JPH03263282A - Character segmenting method for character reader - Google Patents

Abstract

PURPOSE:To enable character segmentation by using a normal medium by reading a slip, setting the positions of a reference character area and an objective character area, storing these positions into a memory and correcting the position of the read character area with these set positions as references. CONSTITUTION:An MPU 4 reads the slip to be used, extracts the character area, sets the positions of the reference character area and the objective character area and stores them into as memory 5 as format information. The MPU 4 reads out the format information from the memory 5. Next, the slip is read and a character area extraction program 7a is executed so as to detect the position of the reference character area. Then, a correct value extraction program 7c is executed so as to calculate error to a format reference position and to extract a correct value. The MPU 4 executes a character position correction program 7d and according to this correct value, the respective character area positions of the format are corrected. By using the objective area position, the character area is segmented from the picture data of a picture memory 3.

Description

[Detailed description of the invention] [Table of contents] Overview Industrial field of application Prior art (Figure 9) Means for solving the problems to be solved by the invention (Figure 1) Working examples (a) Characters Explanation of the reading device (Figure 2) (b) Explanation of the format creation process (Figures 3 to 6) (C) Explanation of the reading process (Figures 7 and 8) (d)
DESCRIPTION OF OTHER EMBODIMENTS Effects of the Invention [Summary] Character reading device for reading a form and recognizing target characters on the form; Second, regarding a character cutting method for cutting out a target character area on a form.

The purpose is to make it possible to cut out characters using ordinary media without using expensive OCR paper.

A character reading device having a reading unit that reads a form, an image memory that stores read form information, and a processing unit that cuts out a target character area from the form information and recognizes characters in the cut out target character area; , When creating the format, read the form, extract one character area, set the position of the reference character area and the position of the target character area from the reference position, store it in memory, and when reading, Extract the character area of the form, detect the position of the reference character area, find the error between the reference position of the memory and the detected position, and correct the target character area position of the memory with a correction value based on the error. to cut out the target character area.

[Industrial application field]

The present invention relates to a character reading device for reading a form and recognizing target characters on the form, and a character cutting method for cutting out a target character area on the form.

Optical character reading devices that read forms and recognize target characters on forms are widely used as input means for computers, etc., and handwritten character reading devices account for 2% of the total, because they can input handwritten characters on forms. There are two flights.

In such a character reading device, in order to accurately recognize a target character on a form, it is necessary to accurately detect the target character area on the form and cut out the target character area with high precision.

[Conventional technology]

FIG. 9 is an explanatory diagram of the prior art.

As shown in FIG. 9 (4), the conventional character cutting method involves determining the position x and y of the target character area from the edge with the edge of the form l as a reference, and then calculating the position of the target character area from the reading signal of the form to be read. The edges of the document are detected, and an area on the document at that position is extracted as a target character area based on the detected position and used as a target for single character recognition.

Another conventional method for cutting out characters is to attach a mark Mk or an L-shaped line mark Lmk to the form 1, as shown in FIG. It is possible to find the position of the target character area above, detect the mark Mk or line mark Lrnk from the reading signal of the form to be read, and cut out the area on the form at the relevant position as the target character area based on the detected position. It was being done.

(Problem to be solved by the invention) However, the conventional technology has the following problems.

■ In the former conventional method shown in Figure 9 (8), since the edge of the form 1 is used as a reference, it is necessary to cut the form 1 with high precision. In the latter conventional method in the 9th section, the line mark Lm
It is necessary to use special OCR paper with the mark Mk printed on it, which poses a problem of being expensive.

Therefore, the present invention can be used without using expensive OCR paper.
It is an object of the present invention to provide a character cutting method for a character reading device that can cut out characters using a normal medium.

[Means to solve the problem]

FIG. 1 is a diagram showing the principle of the present invention.

As shown in FIG. 1 (4), the present invention includes a reading unit 2 that reads a form 1, and an image memory 3 that stores the read form information.
and.

As shown in FIG. 1(B), a character reading device has a processing unit 4 that cuts out a target character area from the form information and recognizes the characters in the cut out target character area. Read form 1.

Extract the character area, and set the position of the reference character area and the position of the target character area from the reference position.

It is stored in the memory 5 as shown in FIG. 10.

When reading, extracts the character area of the read form.

The position of the reference character area is detected, the error between the reference position of the memory 5 and the detected position is determined, and the target character area position of the memory 5 is corrected with a correction value based on the error to cut the target character area. It is something to put out.

[Effect]

The present invention corrects the position of a target character area using a character area such as characters already printed on a form 1 as a reference.

