JPS6324382A - Optical character reader - Google Patents

Abstract

PURPOSE:To exactly detect a character line regardless of the size of a price tag by detecting the character line by using plural lateral OR circuits. CONSTITUTION:Character codes are inputted by pressing the window part in front of a scanner against a blank 23 on which character codes are written. In such constitution, plural lateral OR circuits 1, 1'... are provided in each lateral OR extent. The lateral OR circuit makes the result of operation 1 when there is even one black picture element in lateral direction and makes the result of operation '0' when all picture elements are white. Black position and length detecting sections 2, 2'... are provided in each circuit 1, 1'.... The detecting sections 2, 2'... receive output of the circuits 1, 1'..., and judges whether the length of black output is proper as the height of character. A character position selecting and combing section 3 receives output from detecting sections 2, 2'..., and judges the positions where characters exist. Information of positions where characters exist is sent to a control circuit 29. The circuit 29 starts and stops a recognition processing section 30.

Description

[Detailed description of the invention] a) Industrial application field The present invention relates to an optical character reading device.

That is, the present invention relates to a device that optically reads characters by scanning a sheet of paper with characters written thereon using a hand-held scanner.

B) Conventional technology Supermarkets and department stores use POS (Po1nt o), which collects sales information for each item and manages inventory.
E531es) The system is popular.

In a POS system, an optical character reading recognition device (OCR) is used to read the code number and price of each item from a price tag at the time of sale.

For example, a hand-held OCR (Japanese Patent Application No. 6O-79082) as shown in FIG. 2 has been devised and is already in use.

In this hand-held OCR, characters, symbols, etc. written on the paper 23 can be read by simply holding the scanner 21 in the hand 22 and placing it on the paper 23.

The object to be read is information in a POS system, such as a price tag on which product number, price, etc. are written. Characters include numbers, symbols, and the alphabet. It doesn't read kanji or anything like that. The object is simply expressed as paper 23. When I say paper, I don't mean paper that has blanks on which you can write new letters or figures.

In front of the scanner 21 is a light source 24 . The light from the light source 24 hits the paper 23, is reflected, and enters the lens system 25. This is photoelectrically converted by the sensor 26.

The sensor 26 is an image sensor with two-dimensional expansion. Then, the sensor 26 detects the information writing part of the paper 23,
In other words, the field of view is approximately the same size as the price tag.

The light receiving section of the sensor 26 is composed of independent photoelectric conversion elements divided vertically and horizontally.

In each unit element, the intensity of light is converted into a change in photocurrent or photoresistance. This electrical signal is an analog quantity proportional to the light intensity, but is converted into a digital quantity by comparing it with a reference voltage or the like. Depending on the purpose, it may be converted to a digital quantity of higher order bits, but in many cases,
Divided into two stages: light and dark.

The portion photoelectrically converted by a unit element is called a pixel. A value of 0 or 1 is associated with each pixel divided into two levels, bright and dark. Alternatively, a pixel with a value of O is called a white pixel, and a pixel with a value of 1
A pixel with this is called a black pixel.

A pixel is the smallest unit area in image processing.

The number of pixels is determined by the sensor. The fact that the field of view of the sensor is the same as the information section of the price tag means that the images of all characters and symbols are focused somewhere on the sensor surface.

Information on price tags within the field of view can be read without moving the scanner 21 over the paper.

The control binarization circuit 27 is a circuit that binarizes the output signal of each pixel of the sensor 2G by comparing it with a certain threshold value. The binarized signals are sequentially sent to the image memory 28.

The image memory 28 stores binarized data for pixels covering almost the entire field of view of the sensor 26.

The binary data of the sensor 26 will be explained with reference to FIG.

The sensor has 29 pixels, consisting of P pixels vertically and 9 pixels horizontally. This is approximately equal to the field of view of the entire information writing portion of the paper 23.

The pixels are regularly arranged vertically and horizontally. A series of horizontal pixel columns is called a row. Bl, B2 in the row direction.

・・・・・・・・・It is coded as Bq.

A series of vertical pixel columns is called a column. Ll in the column direction,
B2 , . . . , I-p.

All pixels can be completely specified by row and column numbers. For example, a pixel in m rows and n columns can be written as (m, n).

