JPS5827276A - Optical character reader - Google Patents

Abstract

PURPOSE:To correct data without using an additional line or an additional field by entering an inserting mark and additional data for the data already entered in a form. CONSTITUTION:When the scanning of a prescribed line is completed and the projection is taken, a recognizing part 6 scanns the contents of a vertical projection part 4 and a horizontal projection part 5. If a projection of twice or more the normal character pitch and line pitch is detected, a detecting means for detecting a mark for insertion included in the recognizing part 6 is driven to check whether the detected mark is a mark for insertion or not. Subsequently the recognizing part 6 scanns the inside of a picture memory part 2 and detects the position of X and Y coordinates of the highest point out of a mark for insertion 13. The marks 13 and an additional character line 14 are masked, the quantized patterns of the character lines 12 is cut out character patterns of the unit of one character to recognize them and the recognized results are successively stored in an unshown buffer in the recognizing part 6.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an optical character reading device which has a completely improved method of modifying the data being read.

In the past, optical character readers (hereinafter referred to as OCR) were often used for inputting data such as simple numbers, but in the future, as the types of characters that can be read increase, OCR for text (manuscript) input will be used. It is thought that its uses will continue to expand. Japanese and English word processors are becoming popular as text processing devices, and the OCR Ishikawa is naturally an ideal input machine for these devices. For such OCR, reject processing is a job on the word processor side, so it becomes unnecessary, and the form of Book M is also different from the field specification in the case of manual data processing, since it is a whole book that is not in a free format (manuscript paper, letter paper). Therefore, the number of human data inputs increases, and the number of omissions and errors also increases, so a correction method different from that of data input OCR is required. This is possible, so if you use the dropout color to mark the mistakes and their contents when creating the manuscript, you can make corrections in the word processor after inputting. Even with OCR alone, you can easily make corrections when writing the manuscript. It's convenient if you can do it.

However, with conventional OCR alone, when data is to be corrected, all fields (columns) dedicated to I4!2υ erasure are provided, and data in related lines or fields is edited by writing marks or specific symbols there. Update data by erasing, filling in all new data in separately prepared spare fields, preliminary 11 inch lines, etc.
Is there a way to replace the data, or transfer all the data to an external storage device and then use the CRT-key? The method used was to use the - code to make all changes to the contents.

This invention was made in view of the above circumstances,
You can use additionality and additional fields by filling in the insertion symbol and all additional data for the data that has already been entered in the form.

An optical character reader that can perform all data corrections! The purpose is to provide 1.

Hereinafter, one embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a schematic diagram of an embodiment of the present invention. In the figure, reference numeral 1 indicates a photoelectric conversion signal. This photoelectric conversion section 1
has the function of converting a photoelectric conversion signal into a black and white image on a form.

Reference numeral 2 indicates an image memory section. This image memory section 2 has a function of storing all data for at least two lines of a form in a state in which the photoelectric conversion signal is n-digitized into binary values of white bits and black bits. Reference numeral 3 indicates the entire address control section.

This address control section 3 has a function of controlling read and write addresses of the image memory section 2. Reference numeral 4 indicates a vertical projection section. This vertical projection section 4 has a function of taking a vertical projection while transferring an image to the image memory section 2. Reference numeral 5 indicates the horizontal projection part,
It shows. This horizontal projection section 5 has a function of taking a horizontal projection while transferring an image to the image memory section 2. Reference numeral 6 indicates a recognition section.

In addition to controlling the above-mentioned parts, the recognition section 61i also has the function of extracting all lines and characters from the image memory section 2 and recognizing the next part of each sentence.

Next, the operation of the above embodiment will be explained. FIG. 2 is a diagram showing an example of filling in a form. In the figure, reference numeral 11 indicates a document for manual input, reference numeral 12 indicates a line of written characters, and reference numeral 13 indicates an insertion symbol indicating a corrected entry. 14 indicates an additional character line to be additionally inserted. As shown in FIG.
When the vertical and horizontal projections are taken respectively in , the additional character line 14 and the added character line 12 must be written in such a position and shape that their projections are connected. In addition, in order to clearly indicate the additional position, the insertion symbol 13 is scanned and the highest point is set as the insertion position, so the insertion symbol 13 should have a protrusion or a convex part so that the addition position can be clearly seen. I see the need.

