JPH0122958B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a character reading and recognition device, and more particularly to a character reading and recognition device that can accurately recognize characters even when ink stains or the like exist between characters. .

For example, when writing the account number and sometimes the amount on checks and bills, use JIS E13B/MICR.
(Magnetic Ink Character Recognition) characters, which are automatically read and organized. This MICR character has a shape as shown in Figure 1 A, and although the character width l ₁ , l ₂ , etc. differs depending on the character, when typing this, the distance between the characters l is constant. has been done. One way to read and recognize these characters using OCR is to scan the characters over a certain period of time.
The results are stored in a buffer or the like. Characters are then recognized by comparing them with standard patterns in a separately prepared dictionary.

When character recognition is performed using the mask matching method in OCR reading in this way, when there is a blemish between the characters, the blemish is large and does not touch one of the characters. This may cause misreading or inability to read and recognize the information.

For example, when reading the character C in Figure 2, each bit string is scanned in the direction of the arrow, starting from the line L where the first black point (character information) of the character C exists, and the results are stored in a buffer or the like. . After scanning one column, the next column is scanned. Then move by the width of the character where the black dot exists, and then
_P1 is indexed to determine the size of the character, which is compared with the standard patterns of known characters stored in a dictionary, and those that are most similar are recognized as that character.

However, in Fig. 2, when the character C has a stain D, the data is read by the length of l ₃ ' as the character width, so the center point of the character is
A pattern is defined as P ₂ and compared with the standard pattern stored in the dictionary. Therefore, if this stain D is not attached to the letter C, it may be possible to identify it as a stain in some cases because it is surrounded by white areas.
If it is in contact with the character C, this will also be read as character information. If this stain D is large, the center of the character will be P ₁ because of this stain D.
There is a large shift from P 2 to P ₂ , which means that the character patterns will not be compared correctly.
As a result, misreading may occur or the information may become unreadable.

Therefore, in order to prevent this, it is conceivable to read the reading width by the width l ₁ of the widest character, and compare each character with data limited to its width l ₁ . However, as shown in FIG. 3, when there is a large stain D ₀ attached to the character C ₂ between the characters C ₁ and C ₂ , the following problem exists.

That is, in Fig. 3, for the character C ₁ , the width l ₁
After reading data for the amount of , data for the next character C ₂ is read for a width l ₁ from the position of the line L ₀ to L ₀ ′ due to the black spot D ₀ . However, if the black dot D ₀ is connected to another black dot D′ indicated by a dotted line, the next black dot area will be immediately drawn from the line L ₁ that has read the data of width l ₁ for the character C ₁ . It exists continuously and has a width of l ₁ and is read as the next character, and the black dot area from line L ₁ ' is read as data for the third character. As a result, regardless of whether it is recognizable or not, there is one character on the line L ₁ - L ₁ ',
This is read out as so-called 3-digit data in which one character exists on the line L ₁ ′-L ₁ ″.

Therefore, in order to solve this problem, the character reading recognition device of the present invention sets an area between adjacent characters in which no data is read, and even if there is a black dot area where a black dot exists, this area is not read. This is what I tried to do to ignore it. For this purpose, the character reading recognition device of the present invention compares a character pattern obtained from character information obtained by reading a character with an optical reading means with a well-known standard pattern to recognize a read character. A storage means for storing character information read by the optical detection means, a means for detecting a predetermined unit width of the character information from the storage means, and a period from the predetermined unit width to the beginning of the next predetermined unit width. The present invention is characterized by providing means for prohibiting text information of.

Next, before detailed description of the specific configuration of the present invention,
The outline will be explained based on FIG.

In the present invention, a read data input prohibited area _IH is provided after character _C4 is read by a width _l1 , and the read information in this read data input prohibited area is not used. In this case, the width _l0 of the read data input prohibited area is determined to be, for example, about 1/2 to 2/3 of the character interval _l2 . As a result, even if there is a large blemish D ₀ , it will not be read entirely, but only the part after line L ₂ will be read, and data will be read as the next character by a width l ₁ from line L ₂ . Therefore, even if the dot D' is continuous from the character _C1 as shown in FIG. 3, this will prevent the 2-digit character from being misread as a 3-digit character. Furthermore, even when comparing the character patterns of the character _C2 , only a part of the stain _D0 has an adverse effect, so that the recognition rate can be improved.

Next, one embodiment of the present invention will be described based on FIG.

In the figure, 1 is a buffer, 2 is a preprocessing circuit, 3 is a gate, 4 is a vertical positioning circuit, 5 is a spot detection circuit, 6 is a horizontal positioning circuit, 7 is a gate control circuit, 8 is a BOX creation circuit, 9 is a standard mask circuit,
10 is an arithmetic circuit.

Here, the buffer 1 temporarily holds character information transmitted from an optical system (not shown), and the preprocessing circuit 2 converts this transmitted character information into "1" and "2", for example. This is to perform binarization processing to facilitate processing such as positioning, which will be described later.

The gate 3 is connected to the preprocessing circuit 2 only during the width _l0 in order to provide the read data input prohibited area _IH shown in FIG.
This prevents data from being sent from.

The vertical positioning circuit 4 detects a black dot area in the read data in which a black dot (character and black dot part) exists in the vertical direction, determines the position of the character area, and transmits the character data to the horizontal positioning circuit 6. It is.

The black spot detection circuit 5 identifies whether the data output from the gate 3 is for a character or a small spot such as a stain, and detects the continuous width of the black spot area and determines whether the width is the smallest character If it is smaller than the width, the black spot area detected so far is identified as a black spot and a black spot detection signal is output.

