JPS60159987A - Character recognizing device - Google Patents

Abstract

PURPOSE:To enable character recognition by simple method by extracting a quadrangle circumscribed by character pattern, dividing this into three directions longitudinally or lateraly, and deciding the divided area in which characteristic such as character branch point edists. CONSTITUTION:A binary coded character pattern is stored in a picture memory 1. A circumscribed quadrangle extracting circuit 2 extracts a quadrangle circumscribed by the character pattern on the picture memory 1. A branch point position extracting circuit 3 extracts position of existence of branch point that forms characteristic of a character. These extracted data are stored in a RAM4. A CPU5 interprets program of a program memory 6, and executes reading or writing of data to the RAM4 and picture memory 1, and at the same time, controls operation of above-mentioned extracting circuits 2, 4. By this way, distinction between P and 9, 5 and 2 can be obtained clearly by only adding a simple processing method.

Description

[Detailed Description of the Invention] <Technical Field of the Invention> The present invention optically reads unknown characters and converts them into black and white.
The present invention relates to a character recognition device that identifies the unknown character by converting it into a value to obtain a character pattern, extracting the characteristics of the unknown character from the character pattern, and comparing the characteristics with a standard pattern stored in a dictionary in advance.

<Background of the Invention> In a conventional character recognition device, during dictionary matching processing, candidate characters are narrowed down step by step using the characteristics of an unknown character, and finally a detailed dictionary matching operation is executed.

For example, focusing on the presence or absence of intersections and loops among character features, the numbers "1" to "9" are classified into four groups according to their aspects, as shown in FIG. Therefore, if an unknown character has the characteristics of ``there is an intersection'' and ``no paroop'', the unknown character is determined to be one of the candidate characters r4Jr5J r7J included in the third group. If similar processing is performed using other character features such as the number of loops, the number of endpoints, the presence or absence of branch points, candidate characters can be further narrowed down. However, in this type of narrowing down operation, if only intersections, branch points, loops, depressions, etc. are used as character features, it will not be possible to distinguish between, for example, the alphabetic letter "P" and the number "9" shown in Figure 2, or even the number "9". It is impossible to distinguish between the number "5" and the number "2" shown in FIG. 3, and it is therefore necessary to adopt a more complicated processing method to recognize these characters.

<Objective of the Invention> The present invention focuses on a specific character feature and determines the location of that feature in a character pattern.
It is an object of the present invention to provide a character recognition device that eliminates the above-mentioned disadvantages.

<Configuration and Effects of the Invention> In order to achieve the above object, the present invention extracts a quadrilateral circumscribing a character pattern based on the binary data of unknown characters, and divides the area of this quadrilateral into, for example, the fourth and fifth regions. The screen is divided into three parts in the vertical direction shown in the figure or in the horizontal direction shown in Figs. 6 and 7, and it is determined in which divided area a characteristic part such as a character branch point is located.

According to the present invention, in the case of the English letter rPJ in FIG. 4, the character branch points T, , T2 exist in the leftmost region XA, while in the case of the number "9" in FIG. Point T1.
T2. T3 exists in the rightmost region XC and does not exist at all in the leftmost region XA, and therefore the two can be clearly distinguished. In addition, in the case of the number "5" in Fig. 6, the branch point TI,
T2 exists in the upper end area YA, while the number “2” in FIG.
'', the branching points Tl and T2 exist in the lower end region Yc, but not in the upper end region YA, and therefore the two can be clearly distinguished. Thus, the present invention achieves the remarkable effects of achieving the purpose of the invention, such as making it possible to distinguish between specific characters, which was previously impossible, and contributing to improved character recognition accuracy.

<Description of Embodiments> FIG. 8 shows an example of the circuit configuration of an apparatus according to the present invention, and the image memory 1 in the figure stores a black and white binary character pattern. Further, the circumscribed quadrilateral extraction circuit 2 includes the image memory 1
Above, the quadrilateral circumscribed by the character pattern is extracted, and furthermore, the branch point position extraction circuit 3 extracts the location of the branch point that is characteristic of the character. These extracted data are stored in RAM (Ra
ndom Access Memory) 4, and is stored in the CPU (Central Processing
gUnit) 5 decodes the program in the program memory 6, reads or writes data to the RAM 4 and the image memory 1, and controls the operations of the extraction circuits 2 and 4.

FIG. 9 shows the control operation of the CPU 5. Assuming that the character pattern 7 of the English letter "P" is stored on the XY coordinates of the image memory 1 shown in FIG.
In step 10 of FIG. 9, a quadrilateral 8 circumscribing the character pattern 7 is extracted. The data defining this quadrilateral 8 are the maximum and minimum X coordinates of the character pattern 7.
M, Xm, and the maximum and minimum values of Y coordinates YM, Ym
These coordinate data (in the illustrated example, Xm = 2, XM = 9, Ym = 2, YM = 10
) are sequentially stored in a predetermined area of the RAM 4 shown in FIG.