Character lines 12 and 14 are photoelectrically converted by the photoelectric converter 1 and stored in the image memory 2 as a plate pattern. At this time, as shown in FIG. 3, the vertical projection section 4 and the horizontal projection section 5 take horizontal projection and vertical projection of the sentence 5-hand lines 12 and 14. (Indicated by the f+ line in the figure.) In this case, since the insertion symbol 13 is written between the character lines 12 and 14, the vertical projection part 4 and the horizontal projection part 5 have a third
As shown in the figure, projections that are twice or more than the normal character pitch and line pitch are stored.

In this way, when scanning of a predetermined line is completed and its projection is taken, the recognition unit 6 scans the contents of the vertical projection unit 4 and the horizontal projection unit 5, and calculates the normal character pitch and line pitch. , when a double or more projection is detected, the recognition unit 6
A detection means for detecting an insertion symbol included in the insertion symbol is activated to confirm whether or not it is an insertion symbol.

That is, the detection means tracks all the quantized black bits of the image memory unit 2 corresponding to the vicinity of the center of the horizontal projection stored in the horizontal projection unit 5 (near the margin between lines) and detects that they are continuous. . This detection confirms that the projection stored in the vertical projection section 4 is not due to dust or dirt, but is due to the written 6- insertion symbol. Next, confirm that this insertion symbol is composed of one continuous line, has a convex part in the Y direction, and does not touch the character line. Next, the image memory section 2 corresponding to the vicinity of the center of the horizontal projection
is scanned in the horizontal direction and the number of points that change from white bits to black bits is counted. At this time, the position at which the image mesori section 2 is scanned is set to be a position corresponding to the vicinity of the center of the total horizontal projection, but the following may be used. A histogram is created by scanning the entire image memory section 2 horizontally and counting the number of black bits.

The position of the image memory section 2 corresponding to the position near the center of this histogram and where the amount of histogram is low is set as the full scanning position.

Half of the number of points that change from white bits to black bits counted as described above is the number where insertion 6 self-signs exist. 4% k of the point where the ox octopus changes from white bit to black bit
Remember it. Then, use this point as the starting point for line edge tracing or line edge circle tracing. All of the points that change from white bits to black pits, which are considered to belong to one insertion symbol, are all confirmed to be connected by a continuous line. .

In this way, the insertion symbol 13 is confirmed. The recognition unit 6 scans the inside of the image memory unit 2 and detects the position of the highest point X and Ym marks among the insertion symbols 13. Then, the position coordinates of this detected point are stored as the insertion position of all additional data.

The recognition unit 6 also recognizes the insertion symbol 13 and the additional character line 1.
4 is masked, all 1 sentences of the quantized pattern of character line 12 are quickly cut out into character patterns, all recognition is performed, and all recognition results are sequentially stored in a buffer (not shown) in recognition section 6. Similarly, the recognition unit 6 cuts out the quantization pattern of the additional character line 14 into character patterns of one sentence quickly, performs recognition, and sequentially stores all recognition results in the buffer. Then, the recognition unit 6 rearranges the above recognition results in all batzas based on the insertion position of the additional data that has already been stored, and stores the contents of the additional character line in a form that is fully inserted at the position where it should be inserted. .

Therefore, the above OCR has the following effects.

(1) Additions can be easily made to the input form in the form of a manuscript paper by using special symbols for addition insertion.

(2) Providing an additional row or field for each t\j The row one line below the row to be added can be used as an additional row.

As described above, according to this invention, data can be corrected without using additionality or additional fields by entering an insertion symbol and additional data into data that has already been entered in a form. An optical character reading device can be provided.

[Brief explanation of drawings]

FIG. 1 is a schematic configuration diagram of an embodiment of the present invention, FIG. 2 is a diagram showing an example of filling out a form in the embodiment, and FIG. 3 is an explanatory diagram for explaining the operation of the embodiment. DESCRIPTION OF SYMBOLS 1... Photoelectric conversion section, 2... Image memory section, 3-pass address control section, 4... Vertical projection section, 5... Horizontal projection 9-11'o-

Claims (1)

[Claims] a photoelectric conversion unit that photoelectrically converts the characters written on the form into a doji-ized pattern; an image memory unit that stores quantization patterns of at least two lines of characters in the form;
a vertical projection section that takes a vertical projection of the cardinalization pattern stored in the image memory section; a horizontal projection section that takes a horizontal projection of the quantization pattern stored in the image memory section; Based on the stored projection, the insertion symbol written between the character lines is detected, and the recognized character results in the two character lines are rearranged based on the position of the detected insertion signal. An optical character reading device comprising: a recognition section.