JPS60142784A - Character separating system - Google Patents

Abstract

PURPOSE:To separate characters for one character's share from a manuscript character pattern by providing a means detecting specific line segment end point where a specific line segment contacts a character pattern and a character separating section separating the character pattern based on the line segment end point. CONSTITUTION:The range of a threshold value alpha is decided from a center line C of a lateral area of the character area, the range is scanned horizontally and the center position of the part clipped with black points is obtained as the specific line segment. The specific line segment is obtained from the upper and lower range respectively in this way and when plural specific line segments exist, the segment closer to the center line C is adopted. When the specific line segment from the upper range and that from the lower range are in contact with the character pattern, the character pattern is separated from the contact point closer to the center line C as shown in (D). Thus, the contacted manuscript character pattern is separated accurately with a comparatively simple means.

Description

DETAILED DESCRIPTION OF THE INVENTION (Technical Field of the Invention) The present invention relates to a character separation method for separating joined characters in handwritten alphanumeric characters.

[Conventional technology and problems]

Conventional handwritten character recognition technology recognizes what is written within a predetermined character frame for each character, and can only recognize what is written within a predetermined character frame. Moreover, since it was written within the character frame, there was no joining between the characters.

By the way, in logic circuit diagrams and the like, these symbols are displayed with handwritten characters to indicate names of circuit elements, human input signals, and the like. However, when inputting a logic circuit diagram into a data processing device, this handwritten character cannot be recognized because it is not written within a special character frame, but because there are continuous parts between characters. There wasn't. For this reason, in the past, when a circuit designer created a drawing, he or she had to input the symbol portion of the handwritten characters as data on a punch card, and then input it into a data processing device using a card league.

As a result, circuit designers are burdened with having to input data of handwritten characters in addition to creating drawings, so there is a demand for the development of a device that can automatically read these handwritten characters.

[Purpose of the invention]

The purpose of the present invention is to separate data in which characters are joined in a group of character strings written without a character frame, which is necessary for recognizing such handwritten character parts, and to separate them into individual characters. The object of the present invention is to provide a character separation method that allows character extraction.

[Structure of the invention]

To achieve this objective, the character separation method of the present invention:
handwritten character holding means for holding joined handwritten characters; singular line segment extraction means for extracting a singular line segment by scanning within a certain range from the estimated character center position of the joined handwritten character area; It is equipped with a singular line segment end point detection means for detecting a singular line segment end point where a line segment contacts a character stamp, and a character separation processing unit that separates a character pattern using this singular line segment end point, and is capable of detecting one character from a joined handwritten character stamp. The feature is that the character pattern for minutes is separated.

[Embodiments of the invention]

Before describing the present invention in detail based on one embodiment, the principle of operation thereof will be explained.

(1) First, a drawing is scanned to detect handwritten character groups. It can be seen that handwritten character groups can be easily identified and extracted because black areas are concentrated to a certain extent compared to patterns and line parts.

(2) After extracting a handwritten character group in this way, as shown in Figure 1 (a), scan this in the horizontal direction, and as shown in Figure 1 (b), the height of the character group is (length) H and length L.

Then, as shown in FIG. 1(b), this area is scanned vertically to separate separable characters. As a result,
In FIG. 1, A, ESC, and D can be separated individually, but since "4N" is joined, it is extracted as one unit as shown in FIG. 2 (A).

(3) By the way, as shown in Figure 1 (C), the size of normal characters has a width to height ratio of approximately 2:3, so in this case, the vertical YW in Figure 2 (A) From the ratio with horizontal XW, it is identified that this corresponds to a case where two characters are joined.

(4) Next, as shown in Figure 2 (b), an area with a threshold value α is determined from the center line C of the horizontal area of this character area, and this area is horizontally scanned to find the area between the black dots. Take the center position as a singular line segment. At this time, if there is no black spot area on one or both sides, the center position between the frame portion of the threshold value α and the black spot, or the center position between the frame portions is defined as the singular line segment.

In this way, singular line segments are obtained from the upper part and the lower part, respectively, but when there are a plurality of singular line segments, those close to the center line C are left.

(5) When the singular line segment from above and the singular line segment from below touch the character slam, as shown in Figure 2 (d), this Separate character patterns.

An embodiment of the present invention will be described based on FIGS. 3 to 5 and with reference to FIG. 2.

In the figure, 10 is an image memory, 11 is a rectangular area
12 is a memory, a threshold value holding register, 13 is an address control section, 14 is a singular line segment extraction section, 15 is a singular line segment end point detection section, and 16 is a single character separation processing section.

The image memory 10 stores video input signals and holds the entire original image.

The rectangular area buffer memory 11 is a buffer memory into which data determined to be connected is input, as shown in FIG. 2(a).

The threshold value holding register 12 is a register in which the threshold value α shown in FIG. 2 (b) is written. This threshold value α is determined by simulation, but for example, from 10 to the width value of XW.
It is set at around 20%.