Next, in step 11, the CPU 5 selects the branch point T1 from the character pattern 7. Coordinates where T2 exists (X, Y
), and its coordinate data (in the illustrated example, Tl is (
4,5) and T2 is (4,6)) are stored in a predetermined area of the RAM 4 shown in FIG. In FIG. 12, the table stopper is a code indicating completion of branch point information.

Next, in step 12, the CPU 5 divides the quadrilateral 7 into three parts vertically and horizontally, and in the next step 13,
Said branch point T1. It is determined in which divided region T2 exists.

FIG. 13 shows the contents of step 12.13 in more detail. In the figure, steps 20 to 25 are position data X of a line that vertically divides the quadrilateral 7 into three.
Steps 26 to 31 show the process of calculating position data Y, . The illustrated method is to divide the horizontal and vertical sides of quadrilateral 7 by 3, find the division constants Dx and Dy, and their remainders (step 20.26), and each remainder is 0.1.2.
(Steps 21-22 and Steps 27-28), the position data Xl is determined according to the remainder value.

X2, Yl, and Y2 are calculated (step 23~
25 and steps 29-31). And these device data (in the case of the illustrated example, X1=5.X2=7, Yl-5
, Y2-8) are stored in a predetermined area of the RAM 4 shown in FIG.
Three vertical and horizontal areas XA to xc, YA, % shown in Figure 6
, Yc.

Next, in steps 32 to 37, the CPU 5 executes the branch point T1. It is determined in which of the vertically divided areas XA-Xc T2 is located, and in steps 38 to 43, which of the horizontally divided areas YA and Yc is located. This determination is performed by comparing the X coordinates of the branch points Tl and T2 with the position data Xl and X2 (steps 32 and 33).
and branch point T1. T2's Y coordinate and position data Y,
, Y2 (steps 38 and 39), and based on the determination result of each step, the 17th
Branch point T1. Data representing the location of T2 is set (steps 34 to 3).
6 and steps 40-41). The branch points Tl, T
Each coordinate of 2 is read from the RAM area shown in FIG. 12, but when the read data is a table stopper, the judgments in step 37 and step 43 become 6YES'', and each judgment process is completed. In FIG. 17, data "1" is set in the RAM areas corresponding to areas XA, YA, and YB, and it is understood from this that branch points exist and are located in these divided areas.

FIG. 18 shows a character pattern of the number "9", and the branch point location data shown in FIG. 19 can be obtained by the same method as above.

Comparing the data arrangement in FIG. 19 and the data arrangement in FIG. 17, it is clear that they do not match. Therefore, by referring to the data contents corresponding to area XA, for example, the alphabet r P J and the number "9"
It is possible to distinguish between

[Brief explanation of the drawing]

FIG. 1 is an explanatory diagram showing an example of classification of candidate characters, FIGS. 2 and 3 are explanatory diagrams showing examples of character patterns that cannot be classified by conventional devices, and FIGS. An explanatory diagram showing character patterns for explanation, FIG. 8 is a circuit block diagram showing an example of the device of the present invention, FIG. 9 is a flowchart showing the control operation of the CPU, and FIG. 10 shows the character pattern on the image memory. 11 and 12 are explanatory diagrams showing the data storage state in the RAM, FIG. 13 is a flowchart showing the control operation of the CPU, and FIG. 14 is the R
FIG. 15 and FIG. 16 are explanatory diagrams showing the state of data storage in AM, FIG. 15 and FIG. 16 are explanatory diagrams showing quadrilateral divided areas, and FIG. 17 is an explanatory diagram showing the data contents of RAM showing determination results.
FIG. 18 is an explanatory diagram showing character patterns on the image memory;
FIG. 19 is an explanatory diagram showing the data contents of the RAM indicating the determination result. 2... Circumscribed quadrilateral extraction circuit 3... Branch point position extraction circuit 5... CPU Patent applicant Tateishi Electric Co., Ltd. 17+l Figure-3+2. Square 3 Wate 4 Fig. Min.

Claims (1)

[Claims] ■ In a character recognition device that reads unknown characters and converts them into black and white binary data to obtain a character pattern, extracts the features of the unknown characters and compares them with a standard pattern, the character recognition device reads and converts the unknown characters into black and white binary data to determine the character pattern. A character recognition device comprising means for extracting a quadrilateral circumscribing a quadrilateral, means for dividing a region of the extracted quadrilateral into a plurality of regions, and means for determining the presence or absence of a specific character feature in each divided region. (2) The character recognition device according to claim 1, wherein the quadrilateral is divided into three parts in each of the vertical and horizontal directions. (2) The character recognition device according to claim 1, wherein the specific character feature is a branch point of a character.