JPS60201480A - Character read system - Google Patents

Abstract

PURPOSE:To improve the reading accuracy by separating a read area for designating plural character kinds with a space of a >one character so as to divide a data at each character kind set for the recognition again. CONSTITUTION:A read area on a form is scanned, inputted to a line buffer 11 as a white/black binary-coding image data, and a character image is extracted (12) at each character, a character kind 18 extracted by a parameter and one read area is recognized (13) and the data is separated to each character block. Then the character block is checked for the content by character kind set information 19 designated by the parameter, an optimum character kind set is decided (15), the data is recognized (16) again at each character block and a read area on the form is read. The read area designated for plural character kinds is a block of KATAKANA (square form of Japanese syllabary) and alphanumeric characters where the character kinds are mixed with, e.g., numerals, English letters and KATAKANA. This block is read by a recognition block 13 by a dictionary reading the mixture of numeral, English letter and KATAKANA, the result of recognition is outputted while the KATAKANA is taken as A, the English letter is taken as B by a discrimination output block 14 as to each character kind set.

Description

[Detailed Description of the Invention] (Technical Field) The present invention relates to a character reading method, and it is possible to improve character reading accuracy by separating the reading area on a form specified by multiple character types with a space for each character type set. Regarding possible character reading methods.

(Prior art) In conventional optical character reading devices, one
When reading two reading areas, judgment must be made using a mixed recognition dictionary, which has the disadvantage that reading accuracy is lower than when reading with a single recognition dictionary.

In addition, conventional character reading devices have the disadvantage that in mixed reading, the reading accuracy decreases as the number of mixed character types increases.

(Object of the Invention) The object of the present invention is to eliminate such drawbacks in conventional optical character reading devices, and to read in mixed reading by providing a plurality of character types in the reading area! '# We aim to provide a method for reading characters that improves reading comprehension.

(Structure of the Invention) According to the present invention, in a character reading method in an optical character reading device, there is provided a form having at least one reading area in which a plurality of character types are written, and a form that reads the reading area in a plurality of character types. , means for separating each character type set into character blocks with one or more spaces, means for setting the character type of the character block, and character type set information determined by the determining means based on the determined character type. , a re-recognition means, and t, and a character reading method is obtained which is characterized in that character reading accuracy is improved by reading characters for each character type.

(Example) Next, embodiments of the present invention will be described with reference to the drawings.

Figure m1 shows embodiments of the present invention. In FIG. 1, one embodiment of the present invention is a character reading method in an optical character reading device, which has at least one reading area in which a plurality of character types can be written on a form, and this reading area is optically readable and black and white. Line buffer 11 for temporarily storing binarized image data. ! :, a character extraction s12 that extracts a character image for each character from the data of the line buffer 11; a recognition block 13 that recognizes a reading area of the character image of the character extraction unit 12 according to a character type 18 specified by a parameter; The recognition block 1
A judgment output block 14 separates each reading area by a space for each character block according to the output signal of 3, and judges the character type of the character block 1;
A character type block determination unit 15 checks the contents of character blocks separated by spaces from 4 to 4 using character type set information 19 specified by parameters, and determines the optimal character type set. It also includes a character type block recognition unit 16 that performs the above recognition, and a final judgment processor 17 that finally outputs the recognition result.

That is, in this embodiment, the reading area on the form is scanned and the line buffer l is scanned as black and white binary image data.
, and then extract all character images for each character. Next, one reading area is recognized using the character type 18 specified by the parameter and separated into each character block.

Next, the contents of the character blocks are checked using the character type set information 19 specified by the parameters, and the optimum character type is determined, and each character block is recognized again to read the reading area on the form.

FIG. 2 shows a reading area for specifying multiple character types used in this embodiment. In Figure 2, this block contains a mixture of character types, such as numbers, alphabets, and katakana (
This block consists of Kamekana) and (numbers and alphabets.) This block consists of numbers and alphabets.

In the katakana mixed reading dictionary, the recognition block 13 shows that the recognition block 13 shows that the judgment output block 14 shows that J: ri ren R shogu is katakana as A and numbers as shown in Figure 3.

The alphabetic character is designated as B for each character type set and output. The character type block recognition unit 15 determines that ``A'' is a block of numbers and alphabetic characters, and ``A'' is a block of numbers and letters. , and fcB as numbers and alphabets are again judged by the character type recognition process s6, and finally the result shown in FIG. 4 is output to the final judgment block 17.

FIG. 5 shows character type sets that can be specified using this method. B1
When specified: (numbers, alphabetic characters), (katakana).

(symbol) character type se, soto specification. Similarly B2-B
Even when 6 is specified, each character type set is specified. Note that N is a number and a minus, A is an alphabetic character and a minus, K is a katakana and a minus, and S is a symbol and a minus. In addition to the character type set shown in Fig. 5, combinations shown in Fig. 6 are also possible. FIG. 6 shows a case where (numbers, alphabets, katakana) and (symbols) are used as a character type set. In Figure 6, in addition to character types td, ``BI B2'', ``BIB''
6”, “B2B6”.

Either 'BIB2B6''' can be specified.

(Effects of the Invention) As explained above, the present invention separates each character type set by one or more spaces in the reading area designated by the number of character types, thereby re-recognizing each character type set. It can improve the degree of treetopness.

[Brief explanation of drawings]

FIG. 1 is a partial block diagram showing an embodiment of the present invention, FIG. 2 is a diagram showing an example of a reading area for specifying multiple character types, and FIG. C A diagram showing an example of the result of the judgment output block judged by the recognition block in Fig. 1, 4th
The figures show examples of the results of the final judgment block judged by the character type block recognition unit in Fig. 1, and Figs. 5 and 6.
The figure is a diagram showing character types that can be specified in an embodiment of the present invention. 11... line parr 9, 12... character extraction section, 13... recognition block, 14...
... Judgment output block, 15... Character type block judgment section, 16... Character Yuzu block recognition capital,
17...Final judgment block, 18...
Multiple character type parameters, 19...Character diagram Ku Ku CXJ (E) Dimension Shell Elephant

Claims (1)

[Claims] In a character reading method in an optical character reading device, there is provided a form having at least one reading area in which a plurality of character types are written, and the reading area is read with a plurality of character types into a character block with a space of L characters or more for each character type set. Separating means, means for determining the character fiiwo of the character program, and means for re-recognizing the character type set information determined by the determining means using the determined character type. The character staring method is characterized by improving character reading accuracy by reading the characters.