Also, regarding the columns, Bl, B2,...
..., B9 can be adopted as the column notation. B1 refers to the entire first vertical pixel or pixel row.

Furthermore, to represent a row, Ll, B2,...
....

Lp can be used. Ll is a set of pixels in one row on the top side.

The image memory 28 stores a value of 0.1 for each pixel. Therefore, FIG. 3 can be considered to be the contents of the image memory 28. Data (0, 1) for each pixel can be stored by addressing the values of the matrix (m, n).

Numbers 1, 2.3, 4.5 and alphabet A.

B is illustrated. The parts forming these characters are black pixels, and the data value is 1. The blank space where no characters are formed is a white pixel, and the data value is 0.

The number "1" is between B1 to Bn and L1 to I-r+1;
There it is. In other words, the number 1 is represented by a pixel group of m rows and n columns.

Other numbers are also described by groups of pixels having the same area.

In FIG. 2, the recognition processing section 30 identifies characters and symbols one by one.

In the image memory, r- is scanned through all rows, then B2
Do not use rask-order scanning that scans all rows.

Instead, a group of pixels corresponding to -characters are collectively scanned from the image memory, data corresponding to -characters is temporarily stored in buffer register 8, and this is recognized.

That is, the recognition processing unit 30 retrieves one character's worth of data from the image memory.

The control circuit 29 has a processing capacity of the recognition processing section 30 (n
Data corresponding to 1xn)Iil element is read and stored in the buffer register 8. The buffer register 8 has mxn registers. This stored content is compared with the distribution of pixel values of characters and symbols known in advance to determine what the characters and symbols are.

However, even if (mxn) is a dimension that can express one character or symbol, the character may overlap the boundary of (mxn) depending on how one character data is extracted.

Therefore, while maintaining the size of the (mxn) pixel group, we will move it to the right one by one, or by two or three, and perform identification processing each time.

Such a pixel group (in X n ) corresponding to - characters is herein referred to as an identification window.

The identification window is shifted to the right by one pixel or a plurality of pixels to identify characters arranged in the horizontal direction.

After this, the identification window is moved downward, and then shifted to the right by one pixel (or several pixels) in the lateral force direction;
Identify the letters that line up. Repeat the same action below.

This will be explained more specifically.

In Figure 3, n pieces of data from B1 to Bn for row 51, n pieces of data from B1 to Bn for row B2, ..., n from 81 to Bn for row m. data is read from the image memory 28 and stored in the buffer register 8. That is, the data of the identification window defined by B1 to Bn and L1 to Lm is stored in the buffer register 8.
to go into.

In the identification processing circuit 12, in advance, (m'Xn) pixels,
Characteristics of characters and symbols are memorized.

Referring to this, it compares it with the characteristics of the input data and determines whether there is a match or a mismatch. If the input take matches any of the reference data, it means that the input data represents that character or symbol.

The principle of character recognition for identifying characters and symbols from the data stored in the buffer register 8 is as described above. This is a well-known method that has been adopted, for example, in Japanese Patent Publication No. 59-6418.

Next, slide the identification window one step to the right.

That is, for columns L1 to Lm, data of (mxn) pixel groups from column (B1+1) to column (Bn+1) is
The image is transferred from the image memory 28 to the buffer register 8.

Then, the identification processing circuit 12 again compares the characters and symbols with previously stored characters and symbols.

Furthermore, shift the identification window one position to the right.

That is, for the B1 column to the Lm column, data of (m×n) pixel groups from the (Bt+2) column to the (Bn+2) column are transferred from the image memory 28 to the buffer register 8.

Then, the same operation as described above is repeated in the identification processing circuit.

Thereafter, in the same manner, the identification window is shifted to the right.

That is, the process is repeated for rows L1 to Lm and column 89.

Through such scanning, characters and symbols for one horizontal line are recognized.

Here, the identification window is moved to the right by one pixel (mxn), but it may be moved by two or three pixels.

In this way, when the recognition of characters and symbols in one horizontal line is completed, the characters and symbols are shifted downward to the identification window.

This can be done by shifting the height of one character downward.