The address control unit 13 starts from the image memory 10 as shown in FIG.
Cut out the area shown in and create a rectangular area Ba Sofa Memory 1
1 or rectangular area buffer memory 1
This function generates an address for reading out a character stamp stored in 1 or for cutting out areas α on the left and right sides of the center line C.

The singular line segment extraction unit 14 extracts singular line segments within the area set by the threshold value α. Here, the singular line segment indicates the midpoint of the character pattern, but as shown in Figure 4, there is no character pattern in the part β1 within the area, so the symbol ■ indicating the midpoint of that area is the singular line segment. So, β2
Since only character pattern P2 exists in the part, ■ indicates the midpoint between the left end of the area and the left side of character pattern P2, and ■' indicates the midpoint between the right side of character pattern P2 and the right end of the area. is a singular line segment. In the part 13, ■, which indicates the midpoint between the right side of the character pattern P1 and the left side of the character pattern P2, ■', which indicates the midpoint between the right end of the area and the left side of the character pattern P1, and the above ■゛ are singular line segments. becomes.

When a plurality of singular line segments exist, the one closest to the center line C of the area is adopted.

The singular line segment end point detection unit 15 detects the end point where the singular line segment and the character pattern come into contact, and for example, as shown in FIG.
This is to find the end point E2 of the singular line segment where the character pattern P and the singular line segment 1 and the character pattern P are in contact. In addition, when there are multiple singular line segment end points on each of the upper and lower sides, the center line C
The end point of the singular line segment between the character pattern and the character pattern is detected.

The character separation processing unit 16 separates the joined character patterns P, and passes through the singular line segment end point which is closer to the center line C among the singular line segment end points E1 and β2, and the center line C In the example shown in FIG. 5, character patterns P are separated based on straight lines that are parallel to the end point E1 of the singular line segment.

Next, the operation of FIG. 3, which is an embodiment of the present invention, will be explained.

(a) Image data input to the image memory 10,
Processing is performed by a graphic processing device (not shown) to extract a character stamp group as shown in FIG.

Then, in order to detect whether or not the character patterns in this part are joined, the point S that first contacts the character pattern is detected by scanning in the YW force direction shown in Figure 2 (a), and the point S that first contacts the character pattern is known from this point S. Trace the area around this character pattern using the method
When it passes through almost the full width of W and returns to point S, which is the starting point, it can be determined that a joined character pattern exists in the XW and YW regions.

(b) In this way, when it is determined that two characters are joined in the area shown in FIG. 2(a), this area is written into the rectangular area buffer memory 1. Then, as shown in FIG. 2(b), a region within the range of α entered in the threshold value holding register 2 is determined from the center line C-C of this region, and the singular line segment extraction unit 14 determines the region as shown in FIG. 4, for example. Place singular line segments such as ■, ■, ■, - above and below it.

Of course, in this case, singular lines like ■' and ■' in Figure 4.

minute is further away from the center line C-C than the singular line segments ① and ②, so it will be excluded.

(C) Next, the singular line segment end point detection unit 15 determines that the singular line segment left as a result of the processing shown in FIG.
is detected as the end point of the singular line segment.

(d) - The character separation processing unit 16 separates the character pattern P based on the line at the singular line segment end point E1, which is closer to the center line C-C, among the singular line segment end points El and β2.

It becomes possible to recognize the separated character patterns by comparing each of the character patterns separated in this way with a dictionary button using a well-known method.

In the above description, alpha numeric characters were used as an example, but the present invention is of course not limited to this. If one character is not separated in the vertical direction, it can be separated, and alpha numeric characters can be effectively separated because there is no separation part in one character.

0 Furthermore, the number of joined characters is not limited to only two characters. For example, when it is determined that three characters are joined based on the aspect ratio of the character pattern, the area may be drawn to a position that divides the area into three equal parts, and the left and right threshold ranges may be processed equally.

〔Effect of the invention〕

According to the present invention, joined handwritten character patterns can be accurately separated using relatively simple means.

[Brief explanation of the drawing]

Fig. 1 is an explanatory diagram of a character pattern, Fig. 2 is an explanatory diagram of a separated state of a joined character pattern, and Fig. 3 is a configuration diagram of an embodiment of the present invention.
FIG. 4 and FIG. 5 are explanatory diagrams of its operating state. In the figure, 10 is an image memory, 11 is a rectangular area buffer memory, 12 is a dark value holding register, 13 is an address control section, 14 is a singular line segment extraction section, 15 is a singular line segment end point detection section,
16 is a single character separation processing section. Patent applicant Fujitsu Ltd. Representative Patent Attorney Akira Yamatani 1

Claims (1)

[Scope of Claims] 1. A handwritten character holding means that holds joined handwritten characters, and a specific line segment is extracted by scanning within a certain range from the estimated character center position of the joined handwritten character area. comprising a singular line segment extraction means, a singular line segment end point detection means for detecting a singular line segment end point where the singular line segment contacts a character slam, and a character separation processing unit that separates the character bang by the singular line segment end point, A character separation method characterized in that a character stamp for one character is separated from joined handwritten character stamps. 2. The character separation method according to claim 1, wherein the character BATA is an alpha numeric character.