For this purpose, when creating the format, read form 1 and read form 2.
The character area is extracted, the position of the reference character area and the position of the target character area are set, and stored in the memory 5 as format information.

When reading, the character area of the read form is extracted, the position of the reference character area is detected based on the extraction result, and the error between the detected position and the format reference position of the memory 5 is determined.
This is used as a correction amount to correct the position of the target character area.

In other words, in form 1 such as a slip, the positional relationship of each character area is fixed depending on the printing accuracy. Therefore, the position of the characters printed in kanji such as the title that has already been printed in this character area is used as a reference, and correction is made based on this. This is how it was done.

Therefore, it is possible to read commonly distributed slips without line marks and general slips that are not cut with high precision.

Therefore, it is possible to handle documents not exclusively for OCR, such as general slips, from media exclusively for OCR, and low-cost media can be used.

In addition, single-character reading devices are extremely easy to install because they can read the slips, etc. that are currently in use.

〔Example〕

(a) Description of character reading device FIG. 2 is a block diagram of an optical character reading device to which the present invention is applied.

In the figure, the same thing as shown in Figure 1 is 11 yen I-
It is indicated by the symbol.

The scanner 2 is composed of a CCD line sensor.

While sending the form 1, it is read and converted into two image signals.

The processing unit 4 is composed of a microprocessor (MPU) and executes each program.

5 is a data memory and a reference position (position reference).

Something that stores the position, etc. of the target character area, correction values, etc., 6
is a control program memory, which stores a control program for the MPU 4.

A program memory 7 stores programs necessary for the character reading process of the MPU 4.

That is, a character area extraction program and its work memory 7a extract a character area from the image information of the form 1, a position reference setting program 7b (FIG. 3) that sets a position reference on the form, and a correction value at the time of reading is extracted. A correction value extraction program 7C (FIGS. 7 and 8), a character position correction program 7d that corrects the target character (area) position using correction values, and a character recognition program 7e that cuts out two character areas and recognizes two characters.
and store it.

8 is a display input section, which includes a display section 8a for displaying images and characters, a keyboard 8b for inputting data, etc.
The input position on the display section 8a is indicated by 1', which has a mouse 8C, and 9 is a bus.

MPU 4 and scanner 2. It connects the memories 3, 5, 6, and 7 degrees to the display input section 8.

Furthermore, what is the data memory 5 and the program memory 7?

It consists of one magnetic disk device.

(b) Description of format creation processing FIG. 3 is a flowchart of format creation processing according to an embodiment of the present invention, FIG. 4 is an explanation diagram of the skew amount extraction process, FIG. 5 is an explanation diagram of the character area extraction process, The figure is an explanatory diagram of the position setting process.

■ Input the form to be read into the scanner 2.

The MPU 4 stores the read image data in the image memory 3 under execution of a control program.

(2) Next, the MPU 4 executes the one-character area extraction program 7a, and first calculates the skew amount θ of the form 1. This will be explained with reference to FIG.

As shown in FIG. 4(A)(2), the skew amount θ is the inclination angle when the scanner 2 reads the form 1.

The MPU 4 counts the number of black dots for each line of image data in one image memory 3, as shown in FIG. 4, and creates a histogram of the counted values.

This is scanned by changing the scanning angle as shown in Figure 4 0.
Obtain histograms h1 to h3 for each angle.

Next, the MPU 4 calculates the amount of change from white to black on each histogram, compares them, finds the one with the steepest change, and sets the scan angle at that time as the skew amount θ.

(2) Next, the MPU 4 continues the two-character area extraction program 7a to extract a two-character area.

This will be explained further in Fig. 5).

The MPU 4 stores the image memory 3 at the scan angle of the determined skew amount.
is scanned, and black is projected in the horizontal direction to obtain a projected image ph.

As shown in Figure 5, the two-character area is projected black.

The rest will be white.

Therefore, the MPU 4 scans the projected image Ph from above, and extracts 9 character areas as shown in 10.11.12, assuming that white parts longer than a specified length are between characters. Each character area 10°11.12
In contrast, scan using the skew amount.

Black projected image Pv1 r Pv21 P as shown in FIG.
Get v3.

These black projected images Pvl, Pv2. Pv3 is scanned from left to right in the diagram, and similarly, a white portion longer than a specified length is considered to be outside the character area, and the vertical character area is determined.

In this way, as shown by diagonal lines in FIG.
Extract 11.12.

(2) Next, the MPU 4 uses the skew amount for each character area 10, 11, and 12 in the same manner as in FIG.

In the horizontal direction (two projections are made, the smaller specified length is used, and the white part longer than the specified length is regarded as the character line spacing, and two character lines are extracted.