For example, if m lines are approximately twice the height of the characters, the identification window is shifted downward by m/2 lines. That is, for m rows Lm/2 to 3Lm/2, (mxn) data in columns BI to B2 are transferred to the buffer register 8, and identified by the identification processing circuit 12.

Next, for m rows of Lm/2 to 3 Lm/2, (B1+
(2) Column to (B2+1) Column data is identified by the buffer register 8 and identification processing circuit 12.

Repeat the same operation below.

The number of identification processes will be described.

If we shift the horizontal ζ identification window one pixel at a time, (q-
n+1) times of processing are required for one horizontal row. Assuming that P is a multiple of m/2 in the vertical direction, 2P/m processes are required for one vertical process.

Therefore, the number of identification processes is 2P(q-n+1) It is times.

Through the above processing, all characters and symbols included in the field of view of the sensor 26 can be read.

Punishment) Problems with conventional technology Conventional manual character reading devices have the following drawbacks.

Think about the action of shaking the scanner at the price tag.

The amount of time it takes to scan the price tag and read the text is short. I spend more time not touching the price tag with the scanner.

There is a lot of time spent moving the scanner from the desk to the price tag, or from the price tag to the desk, or simply leaving the scanner on the desk between readings. There is no need to perform recognition processing during this time.

During such a period, it is desirable to stop the recognition process. This will save power consumption.

Furthermore, if recognition is performed only when the character line of the price tag is completely within the field of view of the scanner sensor, there is no need to worry about obtaining incomplete recognition results.

If it were possible to recognize characters only when necessary, instead of performing recognition operations continuously, it would be superior in terms of power and reliability.

Furthermore, as shown in FIG. 4, each line may be recognized by holding a scanner and sliding it vertically over a price tag with characters written across multiple lines as shown by the arrows.

In such a case, there may or may not be a blank character string to be read when viewed horizontally. This changes over time.

The part where a character string exists is called a character area, and the part where a character string does not exist is called a blank area.

In the case of Figure 2, the scanner remains stationary for a while while touching the price tag, and during this time, the characters included in the entire field of view,
Read the symbol.

With the circuit configuration shown in Figure 2, it is difficult to read character strings arranged vertically like this, and even if it were possible,
It takes a long time. This is because the same operation is performed for blank areas as for character areas.

Therefore, it is desirable that the scanner itself be equipped with a function that can distinguish between character areas and blank areas.

Let us consider a means for detecting the presence or absence of a line of written characters within the visual field. In other words, we will consider a means for detecting whether a character area exists or does not exist within the visual field.

FIG. 5(a) shows an example where there are characters within the visual field.

The size of the field of view is 320 pixels in the row direction (horizontal direction) and 60 pixels in the column direction (vertical direction).

On the manuscript paper, the image pitch is 0.19m in both row and column directions.
m, and the field of view is 60.8 mm x 11.4
It is rrrm.

In FIG. 5ial, pixels are indicated by drawing thin lines on the four sides of the pixels. There are a few letters at the top, and some clear letters are lined up at the bottom.

There are no characters in the middle. In other words, there are only white pixels. We know that the letters are arranged horizontally. A character is a collection of black pixels.

Therefore, by checking whether there are black pixels in the horizontal direction (line direction), it is possible to discriminate whether a character exists or not.

However, since the characters are lined up in the row direction for a certain length, it is not necessary to check the entire area in the row direction for the presence of black pixels.

Therefore, a part of the pixels in the row direction is taken, and among the W pixels in the row direction, a row in which at least one is a black pixel and a row in which there is no black pixel are discriminated.

In this manner, the test range consisting of W pixels taken to detect the presence or absence of black pixels in each row is called the "horizontal OR range" by the present inventor.

The reason why it is limited to "horizontal" is that the presence or absence of black pixels is detected for pixel rows in the horizontal direction.

"OR" is used because it performs an OR operation. If at least one of W pixels arranged in the horizontal direction is black, its output is set to "1", and if all pixels are white, its output is set to rOJ. This is nothing but an OR operation.

After that, for each row of pixels within the horizontal OR range, is there any black pixel? Or, the operation to determine whether there is none at all,
In commercial terms, it is sometimes referred to as a "horizontal OR operation."

As an example, among the 320 pixels in the row direction, 100 pixels consisting of 50 pixels before and after the center are set as the horizontal OR range. That is, on the document surface where W=100, it is 19 mm in the horizontal direction.

