JPH0147835B2 - - Google Patents

Description

[Detailed description of the invention]

The present invention is characterized by normalizing the sum of black bits for each character frame division unit area in each subpattern representing a stroke in a specific direction of an input character pattern by the character line width and the size of the character frame corresponding to each subpattern. The elements are extracted, and the character features expressed by these feature elements are used as a sub-pattern feature matrix, and the character features are expressed by the sum of the feature elements of sub-patterns corresponding to the same division unit area, and the character features are expressed as an additive feature. The present invention relates to a character recognition method in which character candidates are selected using an additive feature matrix in the first step, and character names are determined using a subpattern feature matrix in the second step. As one form of character recognition method, a feature matrix (hereinafter referred to as sub-pattern feature matrix for convenience of explanation) whose feature elements F (k, i, j) are quantities expressed by the following formula is created, and in a similar format. There is a character recognition method that recognizes a character by measuring its similarity to a standard character mask. F(k, i, j)=Bk(i, j)/WL·WPk WL represents the character line width of the input character pattern. WPk represents WPh, WPv, WPr, WPl, etc. For example, WPh is the size of the character frame in the horizontal direction, and similarly, WPv, WPr, and WPl are the vertical direction, 45 degrees diagonally to the right, and 45 degrees diagonally to the left, respectively. The size of the character frame in the direction. Bk(i, j) is Bh(i,
j), Bv (i, j), Br (i, j), Bl (i, j), etc., for example, Bh (i, j) is the black color in the divided unit area (i, j) of the horizontal sub-pattern HSP. It is the sum of bits, where i is the number of the divided unit area of the character frame in the X-axis direction, and j is the number of the divided unit area of the character frame in the Y-axis direction. Similarly, Bv(i,
j), Br (i, j), and Bl (i, j) are vertical subpattern VSP and right diagonal 45 degree subpattern, respectively.
RSP is the sum of black bits in the divided unit area (i, j) of the 45-degree left diagonal sub-pattern LSP. In this method, if the number of divisions NX and NY of the character frame in the X-axis direction and Y-axis direction are both 7, and the type of subpattern is 4, one character is a 7 x 7 x 4-dimensional comparison. Furthermore, multiple types of division formats may be set for the number of divisions NX and NY, and as the number of categories (types of characters) to be recognized increases, the time required for identification becomes enormous. Processing speed will decrease. The purpose of the present invention is to improve processing speed by adopting large classification information that is compatible with the above-mentioned method, and to improve the processing speed by adding the characteristic elements of the corresponding division unit areas of each sub-pattern. This was achieved by employing an additive feature matrix in which the value is the feature element of the divided unit area. FIG. 1 shows an embodiment of a character recognition device according to the present invention. A detailed explanation will be given below based on FIG. 1. 1 is optical input from the form. This optical input 1 is input to a photoelectric conversion section 2. The photoelectric conversion unit 2 decomposes one character area into 128 x 128 pixels and converts each pixel into a binary digital signal (hereinafter referred to as input character pattern). One character is represented by an input character pattern of about 60 x 60 bits. The pattern register 3 stores the input character pattern in a format capable of reproducing the X and Y coordinates of each pixel in the expected character area, and has a capacity of 128×128 bits corresponding to the expected character area. The character line width calculation unit 4 has a shift register configuration similar to a well-known filter circuit, and receives an input character pattern and calculates the number Q of states in which all pixels in a 2×2 shift register window are black bits. The black bit sum A of the input character pattern is counted, and the line width WL is calculated using the well-known formula shown below. WL=A/A-Q The character frame detection unit 5 detects the circumscribed frame of the character by expressing it in the left end coordinate Xl, right end coordinate Xr, upper end coordinate Yt, and lower end coordinate Yb in the pattern register, and further determines the size of the character frame. This is to detect. The size of the character frame in the horizontal direction is WPh=Xr−Xl
+1, as vertical size WPv=Yt−Yb+
1. The size in the 45 degree right diagonal direction and the 45 degree diagonal left direction is detected as WPr=WPl=WPh+WPv/2. The horizontal/vertical/right diagonal/left diagonal subpattern extraction units 6 to 9 extract subpatterns HSP and VSP representing strokes in horizontal, vertical, right diagonal, and left diagonal directions based on the input character pattern and line width WL. , RSP, and LSP by extracting continuous black bits that are longer than the corresponding line width in each direction. For example, the sub-pattern HSP is created by horizontally scanning the entire surface of the pattern register 3, detecting the number of consecutive black bits for each scanning line, and extracting the black bits whose length is greater than 2WL. Extract subpatterns. Similarly, vertical, diagonal right, and diagonal left subpatterns are extracted by scanning the pattern register 3 in the vertical direction, diagonal right direction, and diagonal left direction, respectively. An example of subpatterns extracted in this way is shown in FIG. 2, in which ORG is an input character pattern, HSP is a horizontal subpattern, RSP is a right diagonal subpattern, and LSP is a left diagonal subpattern. The character frame division determination unit 10 determines the number of divisions NX, NY specified by the division format and the character frame coordinates Xl, Xr,
Receive Yt, Yb and input character pattern ORG,
Input character pattern projected on the X and Y axes respectively
For each black bit number distribution of ORG, find each series of centroid coordinates that is sufficiently larger than the set maximum number of divisions, and divide it into the number of divisions NX,
Almost equally distributed and matched in NY (NX-1)
and (NY-1) barycentric coordinates are selected and determined as the respective divided coordinates. For example, the X-axis barycenter coordinate series X (Mi) (where Mi is the barycenter number and i=1
~15) Detection is performed by projecting the input character pattern onto the
(Mp), and then the range of the character frame on the X axis using the barycenter coordinate series found in the previous step.
By dividing Xl to Xr and repeating the process of finding the center of gravity coordinates for each range three times,
14 center of gravity coordinates X (M1) ~ X (M7), X (M9) ~
Detect X (M15). 15 obtained in this way
The coordinates of the center of gravity are determined as candidate points for the divided coordinates based on the table below prepared in advance so that each divided section of the X-axis includes an approximately equal number of centers of gravity.

