JPS59177686A - Character recognizing device - Google Patents

Abstract

PURPOSE:To improve the precision of character recognition to make the character recognition processing efficient by scanning an unknown pattern from directions opposite to each other to detect the recessed parts of this pattern and comparing positional relations among respective recessed parts to discriminate the unknown character. CONSTITUTION:The unknown character recorded on a form 1 is read optically by a camera 2, and the optical pattern (unknown pattern) of the unknown character is stored in a picture memory 4. A CPU5 executes the recognition processing of the unknown pattern on a basis of a character recognizing program stored in an ROM6. A picture 8 of the picture memory 4 is scanned from the left to the right to detect the left outline of the unknown pattern, and a left recessed part L is detected. Next, the picture 8 of the picture memory 4 is scanned from the right to the left to detect the right outline of the unknown pattern, and a right recessed part R is extracted.

Description

[Detailed Description of the Invention] <Technical Field of the Invention> The present invention relates to optical patterns such as unknown characters and symbols (hereinafter referred to as "
The present invention relates to a character recognition device that reads and recognizes unknown patterns (referred to as "unknown patterns"), and relates to a character recognition device that is used to identify specific unknown characters that are typical of Japanese ships.

<Background of the Invention> Conventional character recognition devices employ a method of identifying unknown characters by, for example, extracting characteristic parts of characters from unknown patterns. In this type of method, several candidate characters may be extracted from the unknown characters, and in such a case, it is necessary to specify one more character from these candidate characters. However, an example is the number "2" and the alphabet "S".
Especially when the font looks like a Japanese ship, it is difficult to narrow down the candidate characters to just one, which causes a drop in character recognition accuracy.

<Object of the Invention> The present invention improves the character recognition accuracy of this type of device by providing a new character recognition device that can distinguish between unknown characters typical of Japanese ships, and improves character recognition processing. The purpose is to improve efficiency.

<Configuration and Effects of the Invention> In order to achieve the above object, the present invention scans the unknown pattern from opposite directions to detect the recesses of the unknown pattern, and also identifies unknown characters by comparing the positional relationship of each recess. I decided to do so.

According to the present invention, for example, the number ``2'' and the English letter ``S'' have different positions of the recesses when viewed from the left and right, so these unknown letters and shoes that are typical of Japanese ships can be easily identified. This invention achieves excellent effects such as improving character recognition accuracy and streamlining character recognition processing.

<Description of Embodiments> FIG. 1 shows an example of a circuit configuration of a character recognition device according to the present invention. In the illustrated example, an unknown character recorded on a sheet of paper 1 is optically read by a crinkle 2, converted into digital data by an A/D converter 3, and then the optical pattern (unknown pattern) of the unknown character is stored in an image memory. It is stored in 4. Also, CPU (CenLral Processing
Unit) 5 is ROM (Read Only M
memory) 6 Based on the character recognition program stored in fin, RAM (Random A,
Figure 2 shows the principle of the character recognition method according to the present invention, and the unknown pattern of the number "2" (shown in Figure 2 (1)) , the alphabetic letter “S” and the number “
5" each unknown pattern (shown in Figure 2 (2) and (3))
are compared on the XY coordinates. Each unknown pattern in the illustrated example has a concave portion facing left and a concave portion facing right, but the positional relationship of each concave portion L and R in the character pattern “2” and the character patterns “S” and “5” is is upside down. Therefore, the deepest point A (Xa, Y
a), 13(Xb, Yb), Ya)Yb
, the unknown pattern is "2" and Y b ) Y
In case a, the unknown pattern can be recognized as "S" or "5".

FIG. 3 specifically shows a method for detecting the recesses L and R. Now the image memory 41 is the unknown pattern of the number "2" (3rd
(shown in Figure 3) is stored, the screen 8 of this image memory 4 is first scanned from left to right (Figure 3 (
2)).

By sequentially extracting the points where each scanning line 3 reaches the unknown pattern, the left side contour of the unknown pattern can be detected, thereby extracting the leftward concave portion. (Scan the screen 8 of the image memory 4 from right to left (Figure 3 (3))
).

Then, by sequentially extracting the points where each scanning line reaches the unknown pattern, the points can be detected on the right side contour of the unknown pattern, and thereby the rightward concave portion k can be extracted.

FIG. 4 shows a state in which an unknown pattern of the number "2" is formed on the screen 8 of an image memory 40 consisting of 8-bit pixels 9 in a vertical 10-bit x structure.