The horizontal positioning circuit 6 determines the character width of the read character from the character data having the character width, generates an output signal, operates the gate control circuit 7, and controls the gate 3 for a certain period of time corresponding to the width _l0 . , and sends the data to the BOX creation circuit 8.

The BOX creation circuit 8 determines the size and center of the character pattern transmitted from the horizontal positioning circuit 6 and the vertical positioning circuit 4, so that it can be easily compared with a standard pattern.

The standard mask circuit 9 stores standard patterns of characters and sequentially sends them to the arithmetic circuit in response to control signals. The arithmetic circuit 10 sequentially compares the character pattern transmitted from the BOX creation circuit 8 with the standard pattern transmitted from the standard mask circuit 9, and outputs the recognition result of the character pattern. At this time, a standard pattern is sent out by a synchronous control signal applied to the standard mask circuit 9, and comparison is performed.

Next, the operating state shown in FIG. 5 will be explained in detail.

For example, character information shown in FIG. 4 is transmitted to the buffer 1 from an optical system (not shown) and is temporarily held. This character information is binarized into "1" and "0" by the preprocessing circuit 2 and transmitted to the vertical positioning circuit 4 and the stain detection circuit 5 via the circuit 3. Then, the vertical positioning circuit 4 detects the column in which the black dot exists, and transmits character information to the horizontal positioning circuit 6. When the character information corresponding to the width _l1 is input, the lateral positioning circuit 6 sends an output signal to the gate control circuit 7. As a result, the gate control circuit 7 sends out a gate off signal to turn off the gate 3. Thus, in FIG. 4, when character information corresponding to _the character width _l1 is input to the lateral positioning circuit 6 for the character C1, the gate 3 is turned off. This off state is maintained for a width l ₀ .

On the other hand, the data of the character C ₁ read out in this way is transmitted from the horizontal positioning circuit 6 and the vertical positioning circuit 4 to the BOX creation circuit 8, becomes a character pattern, and is held in the standard mask circuit 9 in the arithmetic circuit 10. The standard patterns are compared sequentially. Then, the comparison result is output from the result arithmetic circuit 10, and the most similar character is recognized based on the output.

The character information sent from the preprocessing circuit 2 following the character C ₁ is
No data is sent out from the gate 3 during the read data input prohibited area _IH in FIG. Then, when passing through this read data input prohibited area _IH , gate 3
is turned on and the next character information is transmitted to the vertical positioning circuit 4 and the spot detection circuit 5. As a result, as shown in Figure 4, the read data input prohibited area
Since the black spot D ₀ exists when I _H has passed, the vertical positioning circuit 4 immediately detects the black spot and transmits this character information to the horizontal positioning circuit 6 . When the character information corresponding to the width _l1 is input to the horizontal positioning circuit 6, the gate 3 is turned off in the same manner as described above.

At this time, the lateral positioning circuit 6 has a spot D ₀
and the character information of the character _C2 are input, and based on this, a character pattern is created in the BOX creation circuit 8, its center position is determined, and a comparison with the standard pattern is performed in the arithmetic circuit 10. In this case, the portion of the stain D ₀ is the read data input prohibited area I _H
By providing this, only a part of it is added to the character pattern in the BOX creation circuit 8, and the character recognition rate can be improved by comparison with the standard pattern. Moreover, as mentioned above, the number of digits of characters will not be misread.

In addition, if the stain D ₀ is not a large one that is continuous with the letter C ₂ as shown in Figures 3 and 4, but is a small stain that is separated from the letter C ₂ , it will be surrounded by white dots. Therefore, when the black spot detection signal is continuously transmitted for a period shorter than the character width l' of a character with a narrow width, for example, the number 1, this can be determined to be a black spot. Therefore, the black spot detection circuit 5 detects the continuous state of the black spot detection signal, and if it is shorter than the narrow character width l', the previous data is regarded as data of a black spot and is invalidated.

As explained above, according to the present invention, since a read data input prohibited area is provided, it is possible to accurately recognize characters even when there are blemishes, and moreover, excellent character reading recognition is achieved without misreading the number of digits. can be done. In the above explanation, the case where the gate 3 is controlled to be turned off for a certain period of time corresponding to the width l ₀ was explained, but of course the control is not limited to this only; for example, when the width l ₀
It is also possible to determine the off amount of the gate 3 by counting the number of bits corresponding to . The present invention is not limited to the case of horizontally written character recognition as described above, and vertically written characters can also be recognized in the same manner.

[Brief explanation of drawings]

Fig. 1 is an explanatory diagram of MICR characters and masks, Fig. 2 is an explanatory diagram of characters and blemishes, Fig. 3 is an explanatory diagram of problems when blemishes exist, and Fig. 4 is an explanatory diagram of the operation of the present invention. FIG. 5 is a configuration diagram of an embodiment of the present invention. In the figure, 1 is a buffer, 2 is a preprocessing circuit, 3 is a gate, 4 is a vertical positioning circuit, 5 is a spot detection circuit, 6 is a horizontal positioning circuit, 7 is a gate control circuit, 8 is a BOX creation circuit, 9 is a standard mask circuit,
10 each indicates an arithmetic circuit.

Claims (1)

[Scope of Claims] 1. In a character reading recognition device that recognizes a read character by comparing a character pattern obtained from character information obtained by reading a character with an optical reading means with a well-known standard pattern, the character is read by an optical detection means. storage means for storing character information that has been written, means for detecting a predetermined unit width in the character information from the storage means, and prohibiting character information between the predetermined unit width and the beginning of the next predetermined unit width. A character reading recognition device comprising: means;