JPS6262392B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a fast and accurate character recognition method.

In general, recognition of characters and figures is performed by extracting strokes from character and figure patterns and extracting geometric features such as the positions, lengths, and mutual relationships between the strokes. And to extract the stroke, the following 2
Two methods are widely used.

Calculate the curvature of the contour point series detected by tracing the contour of the character shape, divide the contour series using points with large curvature values as division points, and extract strokes by combining the divided series. I do.

A character/figure pattern is thinned to form a skeleton, and the connectivity and skeleton pattern of the skeleton pattern is traced to detect sudden angle changes and extract strokes.

However, when the character/graphic pattern is large or complex, the amount of processing increases and the processing speed decreases. In addition, the method requires thinning of the character/figure pattern, and furthermore, the thinning causes problems such as distortion of the pattern and generation of whiskers, which complicates subsequent processing. It has the following disadvantages.

Therefore, as a method to eliminate these defects, the line width W of the character/figure pattern is detected, the pattern is scanned in a desired direction, and the length l of the writing point of the pattern in the scanning direction and the line width W are calculated. A sub-pattern representing the stroke component in the scanning direction is extracted by extracting entry points that satisfy lnW (n is a constant) between ), and a method has been proposed in which the character dose in each area is calculated and recognized. FIG. 1 shows an example of recognition of the kanji character "dog" using the above method. When vertical sub-pattern B and horizontal sub-pattern C are extracted from the character pattern shown in A, strokes A and B of character pattern A have continuous lengths of points less than Nω in each scanning direction, so
This means that it does not belong to either subpatterns B or C.

That is, it has the disadvantage that character recognition is difficult because the features representing the stroke are missing.

The purpose of the present invention is to eliminate these drawbacks of the prior art, and when extracting subpatterns representing stroke components in each direction, furthermore, among the black bits of the original pattern, black bits that do not belong to the subpatterns are undefined. Extract the subpattern, create a feature matrix including this undefined subpattern,
This system is designed to recognize characters and figures, and will be explained in detail below.

FIG. 3 shows an embodiment of the character recognition device of the present invention. Optical signal input 1 is photoelectric conversion section 2
The signal is converted into a binary quantized digital electrical signal and stored in the pattern register 3. At the same time, the line width of the input pattern is calculated in the line width calculating section 4. The vertical sub-pattern extracting section 5 performs a vertical scan on the entire pattern register, and extracts a vertical sub-pattern (VSP) from the relationship between the continuous length of black bits and the line width calculated by the calculating section 4. Similarly, the horizontal sub-pattern extraction section 6 extracts the horizontal sub-pattern (HSP) by a horizontal scan, the right-diagonal sub-pattern extraction section 7 extracts the right-diagonal sub-pattern (RSP) by a 45° right-diagonal scan, and the left-diagonal sub-pattern extraction section 8 extracts the left diagonal subpattern (LSP) by performing a left diagonal 45° scan.

The undefined sub-pattern extractor 9 extracts an undefined sub-pattern (USP) from among the black bits stored in the pattern register 3 that do not belong to any of the aforementioned sub-patterns. USP is HSP,
The extraction is performed by applying the following logical formula (1) between VSP, RSP, LSP and the original pattern P.

USP=P〓〓〓〓 (1) The character frame detection unit 10 detects a character frame circumscribing the character pattern in the pattern register 3, and sends the result to the character frame division determination unit 11. The character frame division determining unit 11 determines dividing points on the X and Y axes for dividing the inside of the detected character frame into M×N areas (M and N are integers; in this embodiment, M=N=5). Determine the coordinates. Here, the X axis indicates the horizontal direction of the character frame, and the Y axis indicates the vertical direction.

The feature matrix extraction unit 12 extracts VSP,
Divide the character frame area on each subpattern register of HSP, RSP, LSP, and USP into M×N areas, count the number of black bits Bij in each area, and use the line width W to calculate the Calculate the feature indicating the character line length, M×N
Create a ×5-dimensional feature matrix Lij.

L _ij = B _ij /W...(2) Then, the VSP feature matrix is the length ΔX of the character frame in the x-axis direction, the HSP feature matrix is the length ΔY in the y-axis direction, and the RSP, LSP, and USP feature matrices are Normalization is performed using √() ² +() ² to finally create an M×N×5-dimensional normalized feature matrix.

The identification unit 13 applies the distance (D) defined by equation (3) between the standard character mask (f _n ) and the normalized feature matrix (f _i ), and selects a standard such that D is the minimum value. Output character mask category name as character name 14
Output to.

D=√( _n − _i ) ² (3) In this way, the input character pattern is recognized.

FIG. 3 shows an example of recognition of the kanji character "dog" by the method of the present invention. A is the original pattern, B is
VSP, C is HSP, D is RSP, E is LSP, F is USP
are shown respectively.

As described above, this embodiment has the advantage that the introduction of the undefined subpattern (USP) eliminates the lack of strokes seen in conventional character recognition methods.

When extracting subpatterns representing strokes in a desired direction from a character pattern, the present invention further extracts undefined subpatterns by applying a simple logical formula, and creates a feature matrix that includes these subpatterns for recognition. This has the excellent effect of not only preventing missing strokes but also enabling simple and accurate character recognition.

[Brief explanation of the drawing]

FIG. 1 is an explanatory diagram of sub-pattern extraction by a conventional method, and FIG. 2 is a configuration diagram of an embodiment according to the present invention.
FIG. 3 shows examples of sub-pattern extraction according to the present invention. 1... Optical signal input, 2... Photoelectric conversion section, 3...
Pattern register, 5 to 9...Subpattern extraction unit, 10...Character frame detection unit, 11...Character frame division determination unit, 12...Feature matrix extraction unit, 13...
...Identification part, 14...Character name output.

Claims (1)

[Claims] 1. A digital signal obtained by photoelectrically converting and quantizing a character figure is stored in a pattern register as an original pattern, and a first sub-pattern representing stroke components in each direction from the original pattern, and the first sub-pattern are A second sub-pattern representing a black bit component in the original pattern that does not belong to is extracted and stored in a sub-pattern register, and the area within the character frame of the original pattern is divided into M×N areas ( M and N are integers), the feature amount representing the character line length in each region is normalized by the character size to create a feature matrix, and the feature matrix is compared with a standard character matrix to determine the character shape. A character recognition method that recognizes .