JPS594068B2 - Character detection cutting device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a character detection and cutting device in an optical character reading device.

In general, character reading devices detect and cut out characters one by one from a character string written on a form, for example, and perform recognition.

Among these, the following methods are conventionally known as means for detecting and cutting out characters. (C) Set frames at predetermined intervals on the form in advance and write the characters within the frames.

Then, each character information is cut out by detecting the frame on the form. (2) Write "L", "-", "・" on the form in advance.
A character or symbol such as "" is written as a reference position, and the character is written in a predetermined area based on the reference position. Then, the above-mentioned reference position on the form is detected, and each piece of character information is cut out based on a relatively constant area. (3)
One side edge of the form is set as a reference position in advance, and characters are written in a predetermined area on the form based on the reference position.

Then, one side edge of the form is detected and each piece of character information is cut out based on an area that is relatively constant with that edge. (4) Detect one end of the character string written on the form, and cut out each character information based on a certain area from that position.

15(5) As shown in Figure 1, create a vertical projection J of the character string written on the form (obtaining the horizontal width of the character string and each character), and use this to create a cutout area. Detect and extract each character information.

(6) Detecting and cutting out information on each of the 90 characters using any of the methods (c) to (5) above, and detecting and cutting out a large number of pieces of character information in the vicinity thereof, and extracting each of these character information By recognizing and comparing the characters, an accurate extraction area is set again, and accurate character information is extracted.25 However, each of the above-mentioned means has its own problems.

In other words, with the means (1) to (3), the printing accuracy of characters on the form is severely limited, and the aesthetic appearance of the form is impaired, especially by setting the frame in (C). (4) ), the pitch of the characters written on the form must be kept constant, or the cutout area must be varied according to the pitch of the characters written on the form, which is troublesome, and furthermore, the characters written on the form have to be changed. If it is made by a line printer, the cutout area 35 may be completely different depending on the noise around the characters.With the method (5), for example, katakana "so", "tsu", "ha",
When characters of a type consisting of a plurality of blocks are consecutive, such as ``Ritama ``RU'', there is a risk that the cutout area K may be mistaken as shown in FIG. In this case, even if the character is made up of ordinary single blocks, the printing quality of the character may deteriorate and the line width or character spacing may change, or if there is noise around the character, the cropping area may be incorrectly identified. There is a fear. (
With the method 6), recognition operations are required many times for one piece of character information, which increases the time required to extract accurate character information. In this case, if an attempt is made to shorten the required time, the hardware configuration will become complicated, leading to an increase in cost. This invention was made in view of the above-mentioned circumstances, and its purpose is to accurately detect characters in a short time regardless of the position, print quality, type, etc. of the characters on the recording medium. An object of the present invention is to provide a character detection and cutting device capable of cutting out characters. An embodiment of the present invention will be described below with reference to the drawings.

In FIG. 3, reference numeral 1 denotes a cathode ray tube (hereinafter abbreviated as CRT) such as a flying spot scanner, in which light emitting points on the screen are sequentially moved by coordinate commands from a control section 2. The light spot of this CRTl is projected onto a recording medium, for example, a bulletin board 4 through a lens 3. In this case, the light spot projected onto the form 4 by CATl becomes a line scan that moves downward from the upper left part of the form 4 in the figure, that is, in the direction of arrow Y, and the line scan moves sequentially in the direction of arrow X. I'm starting to go there. Then, the form 4 obtained by such flying spot scanner scanning is
The reflected light is converted into an electric signal by a photomultiplier tube 5, and the electric signal is converted into a digital signal by a quantization circuit 6. The output of this quantization circuit 6 is
When the scanning position corresponds to a character on the form 4, it becomes a so-called black block signal (logic 61' signal), and when it does not correspond to a character, it produces a so-called white block signal (logic 10' signal).
The signal is supplied to the first counter 7 and the pattern register 8, respectively, and to one input terminal of the OR circuit 9. The first counter 7
is a scanning clock signal (issued for each scan) supplied from the control unit 2 only during the period when a black block exists.
The contents are supplied to the control unit 2 as a vertical projection on the characters on the form 4, and
It is designed to be cleared each time one scan is completed by a clear signal supplied from the control section 2. The pattern register 8 forms a quantized character pattern based on the output of the quantization circuit 6, and its contents are recognized by a recognition section (not shown).
On the other hand, the output of the OR circuit 9 is transmitted to the horizontal projection register 10.
The contents of this register 10 are supplied to the OR circuit 9.
is supplied to the other input terminal of the . This register 10 temporarily stores the output of the quantization circuit 6, and by circulating the contents through the OR circuit 9, the black block signal and white block signal for each scan are stored additively. It is becoming more and more common. The contents of this register 10 are cleared by the control operation of the control section 2 every time one character is scanned. In this way, a horizontal projection of each character is formed in the register 10, which is supplied to the second counter 11. This second counter 11 counts the scanning block signal supplied from the control section 2 only during the period when the black block exists, and its contents are supplied to the control section 2 and also indicates when one scanning is completed. It is designed to be cleared by a clear signal supplied from the control section 2. The control section 2 is composed of, for example, a microcomputer, and is equipped with a storage section (for example, a random access memory) having a plurality of addresses.

