been explained above, but according to the OCR of this embodiment, for the form shown in FIG. 10, marks as shown in FIG. A(
4-1, 4-2, 4-3), Mark B (4-4, 4-5)
, 4-6), the text section (10-1.10-4゜1O-5 in Figure 1O) and illustration section (No. 10
10-2*10-3.1O-6) in the figure can be easily distinguished and can be performed separately from character recognition and image processing.

(Effects of the Invention) As described above in detail, according to the present invention, for a form in which a text portion and an illustration portion are mixed, a first mark is attached to the vertical boundary between the text portion and the illustration portion, and a first mark is attached to the vertical boundary between the text portion and the illustration portion. Attach a second mark with a length equal to the width at the top of the
Since these marks are detected by the mark detection means and the reading of the text and illustration parts is controlled according to the detection results, it is no longer necessary to create format information each time the position, size, etc. of the illustration part differs. Therefore, format information can be simplified.

In addition, by detecting the second mark (mark B), the boundary between the text part and the illustration part in one line can be recognized, and the buffer output can be switched between the character recognition part and the image processing part according to the detection result. It becomes possible to perform image processing on data other than the literature department, that is, the illustration department.

Recently, it has become common practice to read printed materials and publications using OCR and create a database.These printed materials and publications contain many advertisements, illustrations, charts, etc., but the OCR of the present invention is Reading a property like this (( is very effective.

[Brief explanation of drawings]

FIG. 1 is a diagram showing the configuration of OCR according to an embodiment of the present invention, FIG. 2 is a diagram showing mark A and mark B written on the form in FIG. 9, and FIG. 3 is a diagram showing an example of the contents of the buffer. , FIG. 4 is a diagram with mark A and mark B written on the form in FIG. 10, FIG. 5 is a configuration diagram of a conventional OCR reading mechanism, and FIG.
The figure is a circuit configuration diagram of a conventional OCR, FIG. 7 is a diagram showing an example of a form including an illustration section, FIG. 8 is a diagram showing an example of writing a switching mark between a text section and an illustration section, and FIGS. 9 and 10. The figure shows an example of a form that cannot be controlled using the conventional method.0 1-1...1 form, 1-2...roller, 1-3...
・Lamp, 1-4... Image sensor, 1-5...
A/D conversion circuit, 1-6...buffer, 1-7...
Switching circuit, l-8...Character recognition unit, 1-9...Image processing unit, 1-10...Mark detection sensor, 1-
11...control unit.

Claims (1)

[Scope of Claims] A reading unit that transports a form, irradiates text and illustration areas on the form, and converts the reflected light into an electrical signal, and an A/D that converts the electrical signal from the reading unit into an A/D signal. a converting section; a buffer section that temporarily stores the output of the A/D converting section as image data; a character recognition section that performs character recognition for character portions of the image data in the buffer section; In an optical character reading device that has an image processing unit that performs image processing for illustrations, a first mark is placed on the vertical boundary between the text and illustrations on the form, and a mark is placed on the top of the illustrations to mark the width of the image. with a second mark of length equal to the first mark.
and a mark detecting means for detecting the second mark, controlling the document conveyance according to the detection result of the mark detecting means, and switching the output of the image data stored in the buffer section to the character recognition section and the image processing section. An optical character reader featuring: