JPH01191280A - Kanji recognizing system - Google Patents

Abstract

PURPOSE:To correctly recognize a KANJI (Chinese character) in a short time by alternately disposing black points and white points on the upper side and the left side of an information element, disposing the white points on a lower side and a right side and reading a code on which a point representing information is disposed by an image scanner. CONSTITUTION:In order to obtain the entry position of the information element, the positions Xa, Ya,...Xd, Yd of the points (a), (b), (c), (d) of the four corners of an external frame are obtained to obtain the coordinates of the points 11, 12, 21, 22 of the reference of four bits. The widths of the external frame are respectively subtracted from the distance of the X axis direction of the point (a) and the point (c), and the point (b) and the point (d) and averaged to have the length of the X axis direction occupied by a piece of information and they are two columns, so that they are divided equally into two to obtain the width Xs of the piece of information. On a Y axis, Ys is similarly obtained. The width of the information obtained in such a way is added to the coordinate Xa, Ya of an origin, thereby, the coordinate of the reference of the information is obtained. However, when the number of the bits of a the information element is increased, the position detecting error of the information element is increased. Then, as one part of the external frame, the black points and the white points are alternately disposed to measure the position of the black points, thereby, more correctly obtaining the position of the information.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a kanji recognition method for recognizing encoded kanji.

[Conventional technology]

In computer-based technology, the time required for input is
This accounts for a very large proportion of input data, and various methods have been developed to shorten the input time.

One method is to input a printed code into a computer using a reading device. These include barcodes and Carla codes. A barcode is a code made up of thin lines and thick lines, and is suitable for reading a small number of alphanumeric characters with a linear scanner.

Also, as stated in the evening edition of the Asahi Shimbun on January 16, 1988, the Karura code is created by making the four squares in the character 田 in Figure 8 either white or black. It represents a bit signal.

However, the Carla code, like a television camera,
This code has pixels of approximately 50 x 250 dots, and is suitable for being read with a scanner that scans a surface, and is suitable for reading not only alphanumeric characters but also several kanji characters on one screen.

By the way, recently, image scanners have become popular,
The number of pixels is approximately 3000 x 5000 dots, which is an order of magnitude higher than that of a television camera.

[Problem to be solved by the invention]

Image scanners also scan a surface, so it is more suitable to use Karura codes than barcodes, but when reading Karra codes with image scanners, Karra codes are also wasteful in terms of reading speed and recording density. There is room for improvement.

Furthermore, Japanese Patent Laid-Open No. 60-27074 discloses an information reading device that uses a pair of black and white to indicate 1 or 0 depending on the order of the pairs, but according to this, the recording density must be halved. do not have.

Here, the procedure for recognizing the Carla code is as follows.

1) Cutting out the information element 2) Determining the information bit position 3) Black/white determination And, of the above procedures, 1) and 2) are performed for each information element, so it is difficult to use the Carla code where the information element is as small as 4 bits. In this case, if this is read by a device that reads a large amount of information, such as an image scanner, the reading time will be long.

Further, blank spaces between codes are necessary for operations such as cutting out information elements, but the waste caused by these blank spaces is also reduced when the number of bits of information elements is large. In other words, codes with larger information elements are more suitable for reading with an image scanner.

On the other hand, the information element of the Carla code is 2×2 4 bits, but the number of bits can be increased by increasing the number of rows and columns.

However, as the number of rows and columns increases, a problem arises in that the accuracy of detecting the position of information elements decreases. Sunarichi,
In the case of the Carla code, the extracted information element is divided into four equal parts vertically and horizontally, and it is determined whether each part is black or white.The more the number of parts is divided, the greater the error in the detected position becomes. Become. Note that it is desirable that the number of bit strings of the information element be 16, since Kanji characters are 16 bits. Regarding the number of rows, there is no particularly desirable number of rows, but it would be appropriate to set it to 16 rows including the number of columns.

If an information element of 16 rows and 16 columns is cut out and divided into 16 equal parts vertically and horizontally, the position detection accuracy of the information element will be considerably reduced.

By the way, when considering the position detection accuracy of information elements, the question is how many pixels one bit consists of.

Furthermore, it is necessary to assume that the code is printed using a printer attached to the same device as the image scanner. That is, generally, a printer and an image scanner attached to the same device have the same resolution. because.