Further 1: For each character line, the MPU 4 projects each character line in the vertical direction using the same method as shown in FIG. 5 and cuts out the characters within two lines.

In this manner, the coordinates of the four corners of the cut-out rectangular area of each extracted character area are stored in the work memory 7a.

(2) Next, the MPU 4 executes the 2-position reference setting program 7b to set the reference position and the like.

This will be explained with reference to FIG. 6.The O MPTJ4 displays 1. For example, as shown in FIG. 6, if the input form 1 is displayed under "Title",
If there is a data field to be read, the 2 display images will be 1 title, the contents of the data field to be read, and a rectangular frame surrounding it.
It is created from the character area coordinates extracted in .

If the character frame of the read data field of the input form is a dropout color, write it on the form in a color that will not drop out, as shown in Figure 6, so that the character frame position can be entered as an image in advance. This display image shows the relationship between the contents of the input form and the area cut out by the extraction process.

Next, the operator uses the mouse 8C to input an instruction for a standard character rectangular frame on the screen of the display section 8a.

This reference character rectangular frame (secondly, it is necessary to select a Kanji title character, etc. at a position that does not change in relation to the position of the character entry frame in which the data to be read is written).

That is, on a slip, etc., title characters such as "Delivery Note" are printed, and this character area is printed.

Since it is constant, it can be extracted at a fixed position with respect to the entry frame of the data to be read.

When this reference frame is input, the MPU 4 recognizes the characters within the reference frame using the character recognition program 7d and displays the result of the two character recognition as an answer below the reference frame on the second display section 8a.

The operator visually compares and confirms this answer with the characters within the reference frame on the display section 8a, and if the answer is correct, presses the mouse 8.
Check with C and if it is incorrect.

Enter the correct answer using the keyboard 8b and correct it.

As a result, the MPU 4 uses the upper left and lower right coordinates of the two selected reference character rectangular frames as reference positions.

It is stored in the work memory of the data memory 5 together with the character code within the reference frame.

(2) Next, the operator uses the mouse 8C to specify and input the upper left and lower right coordinates of the target character area as the target character area position from the above-mentioned display image on the second display section 8a.

The MPU 4 converts these coordinates into a relative position from the reference position, and sets it as the target character area position.

Furthermore, the operator inputs the character type and number of characters of the target character area using the keyboard 8b.

After these inputs, the MPU 4 converts the reference position data and target character area position data into values with the skew amount θ being "0", creates format information, and stores the reference position and its character code in the data memory 5. In the memory for reference storage,
The position of each target character area, its character type, and number of characters are stored in the target character area position memory of the data memory 5.

In this way, the form to be used is input, one print character area of the form is used as a reference position, and the position of each target character area from the reference position is created as format information.

Creating format information with the operator's intervention in this way allows the operator to freely select the reference position suitable for the form, and excludes items other than the target character area, such as the company name in the document, from the target character area. It's for a reason.

(C) Explanation of reading process FIG. 7 is a flowchart of a reading process according to an embodiment of the present invention.

FIG. 8 is a flowchart of the reference position correction value extraction process of FIG. 7.

■ Prior to reading the form, the MPU 4 reads format information from the data memory 5.

■ Read the form as explained in Figure 8.

The character area extraction program 7a is executed to detect the position of the reference character area, the correction value extraction program 7c is executed, the error with the format reference position is determined, and the correction value is extracted.

The MPU 4 executes the 2-character position correction program 7d, and uses this correction value to correct the position of each target character area in the 7-ohmapto.

(2) Then, the MPU 4 uses the corrected target character area position to determine the position where the character to be read exists, and cuts out the character area from the image information (image data) in the two-image memory 3.

The MPU 4 executes the two-character recognition program 7e.

Recognize the characters in the character area cut out from the number of characters in the two character types of the format.

The MPU 4 checks whether all character recognition has been completed, and if it has not been completed, the process returns to the beginning of step (2).

On the other hand, when the MPU 4 determines that all the characters have been completed, it determines whether there is a second page of the form, and if there is, it returns to step ■, otherwise it ends. Next, the reference position correction value extraction process is performed according to FIG. ○ ■ Scanner 2 reads form 1 and stores it in image memory 3 as image data.

Next, the MPU 4 extracts the skew amount θ using the same method as step (2) in FIG.

Furthermore, the MPU 4 extracts a character area using the same method as in step (2) in FIG. 3, and then extracts a character line within one character area using the skew amount in the same manner as in step (2) in FIG.

Next, the MPU 4 calculates the difference between the upper left coordinate of each extracted character line and the upper left coordinate of the standard position of the format, and selects the character line with the smallest difference as the character line closest to the standard position.

