JPS60146380A - Character recognition device - Google Patents

Abstract

PURPOSE:To execute character recognition efficiently by characteristic quantity of a stroke at every code in the every direction by comparing a directional code of a point concerned and its adjoining point and giving the stroke number of the adjoining point having a coincide directional code for a point concerned. CONSTITUTION:A directional code giving part 13 reads out sequentially character patterns in a memory 12, and decides whether or not a point concerned of a character pattern belongs to a continuous black point string having more than reference length, and gives a directional code for the point concerned when it belongs to the black string having more than N number of picture elements. A stroke extracting part 15 extracts a stroke in accordance with the directional code of the point concerned by the directional code giving part 13. A characteristic quantity calculating part 16 obtains characteristic quantity of the character pattern obtained in the stroke extracting part 15. The characteristic quantity at every stroke obtained in the characteristic quantity calculating part 16 is compared with a recognition dictionary at every character pattern and a recognition part 17 recognizes a character pattern.

Description

DETAILED DESCRIPTION OF THE INVENTION FIELD OF INDUSTRIAL APPLICATION The present invention relates to a character recognition device that extracts strokes in each direction of a binary character pattern, and performs a characteristic extraction for each stroke.

Configuration of conventional example and its problems Conventionally, the input is a binary digital data string with polarities of "0" and "1" obtained by horizontally scanning the character to be recognized, and the strokes in each direction of the character to be recognized are If you want to extract
It uses a thinning method, which is extremely complex and requires a lot of processing time, and is difficult to implement in hardware.

Purpose of the Invention The present invention solves the above problems, easily extracts strokes in each direction without thinning them, calculates the feature amount of each stroke, and recognizes the target character using the feature amount. The purpose is to provide a character recognition device that can.

Structure of the Invention The present invention includes a direction code assigning unit that determines the direction code of a point of interest in a character pattern to be recognized, and a direction code of the point of interest and a direction code of a point adjacent to the point of interest, which is compared with the direction code of the point of interest.
Strokes are extracted by assigning stroke numbers of adjacent points with matching direction codes to points of interest, and character recognition is efficiently performed using stroke features for each direction code.

DESCRIPTION OF THE EMBODIMENTS FIG. 1 is a block diagram showing the configuration of a character recognition device according to an embodiment of the present invention.

In FIG. 1, 10 is a character pattern input signal, and 11 is a character pattern input signal.
1 is a binarization unit that binarizes the character pattern input signal 10;
2 is a memory unit for storing character patterns; 13 is a direction code assigning unit for determining the direction of each point of the character pattern; 14 is a direction code buffer for storing the obtained direction code for one horizontal scanning line; 16 is a part for storing the character pattern; A stroke extraction unit extracts strokes made up of points having connected codes in the same direction; 16 is the number and length of each stroke;
The operation of a character recognition device having four or more components, including a feature amount calculation section for calculating all areas, etc., and a recognition section for performing character recognition using 17 feature amounts will be described below.

In FIG. 1, the character pattern signal 10 to be recognized is 2
It is binarized into black and white in the digitization section 11 and stored in the memory section 12 as a character pattern.

The direction code assigning unit 13 sequentially reads out the character patterns from the memory unit 120 in the horizontal direction, and selects consecutive character patterns whose points of interest are in the direction of each direction code shown in FIG. It is determined whether the point is part of a black dot string, and if it is a part of a black dot string of N or more pixels, a corresponding direction code is given to all points of interest. For example, in the character pattern shown in FIG. 3(a), when the reference length N=s, the direction code of each point becomes as shown in FIG. 3(b).

The direction code of each point determined by the direction code adding section 13 is sequentially stored in the direction code buffer 14 for one horizontal scan.

The stroke extraction unit 16 uses the direction code of the point of interest provided by the direction code adding unit 13 and the direction code of the point of interest as shown in FIG.
The direction code of the point in front of the pixel and the direction code of each point one horizontal scanning line before obtained from the direction code buffer 7 are compared, and if the direction code of the point in front of the pixel matches the direction code of at least one point, the point of interest is assigned to the point of interest. The stroke number to which it belongs is assigned, and if they do not match at all, all new stroke numbers are assigned to the point of interest.

For example, the strokes in each direction of the character pattern in Figure 6(a) are Φ), (c), (d), (e) in Figure 6.
It turns out that this is true.

In addition, the feature amount calculation section 16 calculates feature amounts such as length, area, number, position information, etc. for each stroke of the character pattern drawn by the stroke extraction section 16. The feature amount for each stroke obtained by the feature amount calculation section 16 is compared with a recognition dictionary for each character pattern, and recognition is performed by the recognition section 17.

For example, in Fig. 6, the strokes for each direction code of the character pattern (a) are shown in Fig. 6 (b), (C) l (
As shown in d) and (e), the stroke feature amount for each direction of each character pattern in Figures 5 (a) and 6 (-) is as shown in Figure 7, and as shown in Figure 6 ( The number and length of strokes in each direction in a) and FIG. It is easy to see that character recognition is possible.

Further, in this embodiment, the direction codes shown in FIG. 2 are set to four directions, but any number or directions may be used.

Effects of the Invention The character recognition device of the present invention recognizes the direction code of each point of a character pattern. It is configured to extract strokes for each direction code based on the direction code of each point, and recognize characters based on the feature amount of the extracted strokes for each direction, which is different from the conventional method. There is no need for line thinning, the hardware configuration is simple, and its practical value is high.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a character recognition device according to an embodiment of the present invention, FIG. 2 is a diagram showing an example of a direction code,
Fig. 3 is a diagram showing an example of adding direction codes, Fig. 4 is an explanatory diagram of checking connectivity in stroke extraction, Figs.
It is a comparison diagram of the feature amount of each stroke of different types of character patterns. 10...Character pattern input signal, 11...
...Binarization section, 12...Meseri section, 13...
... Direction code adding section, 14 ... Direction code buff 1 part, 16 ... Stroke extraction section, 16
. . . Feature value calculation section, 17 . . . Recognition section. Name of agent: Patent attorney Toshio Nakao and one other WA
I Figure 351' ((L) -th (Io-ri scanning line e (L. @1 scan limit a- Figure 5 ('b) Figure 6 (b)

Claims (1)

[Claims] The binary character pattern data string obtained by scanning the character pattern area to be recognized is input, and the j-th point of interest P0 on the i-th horizontal scanning line is the reference length N for determining the direction code.
A direction code assigning unit that determines whether the point of interest PiJ is included in a part of the black dot array in any continuous direction and assigns the direction code to the point of interest PiJ; If the direction code of at least one point matches the direction code of the point one horizontal scanning line before and the point one pixel before the point of interest, the stroke number to which the matching point belongs is assigned to the point of interest, and if there is no match at all. In this case, a stroke extraction section that assigns a new stroke number to the point of interest, and a stroke extraction section that calculates the number of strokes in each direction, length, area, etc., are used to calculate the number of strokes, length, area, etc. of the target character. A character recognition device that performs recognition.