The field of view extends 60 pixels vertically. Therefore, in this example, the horizontal OR range is a central rectangular area consisting of 100 x 60 pixels.

The 110 pixels before and after this area and the 110 pixels do not fall within the horizontal OR range.

If a character does not fall within the horizontal OR range in the sensor field of view, the character cannot be read.

The person holding and operating the scanner must be careful to place the scanner on a price tag, etc., so that the center of the character string and the center of the scanner are aligned. If the character string is longer than 110 pixels, it is sufficient that the characters fit within the scanner window.

In a certain part of characters, lines in which the result of the OR operation is 1 are continuous vertically. As shown in FIG. 5 fcl, the length of a line in which the result of vertically consecutive OR operations is 1 is assumed to be l. Here, there is a black pixel row with a length of 11 on the top side, and a black pixel row with a length of 12 near the bottom side. This corresponds to the presence of strings.

The height l of the set of black pixel rows corresponds to the height of the character. That is easily understood.

If the lengths of Jl and f2 are appropriate values for the height of a character, it can be said that the character is within the visual field.

The height l, however, is not always equal to the height of the characters. When the scanner is paying attention to the characters, l becomes large. If the line direction of the scanner is parallel to the direction in which the character lines are arranged, l will be approximately equal to the height of the character and will take the minimum value.

It is necessary to judge whether 11, 22, etc. are appropriate values for the character height.

A reasonable minimum value for the character height is when the character line is horizontal within the visual field, as shown in Fig. 6 f1).

A reasonable maximum value for the character height is when the character line is within the field of view at the maximum allowable slope, as shown in Figure 6 ib).

If the line direction of the text and field of view match, the text height is 2.4
Since fi, l is also approximately 2.4m+n.

The maximum inclination of characters and field of view is 8 degrees, but 19 tran
If a character line with a height of 2.4 is tilted by 8 degrees in the horizontal OR range of , l will be approximately 5.9 fi.

Then, the appropriate height for l is 2.4 cm to 5.9 B.
That's 12 to 31 pixels.

However, there are different types of price tags. It must be possible to read any price tag. This is horizontal O
This poses a serious problem in determining the R range.

An example of a price tag will be explained with reference to FIG.

Take JISB9551 POS price tags as an example.

The price tags of POS No. 41 and PO3IQ are illustrated.

FIG. 7 la shows the PO541 price tag. Width 70mm.

There are 35 vertical lines. There is a row of long numbers at mid-height.

At the bottom there is a row of numbers that are skewed to the right. The problem is that it is biased to the right and not in the center.

The existence or non-existence of the numbers at the bottom right must also be compared. Therefore, the lateral OR range must have an extension of 30.4 m. In FIG. 7 1al, the horizontally lined portion is the IJOR range. The number of pixels is 160.

This is 60 pixels from the 1981100 pixels illustrated in Figure 6.
% wider.

With such a wide horizontal OR range, the lower right number of pos 41 can also fall within the horizontal OR range, as shown in Figure 7 ic).

Figure 7 1al is the price tag of POS No. 10. 32 fins horizontally,
It is 25 pieces of sake. When the scanner is pushed back, Figure 7 1a
It becomes like l. The fortune on the price tag is 32M, and the width of the horizontal OR range is 30.4W.

If you press the scanner carefully, the horizontal OR range will not extend beyond the price tag as shown in Figure 7 (e+).

However, doing so is difficult. 1.6m
This is because there is only a margin of +n.

If the horizontal OR range extends beyond the price tag, the background of the price tag will become horizontal OR.
fall within range. The background is not necessarily white. Therefore, if there is a black part in the background, even if the part of the price tag is white, this will be determined to be a line containing black pixels. This causes malfunction.

In other words, the horizontal OR range must not exceed the price tag.

If the price tag is small, there is little margin, so the operator can
It's difficult to do something like this unless you're careful and skilled.

It is unreasonable to expect such skill from supermarket cashier employees, and there is no benefit to using optical reading devices if they are difficult to handle.

OBJECTS OF THE INVENTION The present invention overcomes the drawbacks of conventional hand-held optical character reading devices.

A first object of the present invention is to provide a practically excellent optical character reading device that can detect lines of written characters on both large and small price tags.

A second object of the present invention is to provide an optical character reading device that can improve the accuracy of character reading by accurately distinguishing between lines with characters and lines without characters.

A third object of the present invention is to provide an optical character reading device that can accurately discriminate lines without characters, stop unnecessary character recognition processing, and save power.

E) Structure of the Invention The present invention employs a plurality of lateral ORi ranges. multiple horizontal O's
The R range is used to detect character lines in the same way as before.

Since the horizontal OR range is effectively expanded in the horizontal direction, character lines can be detected even if the price tag is wide horizontally and the character writing area is biased to either side.

Furthermore, since there are a plurality of horizontal OR ranges that are independent in the horizontal direction, it becomes possible to distinguish between the background outside the price tag and the characters inside the price tag by referring to these results.

The present invention will be explained in detail with reference to FIG.

The process up to capturing the image from the image sensor 26 into the image memory 28 is the same as the conventional method.

The scanner 21 is held and operated in the hand. That is, the user presses the opening window in front of the scanner 21 against the paper 23 on which the characters and symbols are written, and inputs the characters and symbols.

The light source 24 illuminates the paper 23 in front. The light from the paper 23 enters a beam system 25 and forms an image of the paper 23 on a sensor 26.

The image sensor 26 is a two-dimensional image sensor having a large number of sensor units vertically and horizontally. Each sensor unit generates an electrical signal depending on the amount of light. The signal for each sensor unit is the smallest image information unit. In other words, this is pixel information.

The photoelectric conversion signal for each sensor unit is sent to the control binarization circuit 2γ.
It is binarized into either 0 or 1.

A white pixel is associated with 0, and a black pixel is associated with 1.

The image memory 28 stores brightness values (0, 1) for each pixel within the entire field of view of the image sensor.

The above configuration is no different from the conventional one.

In the present invention, a plurality of horizontal OR ranges are provided instead of a single one. This may be two, three, or more.

A plurality of horizontal OR circuits 1.1', . . . are provided for each horizontal OR range.

A horizontal OR circuit means that for pixels included in the horizontal OR range, if there is even one black pixel in a horizontal row, the calculation result is set to 1", and if there are only white pixels, the calculation result is set to "0". This is a device that does this.

A line with at least one black pixel is called a character line.

This has already been explained by FIG. 5 ic). 11°!
! 2 represents the vertical length of a line containing at least one black pixel, that is, a character line. This is approximately equal to the height of the letters.

Even if the horizontal OR operation result is 1, this can be changed to 1 depending on the character.
It doesn't necessarily mean that it is. As already mentioned, horizontal, O
For the length l at which the R operation is 1, there is a valid range for it to be a character.

Furthermore, regarding the position of a line containing a black pixel, there is a valid range in order for this to be a black pixel caused by a character.

It is necessary to distinguish these. For this reason, the black position/length detection sections 2.2', . . . constitute the horizontal OR circuit 1.1'.

It is provided for each...

Horizontal OR circuit 1. In response to the output of V,..., the length of the kiln output is increased (Fig. 5 fc) and l! It is the function of the black position/length detection unit 2.2', . . . to determine whether or not x, 12) is appropriate as a character height. Such determination processing is called "character line detection processing."

The character position selection/combining unit 3 includes a black position/length detection unit 2.2.
', . . ., the information on the length 11 positions of the kiln output and the information on the character line detection results are received, and the position where the character exists is determined by selection or combination.

Information about the position where the character is present is sent to the control circuit 29.

The control circuit 29 performs processing such as starting and stopping the recognition processing section 30.

The present invention provides multiple lateral OR ranges and multiple lateral OR circuits. Here are the characteristics.

An example of the horizontal OR range will be explained with reference to FIG.

FIG. 8 ta) shows the field of view of the image sensor, which has two lateral OR ranges, and corresponds to the document surface.

The horizontal direction is 60.8 mm and the vertical direction is 11.4 rrrm. Pixel dimensions are sail 19 mm x 0-19 wn
Therefore, it has a horizontally long field of view of 320 pixels and 60 pixels.

This point is the same as that in FIG.

Width 8.55 mm for lateral viewing line 60.8vn
Horizontal ORi surrounds RL and R2 are provided with an interval of 9.5 wm. A horizontal line is attached to indicate the horizontal OR range.

One horizontal OR range is 8.55 mm x 11.4
It is rrrm.

In terms of the number of strokes, it is 45 x 60 pixels.

Since there are two equivalent horizontal OR ranges, the effective width is 2.
6.6++ on, there are 140 pixels.

The meaning of having two is not that they increase the effective range, but that they exist independently.

Figure 8 ib) is above the small price tag, PO510.
It is an explanatory diagram when a scanner is applied.

Both lateral OR ranges R1 and R2 easily enter the price tag. In this case, character lines can be detected in both R1 and R2. It is also unaffected by background factors outside of the price tag.

In the case of Figure 8 +C), fd), R1 protrudes from the price tag. This kind of thing is common.

Pairs are influenced by the background.

However, R2 remains inside the price tag and is capable of detecting lines of text. R2 is independent of R1.

Even if one of them protrudes, as long as the other remains within the price tag, the character line can be detected using either one.

Furthermore, for PO540, which has a large price tag, the arrangement is as shown in Figure 8+e). There is no problem with long numbers in the middle. The problem was that the number digits were skewed towards the bottom right. This character (number) line also falls within the right horizontal OR range R2. R2 can detect character lines. R1 does not detect character lines, but R2 can.

The process of selecting and combining character line positions in the character position selection and combination unit 3 will be explained with reference to FIGS. 9 (fat to td).

In each partial diagram, the field of view of the sensor on the document surface is shown on the left. On the right side, the results of the horizontal OR operation of R1 and the results of the horizontal OR operation of R2 are shown. The reason why I wrote "white" is that the result of the horizontal k operation is °゛0'', which means that there are only white pixels in that row.The reason why I wrote "black" is because the result of the horizontal OR operation is "1".゛′
That is. This means that there is at least one black pixel in that row.

The right end shows the results of selection and combination processing. (
al~fd) are as follows.

(a) Neither Rt nor R2 detects a character line.

The result is that there are no lines of text.

ib) Only one of Rt and R2 detected a character line. In this case, the detection result is used as it is as the character range.

Since we have the result of R2 here, we will use it as is.

[ci Both Rx and R2 detected a character line, but
Character ranges have no intersection.

In this case, the upper range of characters in the column direction is adopted and the lower range is discarded.

Here, R1 is a row with black pixels from height hl to h2, and R2 is a row with black pixels from height h3 to h4. h2 (h3.

The results are taken only for h1 to h2 of R1. The R2 portion is thrown away.

This does not mean that the line "ABC..." is not read. The scanner scans from top to bottom, so when it's in this state (well, ``123'' at the bottom...
This means that only the lines with "・" are read.

When the scanner moves further downward, "123...
・'' disappears and the line ``ABC......'' moves up within the field of view, so at this time 1-ABC......
” line will be read.

(d) Both R1 and R2 detect a character line, and there is a common part in the character range.

The sum of both is the character range of the entire visual field.

This means that the letters are tilted relative to the field of view. Although they are tilted, they exist continuously from hl to h4, so the sum of both is taken to define the character range from h1 to h4.

(For C1, when there are two character lines, the upper character range is selected preferentially.

However, depending on the usage pattern, there may be a configuration in which priority is given to a character range further down, or a configuration in which priority is given to a character range near the center in the column direction.

For example, some scanners have a large vertical width so that when placing the scanner on a document with text written on it, the position can be moved slightly up or down.

FIG. 10 shows a schematic perspective view of such a scanner. This has a vertical opening dimension of 20ran instead of 11.4m.

The operator places the scanner on the document surface so that the character line is in the center of the field of view (Fig. 10 (b) A), but since the field of view is wide in the vertical direction, even if it is slightly shifted vertically (Fig. ib) A) Can be read.

In this case, a character line that the operator intends to read exists in a character range near the center of the field of view in the column direction (vertical direction).

When using such a scanner, priority may be given to a character range near the center in the column direction.

Other configuration examples In the example shown in Fig. 1, multiple horizontal OR ranges R1, R2...
Horizontal OR circuit 1.1', . . .
. . . and black position/length detection section 2.2', . . .
and.

This has the advantage that the time required for processing is short because horizontal OR calculations and black position/length detection can be performed in a plurality of horizontal OR ranges at the same time.

However, on the other hand, there is some overlap in circuit configuration. This may not be desirable from an economic point of view.

FIG. 11 shows another configuration example of the present invention.

There is only one horizontal OR circuit 1. This makes horizontal OR calculations for a plurality of horizontal OR ranges sequential.

There is also only one black position/length detection section 2.

A new horizontal OR range control circuit 4 and memory 6 are provided.

Setting of a plurality of lateral OR ranges is performed in advance in the lateral OR range control circuit 4.

The lateral OR range control circuit 4 has lateral OR ranges R1 and R2.

A horizontal OR range command is issued to the horizontal OR circuit 1 to perform horizontal OR calculations one after another for .

Based on the results of each horizontal OR operation, the position of the character line is determined by the black position/length detection section 2. This is stored in memory 6.

The memory sequentially stores character line ranges corresponding to R1, R2, . . . in synchronization with the horizontal OR range command from the horizontal OR range control circuit 4.

The character range selection combination unit 3 selects all horizontal OR ranges Rt,
The character line range for R2, .

The processing after the character line range selection and combination section 3 is the same as in the example shown in FIG.

G) Other Examples of Processing Procedures In the examples described so far, all the outputs of the image sensor 26 are stored in the image memory 28, and then the horizontal OR operation is performed.

However, it is not limited to this.

The output signal from the binarization circuit 27 that drives the image sensor 26 and binarizes its output is passed to the horizontal OR circuit 1.1'.
, . . . and then storing them in the image memory 28, it is also possible to detect character lines in parallel.

In this way, the time required for character line detection processing and image memory input is shortened.

(ri) Effects (1) Whether the price tag is large or small,
Character lines can be detected accurately.

(2) Since the presence or absence of a character line is detected, unnecessary recognition processing can be stopped for blank areas. This makes it possible to save power.

(3) By detecting whether a character line has completely entered the field of view, the risk of obtaining incomplete recognition results can be reduced.

[Brief explanation of drawings]

FIG. 1 is a configuration diagram of an optical reading device of the present invention. FIG. 2 is a configuration diagram of a conventional optical reading device. FIG. 3 is an explanatory diagram of binarized data divided into pixels and binarized by an image sensor. The vertical and horizontal grids represent pixels. FIG. 4 is an explanatory diagram of the operation of scanning the scanner from the top to the bottom of the paper. FIG. 5 is an explanatory diagram of two-dimensional visual perspiration of an image sensor. f1
+ represents a pixel in NI glands, where the horizontal direction is the row direction and the vertical direction is the column direction. fbl is a diagram showing a horizontal OR range set in the center of the field of view. tc) is the horizontal OR for the fat example.
The figure which shows the result of performing calculation. FIG. 6 is an example showing the inclination of the character lines of a price tag in the field of view of the image sensor. fa) shows an example in which the horizontal direction of the image sensor and the character line of the price tag are parallel. fbl shows an example in which the character line of the price tag is tilted with respect to the field of view of the image sensor. FIG. 7 is a diagram illustrating an example of the positional relationship between the horizontal OR range of a conventional optical reading device and a price tag. fa) is an example of a large price tag and shows POS No. 41. fb) is a diagram showing the horizontal OR range in the field of view of a conventional image sensor. fcl is po
A diagram showing an example in which a conventional image sensor field of view is superimposed on the price tag of S41. fdi is an example of a small price tag and PO51
Indicates No. Q. (e) is a diagram showing an example in which the field of view of a conventional image sensor is superimposed on the price tag of PO5IQ. FIG. 8 is an explanatory diagram showing an example of correspondence between the image sensor field of view, the horizontal OK range, and the price tag when a plurality of horizontal OR ranges are provided in the invention. fat is 8.55mm, one on the left and right
An example of a field of view in which the width of the lateral OR range is separated by 9.5 mm. fbl is an explanatory diagram in which the field of view of the image sensor is superimposed on the price tag of pO310, which is a small price tag. (C) (MaPO5I Q
An explanatory diagram illustrating an example in which a character line overlaps in one horizontal OR range even when the image sensor field of view is stacked on a number price tag with an inclined view. fd) is an explanatory diagram showing an example in which a character line overlaps in one horizontal OR range even when the image sensor field of view is tilted and stacked on the price tag of No. I'05IQ. (el is an explanatory diagram showing that when the image sensor field of view is superimposed on PO541, which is a large price tag, the characters that are biased toward the lower right of the price tag will also fall within one horizontal OR range. Figure 9 is the character position selection combination section Figure 10 is an explanatory diagram of a scanner with a vertically long field of view. (a) is a schematic perspective view of the scanner, and (1)) is an example where the opening window of the scanner is tilted with respect to the characters on the price tag. Diagram showing. FIG. 11 is a diagram showing another configuration example of the present invention. 1.1'... Horizontal OR circuit 2.2'... Black position/length detection section 3...
・Character position selection combination unit 4...Horizontal OR range control circuit 6...Memory 8...Buffer register 12...Identification processing circuit 21... ... Scanner 22 ... Hand 23 ... Paper 24 ... Light source 25 ... Lens system 26 ... Image sensor 27 ... ... Control binarization circuit 28 ... Image memory 29 ... Control circuit 30 ... Recognition processing section Inventor: Miki Yamaguchi
Yuui Hiroshi Fuji Patent applicant Sumitomo Electric Industries Co., Ltd. Application agent Patent attorney Shigeki Kawa, 1 circumferential structure
3 Figure 4 Figure 10 Figure □60. Purification: Figure 6 Figure 7 PO341 Hee 70-1 ← posto No. 11 32--H No. 8 (a) +b) POS No. 10 Figure (c) ■
) RI R2 RI R2 results Figure 9 (C)