Three addresses called a black address (BLACK), a white address (WHITE), and a total address (TOTAL) are specified in the storage section. The black address is the number of scans corresponding to consecutive black blocks, the white address is the number of scans corresponding to consecutive white blocks, and the total address is the total number of scans. The control unit 2 also has a black flag (BLACK-FF) for detecting black blocks.
is provided so that the status of whether or not there was a black block in the previous scan is always stored. In this way, the control section 2 compares the contents of the first counter 7 and the second counter 11 with the contents of each address and black flag of the storage section based on various preset conditions, thereby controlling the The system provides appropriate control over the detection and extraction of Next, the operation in the above configuration will be explained.

This case will be explained with reference to the time chart of FIG. 4 and the flow chart of FIG. 5. First, prior to the start of scanning the form 4, in step 50, the register 10 is cleared by a clear signal output from the control section 2, and the black address, white address, and total information in the storage section of the control section 2 are cleared. Each address is cleared and a black flag is set. Furthermore, in step 51, the first and second counters 7 and 11 are cleared, respectively, by the clear signal of waveform a output from the control section 2. And the above waveform a
When the clear signal ends, the waveform C is sent from the control unit 2 to the CRT1.
An OY direction scanning signal is supplied. In the Y-direction scanning signal, the period between C1 and C2 shown in the figure becomes a scanning period for one scan, and the period between C2 and C3 shown in the figure becomes a return signal for the next scan due to the retrace signal of waveform b output from the control section 2. Line period. On the other hand, the X-direction deflection signal increases by a predetermined voltage every time the clear signal of waveform a ends, and the
Shifts the scanning position for 1 bit in the X direction in the figure. In this way, a scanning clock signal of waveform e is outputted from the control section 2 for each scan in step 52, and the scanning clock signal is counted by the first counter 7 only during the period of the black block signal outputted from the quantization circuit 6. be done.

Further, the scanning clock signal is counted by the second counter 11 only during the period of the black block in the register 10. The control unit 2 performs the following control operations. First, each time one scan is completed, step 5 is performed.
3, the total address of the storage section is incremented by one. The content of this total address, that is, the total number of scans is
In step 54, it is compared to see if it is greater than the constant T1, and if it is greater, it is rejected (REJE) in step 54.
CT), making detection and cutting impossible. The constant T1 is the number of scan lines corresponding to about 1.3 times the character pitch. On the other hand, if the content of the total address is less than or equal to the constant T1 in step 54, it is compared in step 56 to see if the content of the first counter 7 is greater than the constant T2, and if so, the character block is is judged to exist. The constant T2 is the number of scan lines corresponding to the minimum width of a black block that can be determined to be a character block. and,
At step 57, the black address is incremented by one. The content of this black address is compared in step 58 to see if it is greater than a constant T3, and if it is greater, it is rejected in step 59, making detection and extraction impossible.
The constant T3 is the number of continuous scans of black blocks approximately corresponding to the character pitch. In step 58, if the content of the black address is less than or equal to the constant T3, the black flag is checked in step 60, and if the black flag is set, that is, if there is no black block in the previous scan, the current black It is determined that the block has changed from the white block. Thereafter, a black flag is set in step 61 due to the current black block. Furthermore, in step 62, the content of the white address is compared to see if it is greater than the constant T4, and if it is, the black data up to the present time is regarded as noise, and in step 63, the content of the black address is set to "1".
be made into In this case, in step 63 above, the contents of the black address are not cleared, that is, set to "0", but set to "1".
The current black block becomes the beginning for the next scan. The constant T4 is the number of scan lines that corresponds to a value larger than the distance between each plot of a character consisting of a plurality of plotters. On the other hand, in step 62, if the content of the white address is less than or equal to the constant T4, the black block up to the present time is not regarded as noise, and therefore the content of the black address is not changed. Then, in step 64, the contents of the white address are cleared and a 1-bit shift is performed to the next scanning position. If the black flag is set in step 60, it is determined that black blocks are continuous, and the process proceeds to step 64. In this way, the above control operation is repeated, but at step 56 during the control, if the content of the first counter 7 is equal to or less than the constant T2,
It is determined that the area is not a character block, that is, a white block.