In the figure, the shaded area means the "black pixel" that makes up the unknown pattern, and the white area means the "white pixel" of the background part.
When the j-plane 8 is scanned by the method shown in FIG. 3, the pattern outline can be detected. The scrap data of this contour line is the fifth
In the illustrated example, the data is stored in the data setting areas M, , M2 shown in the figure, and in the illustrated example, the left contour line is stored at addresses n to I]+9 of the iσ data setting area M1 (shown in FIG. 5 ( )), and the right contour The line position data is stored at addresses n1 to 01+9 (shown as 51'QI+21) of the data setting area M2.

Figure 6 shows the data comparison area M3 of RAM 5, and in the figure, MINUS, PLUS, EQUAL, Y, T
Each area indicated by r is calculated based on the position data.
Binary data of ljrOJ is set.

FIG. 7 is a diagram showing binary data calculated from the position data of the left contour line. In the same figure, M
The INUS, PLUS, and EQUAL columns are based on the above n.
w n Each position of 1,000 9th address is compared in size with the position data of the previous address, and the increase/decrease trend is expressed as binary data of "1" and "0", and each column of Y and T is The deepest point of the recess detected based on the increasing/decreasing trend is similarly expressed as a binary theta.

FIG. 8 shows the control operation of the character recognition device when identifying the number "2" and the alphabetic character rsJ. At the start, an unknown character is read and the unknown pattern is stored in the image memory 4. First, in step 1], the image (image memory 4) is scanned from left to right, and the position data of the left contour of the unknown pattern is stored in the data setting area M1 of RAM 5. In step 12, the position data at each address in the data setting area M1 is compared in size with the position data at the previous address.If the result is that the size of the position take changes by one in the negative direction, the determination in step 13 is YES. ″′ and step 1
5, "1" is set in the MINLJS area of the data comparison area M3. If the position data changes in the positive direction, the determination in step 13 is "NO", the determination in step 14 is -YES", and the process advances to step 16, where the PLUS area of the data comparison area M3 is changed. ni "]
” or sesera. If the data content has not changed in the history, each determination in steps 13 and 14 becomes "NO" and the process proceeds to step 17, where the EQUA in the data comparison area N13 is
"1" is set in the L area. And step 15,
As a general rule, step 1 is performed after each set processing in 16.17.
Returning to step 2, the next data comparison process is executed.

In this way, when the setting process in step 15 is performed, if "1" has already been set in the PLUS area in the previous comparison process, the determination in step 18 is ``YES''.
[]] is set in the Y area in one step. As a result, the deepest point A of the leftward concave portion is detected, and in the next step 20, the address corresponding to "1" in the Y area (the address corresponding to the 4th address in the example of FIG. ).

The next step 21 checks whether scanning in both left and right directions has been completed, and in this case, the determination is "No".
” Therefore, in step 22, the comparison take area M
After clearing the contents of step 3, the process returns to step 11, where scanning is performed from right to left and data comparison processing similar to that described in section 2 is executed.

When the deepest point B in the rightward concave portion is detected in this way, "1" is written in the Y area and its corresponding address is stored, as described above. Then, the sizes of both memory addresses are compared, and the leftward recess I, the bitter memory address, or the rightward recess]
(If the determination in step 23 is "NO", the Stellar 1230 determination is ``YES'' and the unknown pattern is determined to be the number "2" and processed. Proceed to step 25,
The unknown baku-n is determined to be the English letter "SJ" and is processed.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing the circuit configuration of the character recognition device, Fig. 2 +1) +2+ +31 is a diagram showing the principle of character recognition according to the present invention, Fig. 3 +1) +2) (3) shows the four parts of the unknown pattern. A diagram showing the detection method, Figure 4 is a diagram showing an unknown pattern formed in the image memory, @5 Figure fi+ [
2+ is a diagram showing the data setting area of RAM, and Figure 6 is R.
Figure 7 shows the data comparison area of A M, Figure 7 is a table showing the data contents of the data comparison area, Figure 8 is flowchart 1 and does not show the control operation of the character recognition device. . 2...Camera 3...A/D converter 4...Image memory 5...CPU7
...RAM

Claims (1)

[Claims] A reading means for reading the optical pattern of the unknown character, a storage means for storing the read data of the optical pattern, and a discrimination unit for detecting the recessed part of the optical pattern from opposite directions and comparing the positional relationship of each part to determine the unknown character. A character recognition device comprising means.