Claims (5)

[Claims] (1) A light source that illuminates a sheet of paper on which characters, symbols, etc. are written horizontally, and a light source that illuminates paper on which characters, symbols, etc. are written horizontally, and a plurality of characters, symbols, etc. that have a large number of photoelectric conversion elements arranged in a plane extending in the horizontal and vertical directions. An image sensor with a field of view of It has a built-in binarization circuit that converts the output from the photoelectric conversion element into a binary signal of black pixels and white pixels through digitization processing, a light source, an image sensor, a lens system, and a binarization circuit. In an optical character reading device, which consists of a scanner that is operated by holding it and an identification processing circuit that identifies characters and symbols from binarized data, the field of view of the image sensor is separated by two columns. A plurality of horizontal OR ranges R_1, R_2, ...... are set, and in the horizontal OR ranges R_1, R_2, ......
Horizontal OR ranges R_1, R_2, for performing a horizontal OR operation in which the output is 1 when there is one or more pixels among the pixels arranged in the row direction, and the output is 0 when there is no black pixel.
...... or the entire lateral OR range R_1, R_2
, ......, and the length of the column in which the output of the horizontal OR operation that is continuous in the vertical direction after receiving the output of the horizontal OR circuit is 1 is within a certain range. If so, horizontal OR range R_1, R_2, to determine that there is a character line.
・・・・・・・・・ or total lateral OR range R_1, R
It has a black position/length detecting section that is provided in common for _2,..., and all horizontal OR ranges R_1, R.
＿２、・・・・・・Detects the presence or absence of a character line, and performs a sum operation or selects the character line detection results in the horizontal OR range where the character line exists, and thereby An optical character reading device characterized by determining the position of a character line on a sheet of paper and performing recognition processing. (2) When the character line detection results of a plurality of horizontal OR ranges overlap in the column direction, these character lines are summed to form a character range for the entire field of view. 1) The optical character reading device described in item 1). (3) If the character line detection results in a plurality of horizontal OR ranges do not overlap in the column direction, the character line detection result located above in the column direction is set as the character range for the entire visual field. The optical character reading device according to item (1). (4) If the character line detection results in a plurality of horizontal OR ranges do not overlap in the column direction, the character line detection result located lower in the column direction is set as the character range for the entire visual field. The optical character reading device according to scope (1). (5) If the character line detection results in a plurality of horizontal OR ranges do not overlap in the column direction, the character line detection result closest to the center in the column direction is set as the character range for the entire visual field. The optical character reading device according to item (1).