Then, in step 65, the white address is incremented by 1 and the black flag is set. Thereafter, in step 66, the content of the second counter 11 is compared to see if it is less than or equal to the constant T5, and in the following cases, the process moves to step 67. In this step 67, the contents of the black address are compared to see if they are less than or equal to the constant T6, and in the following cases, the black blocks up to the present time are regarded as noise, and the contents of the total address are loaded into the white address in step 68. At the same time, register 10 is cleared. In other words, this determines that everything up to this point is a white block. The constant T5 is the height of the black block which may be considered as noise. Further, the constant T5 is the number of scans corresponding to the width of the black block, which may be regarded as noise. If the content of the counter 11 is not less than the constant T5 in step 66, if the content of the black address is not less than the constant T6 in step 67, and after step 68 has passed, step 6 is executed.
Move on to 9.

In this step 69, the content of the total address is compared to see if it is greater than the constant T7, and if so, in step 70, the content of the white address is compared to see if it is greater than the constant T8. . The constant T7 is the minimum number of scans that can complete character scanning, and corresponds to about half the character pitch. The constant T8 is the number of continuous scans of white blocks whose width should be blanked, and corresponds to the approximate character pitch. Further, if the comparison result in step 70 is negative, the content of the white address is compared in step 71 to see if it is the same as the constant T. The constant T is the character to be read shown in FIG. In the example, the character consisting of multiple blocks "
This is the number of scanned lines that corresponds to a value that is larger than the distance between the blocks of "T" and "S" and smaller than the minimum value of the adjacent character spacing. If the result of the comparison in step 71 is negative, then in step 72 it is compared to see if the content of the white address is the same as the constant TlO. The above constant Tl
O is 6L0 (large), ゛M'' (medium), ゛S゛ (small) for the vertical projection (height) and horizontal projection (width) of the character example to be read in Figure 6. ), in the table below, the vertical projection is 1L,
This corresponds to a character consisting of a 1S'' block, a 1S'' block, a 1S'' plotter, and a 6M'' block, and the number of scan lines corresponds to an appropriate value less than the minimum value of the adjacent character interval. Further, if the comparison result in step 72 is affirmative, it is compared in step 73 to see if the contents of the black address are greater than constant Tll and less than constant T,2.

The above constants Tll, Tl, correspond to the characters in the above table whose vertical projection consists of a block of "L", a block of "S゜" and a block of "S1", and a block of "S゛" and a block of 6M1. Among these, the constant Tl
l is the number of scanning lines corresponding to a value smaller than the minimum value of the sum of the widths of the vertical projections of each block, and the constant T, 2 is the number of scanning lines corresponding to a value slightly smaller than the character pitch. If the comparison result in step 73 is positive, one character scanning is completed, the quantized character pattern in the pattern register 8 is recognized by the recognition section (not shown), and the next one character scanning is started. Move. In this case, if the comparison results at steps 69, 72, and 73' are negative, the process returns to step 51 without performing detection extraction. Furthermore, in the comparison result of step 70 above,
If the content of the white address is larger than the constant T8, it is determined in step 74 that it is a blank area, and one character scanning ends. If the comparison result in step 71 is affirmative, step 75 compares the contents of the bratta address to see if it is greater than constant Tl3, and if so, step 76 compares the contents of the second counter 11. is compared to see if it is equal to or less than the constant Tl4.

The above constant T, 3 corresponds to the letter "S" in which both the vertical projection and horizontal projection widths are smaller than the minimum value in the table above. This is the corresponding number of scans.Also, the constant Tl
4 corresponds to the character 6S'' in the horizontal projection block size in the above table, and corresponds to a value whose horizontal projection height is slightly smaller than the maximum value. Then, if the comparison result in step 76 is positive, one character scanning is completed, the quantized character pattern in the pattern register 8 is recognized by the recognition section, and the next one character scanning is started. . In this way, the detection and extraction of the character J consisting of multiple blocks such as "B" is completed. In this case, if the comparison results in steps 75 and 76 are negative, the process returns to step 51. If the comparison result in step 72 is negative, step J
In MoV, the contents of the white address are compared to see if they are the same as the constant Tl5.

The constant Tl5 is the number of scan lines corresponding to a value larger than the maximum value of the inter-block distance of a character whose vertical projection consists of two blocks in the above table. Then, if the comparison result in step JMoV is affirmative, step 78 is performed.
The contents of the black address are compared to see if they are greater than the constant T, 6. The above constant Tl6 is, at the soil surface,
The vertical projection block size corresponds to a character of 6M1, and the number of scan lines corresponds to a value smaller than the minimum character pitch. If the comparison result in step 78 is affirmative, step 79 compares the content of the second counter 11 to see if it is greater than the constant Tl4. If the comparison result in step 79 is affirmative, one character scanning is completed, and characters consisting of a plurality of blocks, such as "Ha" and "R", are detected and cut out. In this case, if the comparison results at steps J, V, 78, and 79 are negative, the process returns to step 51. Here, to summarize the above-mentioned control operation with respect to various detection conditions: [Reject condition] 7T0TAL>Tl 4BLACK>T3 [Noise condition] 0 (BLACK<T6)△(Counter 11 x T5) 1
After (WHITE>T4), when a black block is reached, the previous black block is regarded as noise.