@ The MPU 4 cuts out the inline characters of the 2 selected character lines using the same method as in step 3 of FIG.

Then, the MPU 4 performs character recognition on the extracted in-line character line using the same method as in step (2) in FIG.

■ If the comparison results show that they are equal, the MPU4.

Calculate the deviation value between the detected coordinates obtained by cutting out in-line characters and the standard position image of the format.

Next, the MPU 4 adds the skinny amount correction to the deviation value and outputs it as a reference position correction value.

■ On the other hand, if you show that the comparison results are not equal.

The MPU 4 selects the next closest character line to the character line found in step (2).

Then, the MPU 4 determines the amount of deviation between the detected coordinates (Y) of the character line and the standard position coordinates (Y) of the format, and compares this amount of deviation with a predetermined limit value.

If the amount of deviation is smaller than the limit value, the process returns to step (2); if the amount of deviation is larger than the limit value, it is designated to be treated as rejected and the process ends.

In this way, the reference character area of the form to be read is detected, the deviation from the standard position of the format is calculated, the correction value due to the deviation of the form is calculated, and the position of the target character area of the format is corrected. This will cut out the characters.

For this reason, 2 normal slips are used, and the printed character position of the slip title, etc. is used as a reference, regardless of the slippage of the slip.

The target character area can be extracted.

In this embodiment, the same algorithm as used for position extraction during format creation is used to detect the position of the reference character area of the form, so two position detection can be performed with the same precision.

Furthermore, characters within the detection reference character area of the form are recognized.

Because it is compared with the character code of 7i-mat.

Detection becomes more reliable.

(d) Description of other embodiments In the present invention, a quadratic modification is also possible.

■ When creating a format, extract the standard character area and
Although selection is made by displaying the reference character area superimposed on the image, it is also possible to display the form image and allow the operator to specify and select the position of the reference character area using a mouse or the like.

Conversely, the position obtained by extracting the reference character area may be directly adopted as the reference character area position without any instruction from the operator.

(2) In step (2) of FIG. 3, the operator inputs the position of the target character area, but the coordinates obtained by cutting may be used as the position of the target character area.

(2) In step O2 (2) of FIG. 8, character recognition is performed in the area set as the reference area and compared with the 7-orbit character code, but this may be omitted.

Although the present invention has been described above with reference to examples, the present invention can be modified in various ways according to the gist of the present invention.

These are not excluded from the present invention.

〔Effect of the invention〕

As explained above, the present invention provides the following effects.

■ Characters are extracted by correcting the position of each target character area based on the character position 1 in the character area of the form, so O
This has the advantage that commonly available slips and the like, which are not exclusively for CR, can be used as reading media, and low-cost media can be used, and character reading devices can be easily installed.

■ It is easy to create a format by having the user read the slip, etc. to be read.

[Brief explanation of drawings]

FIG. 1 is a diagram showing the principle of the present invention. FIG. 2 is a configuration diagram of an optical character reading device to which the present invention is applied. FIG. 3 is a flowchart of i-mat creation processing according to an embodiment of the present invention. FIG. 4 is an explanatory diagram of the skew amount extraction process in FIG. 3. FIG. 5 is an explanatory diagram of the character area extraction process in FIG. 3. FIG. 6 is an explanatory diagram of the position setting process in FIG. 3. FIG. 7 is a diagram illustrating a reading process 70 according to an embodiment of the present invention. FIG. 8 is a flowchart of the reference position correction value extraction process in FIG. 7. FIG. 9 is an explanatory diagram of the prior art. In the figure, 1... form. 2...Reading unit (scanner). 3... Image memory. 4...processing section. 5...Data memory. (B) (C) Foramatsu 1 I-Hachiyuri 70-Figure 3 Kominami character division ``Yokosaka prisoner of the Xion device Figure 2 225 Cumulative 1 count and stop plum /'P/ -]--''- m- Flowchart for return to request 9 Figure 7, hh-alJ-7 Explanation of punishment 9i Maruri g

Claims (1)

[Scope of Claims] A reading unit (2) that reads a form (1), an image memory (3) that stores read form information, a target character area that is cut out from the form information, and a character of the cut out target character area. A character reading device having a processing unit (4) that recognizes a character, when creating a format, reads the form (1), extracts a character area, and determines the position of the reference character area and the target character area from the reference position. Set the position of the form and store it in the memory (5), and when reading, extract the character area of the read form (1),
The position of the reference character area is detected, the error between the reference position of the memory (5) and the detected position is determined, and the target character area position of the memory (5) is corrected with a correction value based on the error. A method for cutting out characters for a character reading device, characterized by cutting out a character area.