This is because by doing so, the image read by the image scanner can be printed as is by the printer in the same size.

Let us now consider the minimum number of pixels that need to be allocated to one bit under the conditions described above.

First, it is assumed that one image is assigned to one bit, as shown in FIG. 9(a). In this case, the image scanner sensor, which is the same size as the pixel, determines whether it is black or white, but the pixel position and the sensor position rarely match exactly, and in the worst case, the intersection of the sensors may It may be in the center of Since the sensor is adjusted to determine white when half or more of its area is white, in the case of FIG. 9(a), all points are determined to be white.

Next, even if two pixels are assigned to one bit as shown in FIG. 9(b), in the worst case, half of the black will be applied to the two sensors, and therefore, as shown in FIG. ), there is a possibility that all points will be determined to be white depending on the characteristics of the sensor.

Furthermore, as shown in FIG. 9(Q), if four pixels are assigned to one bit, even in the worst case, one point will be determined to be black.

Therefore, 4 pixels are assigned to 1 bit, and the sensor
If even a point is determined to be black, it is logically possible to detect the position of the information element, but if you do this,
Even the slightest scratch can cause false detection, and there is also a possibility that it will be misjudged as black due to the influence of adjacent black pixels, so it is inaccurate.

Furthermore, as shown in FIG. 9(d), if 9 pixels of 3×3 are assigned to 1 bit, even in the worst case, 4 sensors will determine black. Therefore, if the method shown in FIG. 9(d) is used, the position detection accuracy of the information element will be improved to some extent, but
Four of the nine dots are less than half of all dots, and considering the influence of adjacent black pixels, if four or more dots are black, there is a possibility that they will be misjudged as black. growing.

On the other hand, as shown in FIG. 9(e), 1 bit has 1
If 6 pixels are allocated, at worst, 9 points will be determined to be black. Since 9 points is a majority of 16 points, if 9 points or more are black, 7 points or less are white, and 8 points are unknown, the position detection error of the information element will not be biased towards white or black. For this reason, at least 4×4 16 pixels should be allocated to one bit.

However, when 16 pixels are assigned to 1 bit as shown in Figure 9(e) with the aim of maximizing recording density, the positional accuracy of information elements becomes a very important issue, and A positional error in one column can cause a misjudgment.

The present invention has been made in consideration of the above points, and includes:
The purpose is to use 16x16 in one information element.
Bit information, that is, a code that can represent information in 16 Kanji characters, can be recognized accurately in a short time by increasing the recording density using a device that reads a large amount of information, such as an image scanner. The objective is to provide a kanji recognition method that allows for

[Means to solve the problem]

In order to achieve the above object, the kanji recognition method according to the present invention alternately arranges black dots and white dots on the top and left sides of information elements,
It was characterized by having white dots on the bottom and right sides, and a code containing dots representing information that was read by an image scanner.

[Effect]

Here, regarding how to find the location where information is written,
An explanation will be given by taking as an example a case where the information element is 2×2 bits like a Carla code. Note that, as shown in FIG. 10, the coordinates are taken in the horizontal direction on the X-axis with the left end as the origin, and in the vertical direction on the Y-axis with the upper end as the origin.

Therefore, in order to find the entry position of the information element shown in Fig. 10, the four corner points a, b, c, de positions A a g
Y a H...Determine Xd, Yd, and from this value, 4
The reference point for each bit information (upper left point) 11.12
, 21. Find the coordinates of 22. Note that the outer frame needs to be so that the line does not break when read with an image scanner, and for this purpose it needs a width of 2 bits or more, so here it is assumed to be 2 bits wide. In addition, a 2-bit wide white line is placed inside the frame to separate the frame and the information. Furthermore, the information is assumed to be 4×4 dots per bit. The coordinates of the reference point of the information in this case can be calculated as follows.

That is, in Fig. 10, point a, point C1, and point d
The length of the information in the X-ray direction is obtained by subtracting the width of the outer frame (including the white line) from the distance in the X-ray direction. Find the width Xs of the information. Ys is similarly determined for the Y axis.

2×2 X2 The width of the information obtained in this way is the coordinate of the origin
Find the coordinates of the information reference by adding to .

Xnm=Xa+n・Xs, Ynm=Ya+m Emperor Ys −
(3) However, if the number of bits of the information element is increased while adopting such a method, the position detection error of the information element will increase, and when it reaches about 16 x 16 bits, the position detection accuracy of the information element described above will increase. As a result, good values cannot be obtained.

Note that the position detection error of the information element as described above occurs because there are few detection points that serve as position detection references, and the position detection accuracy can be improved by increasing the number of reference detection points.

Therefore, in the present invention, by placing black points and white points alternately as part of the outer frame and measuring the position of the black points,
This allows the location of information to be determined more accurately.

〔Example〕

Hereinafter, the present invention will be described based on one embodiment of FIGS. 1 to 6. FIG. 1 shows information element codes used in the kanji recognition system of the present invention.

In FIG. 1, ■ is a black dot, the mouth is a white dot, and O is 1-bit information; when it is 1, it is ■, and when it is 0, it is a mouth. a, b, c
, d is the information element, and O is 16
Since X16 pieces are lined up, it can be seen that the amount of information is 16x16 bits. Furthermore, instead of an outer frame, black dots and white dots are alternately arranged on the top and left sides of the information element.

Note that there are white dots on the bottom and right sides of the information element, and these white dots indicate boundaries between adjacent information elements.

FIG. 2 is a block diagram showing the overall configuration of a kanji recognition device used in the method of the present invention.

In FIG. 2, when printing a kanji code, the control unit 102 prints the kanji code stored in the code section 103-2 of the memory 103 in the memory program section 103-2.
1, a pattern corresponding to the kanji code is read out by executing a program stored in the printer 101, and a pattern corresponding to the kanji code is generated on the image memory 101.
5 to print.

On the other hand, when recognizing a kanji code, the image scanner 100 detects the code recorded on paper, and the communication unit 104 detects the code recorded on paper.
The image is transferred to the image memory 101 and stored. Then, the control unit 102 transfers this code to the program unit 10.
The recognition is performed by executing the program stored in 3-1, and the result is stored in the code section 103-2.

Here, to explain in more detail the procedure for printing a code using the kanji recognition device shown in Figure 2, in the conventional method, when printing kanji, the kanji code is read from the memory code section, and from this kanji code. , find the address of the kanji ROM that stores the kanji pattern, and write the kanji R.
Printing is performed by reading out the pattern stored in the OM and transferring it to the printer 105. On the other hand, when printing a code using the kanji recognition device shown in FIG.

Since the pattern is determined by the Kanji code, there is no need for anything equivalent to the conventional Kanji ROM.

Figures 3 to 5 show flowcharts for generating a pattern directly corresponding to a kanji code from a kanji code. In Figures 3 to 5, the information element is
As shown in the figure, it is assumed that the information is composed of 16x16 bits of information, and each information bit is composed of 4x4 dot pixels. In addition, information elements, including frames, (
16+2)x4=72 dots, i.e. 72 dots x
It has a 72-dot configuration, and this capacity is secured in the image memory 101 shown in FIG.
, Y coordinates 0 to 71, and the state of each point is I (X+
y), 0 is a white point, 1 is a black point, and the mouth (xt
wl: 7
This indicates that all points in columns Xi to X2 in rows 1 to 12 are set to 1.

In Figure 3, when recognizing a kanji code, first,
Clear all codes to 0 and create an outer frame pattern.

The flowchart is shown in FIG. 4 as step 200, and eight black information bits are placed on each of the upper and left sides of the code.

Returning to Figure 3, next select the kanji code to be printed.
This is read and converted as a pattern for one column.

A flowchart for reading one Kanji code to be printed and converting it into a pattern for one column is shown in FIG. 5 as step 300.

Returning to FIG. 3, next, one kanji code is read again and converted into a pattern for one column. Then, repeat the steps described above for 16 columns to create a pattern for one information element, and print the pattern created as described above to the printer 105.
You can print the required code by sending it to .

FIG. 6 is a flowchart for reading Kanji codes.

In order to read the Kanji code, it is necessary to find the position of each piece of information, which is found by reading the coordinates of the upper left corner of each black dot on the top and left sides of the code.

In FIG. 6, when reading a Kanji code, first search for point a in FIG. 1, that is, the bit in the upper left corner of the code. That is, in the range of 4 dots x 4 dots, point a is determined as the leftmost and topmost point where at least 9 points are black. Next, using this as a reference, move to the right while searching for eight black dots on the upper side of the code, and find the coordinates of their upper left corner. In addition, the coordinates of the upper left corner of the black point on the upper side of the code are
Since n HO+ yn HO (in this case, n is an even number) and there are black points every other black point, the coordinates of the upper left corner of the white point are determined from the midpoint as follows.

X　nl　o　Xn”2..

X n −I HO=□ (4) Next, on the upper side of the code, calculate the difference in coordinates with the point on the right and create a table.

d Xn= Xo, o -X, -t, o
...(5)d Yn" Yn) OYn-++
O・-(6) When there is no error, the above dXn is 4, dY
n should be 0.

Then, the coordinate difference with the point on the right side obtained in this way is added to the coordinates of the point on the left side, and the coordinates of each information point are obtained one column at a time.

Xnt, = xn-t, m + dXn
・++ (7) Yn, , = Yn-1, m+d Y
n =・(8) The start of each column is the coordinates of the bits on the left side, XO1m)Yo, lI. Find by. Furthermore, the coordinates of the white point are detected using a method similar to equation (4).

Therefore, 1 of each bit whose position was determined as described above.
1 if there is more black in the 6x16 dot pixels, O if there is less
If you divide each column of 16-digit binary numbers into 4-digit numbers and convert them into decimal numbers, you will get a kanji code, which is shown in Figure 2. is recorded in the memory code section 103-2.

By the way, when reading a kanji code, the first task is to find the black dot at point a in Figure 1, which is the first black dot of the information element, but the code is usually read diagonally with an image scanner. Therefore, it tends to be difficult to find the black point at point a in Figure 1. Furthermore, if the code is slanted to a degree that makes it difficult for an image scanner to read the code, it becomes difficult to read the code, and it tends to be difficult to determine whether the degree of slant is within a tolerance value.

However, in order to deal with the problems mentioned above,
As shown in Figure 7, lines are drawn at the top and left of the kanji code, and these lines make it easy to find point a, the first black point of an information element, and to determine the allowable diagonal value. Become.

〔Effect of the invention〕

The present invention is as described above, and according to the present invention, black dots and white dots are placed alternately as part of the outer frame of the kanji code,
16 in one information element by measuring the position of the sunspot
X16-bit information, that is, a code that can represent information in 16 kanji characters, can be recognized accurately in a short period of time by increasing the recording density using a device that reads large amounts of information, such as an image scanner. can do.

[Brief explanation of the drawing]

1 to 6 show an embodiment of the present invention, FIG. 1 is a diagram showing information element codes used in the kanji recognition method of the present invention, and FIG. 2 is a kanji recognition device used in the kanji recognition method of the present invention. Figures 3 to 5 are flowcharts for printing information element codes, Figure 6 is a flowchart for reading information element codes, and Figure 7 is a block diagram showing the information used in the method of the present invention. A diagram showing an application example of the element code, FIG. 8 a diagram showing the conventional Carla code, and FIG. 9 a diagram showing the position of the image scanner sensor and the determination result.
Figure 10 is a 2x2 code like the Carla code shown in Figure 8.
FIG. 3 is a diagram illustrating a method for finding information positions for codes whose information elements are bits. a = d...points at the four corners of the information element, gourd...black dots,
Mouth: white dot, O: 1 bit of information, 100...
Image scanner, 101... Image memory, 10
2...Control unit, 103...Memory, 103-1...
-Program section, 103-2...Code section, 104-
...Communication Department, 105...Printer. 7'4.・

Claims (1)

[Claims] 1. A code in which black dots and white dots are arranged alternately on the upper and left sides of an information element, white dots are arranged on the lower and right sides, and dots representing information are arranged within these is scanned by an image scanner. A kanji recognition method that is characterized by reading. 2. The kanji recognition method according to the invention described in claim 1, wherein the number of lines of information in the information element is a multiple of eight. 3. The kanji recognition method according to the invention described in claim 2, wherein the number of columns of information in the information element is a multiple of eight. 4. In the invention set forth in claim 1, a kanji recognition method in which lines are drawn on the top and left side of the collection of information elements to detect inclinations and reference points.