[Table] Similarly for the Y axis, 15 barycenter coordinates Y
(M1) to Y(M15) are determined, and the division coordinates are determined by referring to the table according to the number of divisions. In addition,
The division format is changed and set according to the input character pattern, or once rejected, it is changed to another division format, but the details will be omitted. Here, it is assumed that the number of divisions NX and NY that are common to each sub-pattern is specified as 5 × 5 according to the division format, and the corresponding division coordinates in the X-axis direction are X(M3), M6), X (M9), X
(M12) as the division coordinate in the Y-axis direction, Y(M3),
Determine Y (M6), Y (M6), Y (M9), and Y (M12). The feature matrix extraction unit 11 extracts division coordinates in the X-axis direction and both end coordinates Xl, X (M3), X (M6), X (M9), X (M12),
Xr, division coordinates in the Y-axis direction and both end coordinates Yb, Y
(M3), Y (M6), Y (M9), Y (M12), Yt, each sub-pattern HSP, VSP, RSP, LSP, and the size of the character frame WPh corresponding to each sub-pattern,
Receiving WPv, WPr, WPl and character line width WL,
Each sub-pattern is divided into a plurality of division unit areas (5×5 areas in this case) determined from the division coordinates, and the number of black bits Bk (i, j) for each division unit area is calculated.
and calculate the line width WL and the character frame size WPh corresponding to each sub-pattern as shown in the formula below.
Normalized by WPv, WPr, and WPl, the feature element F(k, i, j) =
Bk (i, j)/WL・WPk (where k=h, v, r, l,
i=1~NX, j=1~NY), in this case 5
A feature matrix consisting of ×5 × 4-dimensional feature elements is created. F (h, i, j) = Bh (i, j) / WL・WPh F (v, i, j) = Bv (i, j) / WL・WPv F (r, i, j) = Br (i , j)/WL・WPr F(l, i, j)=Bl(i, j)/WL・WPr Black bit sum Bk(i, j) for each divided unit area (i, j) of each subpattern The calculation is performed by reading each sub-pattern at the division coordinates corresponding to the division unit area (i, j) and adding up the number of black bits. For example, the black bit sum Bh (1, 1) corresponding to the feature element F (h, 1, 1) is the X coordinate Xl, X (M3) and Y
It is obtained by reading out the horizontal sub-pattern HSP in the range determined by the coordinates Yt, Y(M3) from the horizontal sub-pattern extraction section 6 and counting the number of black bits. The addition matrix calculation unit 12 receives the subpattern feature matrix F(k, i, j) extracted by the feature matrix extraction unit 11, and calculates the subpattern feature matrix F(k , i, j) at the corresponding position (i,
The value obtained by adding the feature elements of
Create an additive feature matrix with the feature element FA(i, j) of j). FA (i, j) = F (h, i, j) + F (v, i, j) + F (r, i, j) + F (l, i, j) The main classification section 13 has a standard character mask. Similar to the addition feature matrix in the input character pattern, the value obtained by adding the feature elements of each sub-pattern of the corresponding divided unit area (i, j) is calculated as the feature element.
A dictionary expressed as an additive feature matrix FSA (i, j) is prepared, and by executing the operation of the following formula, the similarity between the additive feature matrix of the input character pattern and each standard character mask is calculated.
DA is measured, and a specific number (typically 30) of category names with a high similarity rank (low DA value) are detected as candidate categories. DA=√{(,)−(,)} ^2In this example, three types of dictionaries are prepared corresponding to the number of division formats, and one dictionary specified by the division format is referenced. The above-mentioned similarity is measured as follows, but the details thereof will be omitted. In the identification unit 14, the standard character mask is a sub-pattern feature matrix in the same format as the sub-pattern feature matrix in the input character pattern.
A dictionary expressed as FS(k, i, j) is prepared, and by executing the following formula, the standard character mask FS(k, i, j) and the sub-pattern feature matrix F(k, i, j) of the input character pattern is measured, and the category name of the most similar standard character mask is output to the character code output terminal 15. Note that in this embodiment, this identification section 14
Also, three types of dictionaries are prepared corresponding to the number of division formats, and the above-mentioned similarity is measured with reference to one dictionary specified by the division format. As explained above, in this embodiment, the broad classification unit 13 narrows down the candidate categories and then performs identification, and special processing and calculations are performed to extract the feature matrix used by the broad classification unit 13. Since the character is calculated by simple addition of the output matrix of the feature matrix extraction section 11 without the need for , there is an advantage that character recognition can be performed stably at high speed without losing the extracted features. As described above, the present invention performs major classification using an addition matrix obtained by adding feature matrices extracted from sub-patterns, and then performs identification, so that a high-speed and stable character recognition device can be realized.