[Blank detection conditions]

4 (TOTAL>T7)△HTE>T8) [Character detection conditions] (TOTAL>T7)A(WHITE-Tl
O) △(TllkuBLACK<Tl2)1(TOTAL
>T7)A(WIIITE=T9)△(BLACK>T
l3)△(second counter 11≦Tl4) @ (TOTA
L>T7)△HTE=Tl5)△(BLACK>Tl
6) △vine 2 counter 11>Tl4).

Thus, the characters are accurately detected and cut out, and the character pattern in the pattern register 8 is recognized.

Then, the above detection conditions will be explained. As is clear from the above table, the example characters to be read in Figure 6 are classified into five types of character groups by vertical projection, and the characters ``ha'', ``l month'', ``ru'', and ``, which are composed of multiple blocks, are classified into five types of character groups by vertical projection. TS”, “So” baS”
This block consists of an Sn plotter or a 6sn block and an "M" block. However, since there are characters that are composed of only one block of ``S'' or one plotter of 6M'', accurate detection and cutting cannot be performed using only the width of the black block in the vertical projection.For this reason, In this embodiment, horizontal projection is used in addition to vertical projection, and each block spacing and block height are obtained by these projections.In other words, among the character detection conditions described above, the corresponds to a character whose vertical projection block width is 1L''.Furthermore, condition 1 corresponds to a character whose vertical projection block width and horizontal projection block height are both 0S''. “]”, “・”, “I”, [゛”, “0”,
"s" corresponds to this. Condition 7 corresponds to characters whose vertical projection block width is 6M. By the way, among the characters whose vertical projection consists of two blocks, for example, for "T" and "S", when the scanning position reaches the middle of each block, step 71 is executed.
WHITE-T9 or TOTAL of step 69
>T7 is not affirmed, so conditions 7 and 7 may not be met.

Similarly, among the characters whose vertical projections consist of two plotters, for example, "c", "January", and "ru", when the scanning position reaches the middle of each plotter, the counter 11 of step 76 < Since Tl4 or WHITE=Tl5 of step JMoV is not affirmed, the above condition 7,
7 may not be compatible. In such a case, at the end of scanning the two blocks, the cumulative width of the two blocks (
Since the content of the black address) is greater than the constant Tll and less than the constant Tl2, that is, step 7
Detection cutting is performed by corresponding to Tll of 3 and BLACK<Tl2. Although the above embodiment employs a flying spot scanner scanning method, the scanning method is not limited, and may be, for example, as shown below.

(1) A method in which the semiconductor wire scanner is moved parallel to the character string.

(2) A method in which a laser beam is raster scanned over a character string using a rotating reflector.

(3) A line scanner with a field of view the width of a character string is fixed, the form is scanned by the line scanner while moving in a direction perpendicular to the character string, and the character pattern resulting from the scanning is stored in a storage device. The character pattern stored in the storage device is read in the same manner as the scanning method used in the above embodiment.

In addition, the height of a character block is determined by creating a horizontal projection. The difference between and the minimum value may be calculated for each character block.

As described above, the present invention is not limited to the embodiments described above, and it goes without saying that various modifications can be made without departing from the gist of the invention.

As described above, according to the present invention, the vertical projections are counted by the first counting means, the horizontal projections are counted by the second counting means, and the counting results of the first and second counting means are combined. Since the width, height, and block spacing of the blocks that form characters are determined accordingly, each character information can be accurately recorded in a short time regardless of the position of the character on the recording medium, print quality, and type. A character detection and cutting device capable of detecting and cutting out characters can be provided.

[Brief explanation of drawings]

1 and 2 are diagrams showing the vertical projection of characters and their cutout areas, FIG. 3 is a block diagram showing an embodiment of the present invention, and FIG. 4 is a diagram for explaining the operation of the embodiment. A time chart, FIG. 5 is a flowchart for explaining the operation in the same embodiment, and FIG. 6 is a diagram showing character strings to be read. 1... Cathode ray tube (CRT), 2... Control unit, 3... Lens, 4... Form, 5
...Photomultiplier tube, 6...Quantization circuit, 7...First counter, 8...Pattern register, 9... OR circuit, 10...
- Horizontal projection register, 11...Second counter.

Claims (1)

[Claims] 1. In a character reading device that reads characters by moving vertical line scans along a character string, a first counting means counts vertical projections and a second counting means counts horizontal projections. and the width and height of the block forming the character according to the counting results of the first and second counting means,
A character detection and cutting device comprising: a control means for determining a block interval.