[Brief explanation of drawings]

FIG. 1 is a functional block diagram showing an embodiment of a character recognition device according to the present invention, and FIG. 2 is a diagram showing input character patterns and sub-patterns. 1... Optical signal input, 2... Photoelectric conversion section, 3...
Pattern register, 4...Line width calculation section, 5...Character frame detection section, 6...Horizontal sub-pattern extraction section, 7
...Vertical sub-pattern extraction section, 8... Right diagonal sub-pattern extraction section, 9... Left diagonal sub-pattern extraction section, 10... Character frame division determination section, 11... Feature matrix extraction section, 12... Addition matrix Calculation section, 13... Major classification section, 14... Identification section, 15
...Character name output terminal.

Claims (1)

[Claims] 1. Obtained by normalizing the number of black bits of an arbitrary divided unit area in a subpattern representing a stroke in a specific direction of an input character pattern by the character line width and the size of a character frame corresponding to the stroke direction. The feature elements are extracted from the input character pattern for each of a plurality of sub-patterns with different stroke directions and for each divided unit area obtained by dividing the character frame, and the amount is taken as a feature element to create a sub-pattern. In a character recognition method that creates a feature matrix and recognizes an input character pattern by referring to a dictionary in which standard character masks are described in the same format as the subpattern feature matrix, all Separately prepare a dictionary in which standard character masks are described using an addition feature matrix in which the value obtained by adding the feature elements of the sub-pattern is the feature element of the division unit area.
An additive feature matrix of the input character pattern is created with the value obtained by adding the feature elements of all the sub-patterns corresponding to the same division unit area as the feature element of the division unit area, and the addition feature matrix of the input character pattern is The degree of similarity with each standard character mask in the dictionary of the addition feature matrix is measured to detect a group of standard character names whose degree of similarity is at least a certain value or a certain rank, and then A character recognition method characterized in that an input character pattern is recognized by measuring the degree of similarity between a standard character mask corresponding to a standard character name and the sub-pattern feature matrix of the input character pattern.