JPH0792812B2 - Vouchers and how to read them - Google Patents

Description

TECHNICAL FIELD The present invention relates to a slip used for giving and receiving money, and a method for reading the slip.

[Prior Art] Conventionally, a slip of handwritten characters, statistical data, and the like are read by an automatic reading device or an optical character reading device and input into an information processing device such as a computer. However, the typeface of handwritten characters varies greatly due to variations in personal characteristics, and even in the popular postal service system, there are still many handwritten numbers that cannot be immediately read. It is currently being processed in.

Conventional automatic reading devices and numeral recognition devices are still incomplete, and they are not fully automated, and it is expected that highly accurate reading devices will appear.

[Problems of the prior art] The ability of the conventional numeral recognition device is still inadequate and has not yet been fully automated. In particular, the slips used for the transfer of money in banks are 100 It still relies entirely on operator keyboard input as it requires% accuracy.

In today's widespread use of high-density online information processing systems, it is expensive to operate a general operator by sequentially supplementing the unrecognizable characters and the numbers on the slips with a keyboard. However, there was a problem that such a system could not fully show labor saving and immediacy.

[Object of the Invention] The present invention has been made in order to solve the above problems, and does not require strict accuracy and preparation conditions at the entry stage.
A voucher that anyone can easily and accurately fill in and a voucher reading method that can read the information on this voucher with high precision and accuracy and can automatically check erroneous entries and omissions and that does not require correction by an operator. It is intended to be provided.

[Means for Solving the Problems] In the present invention, the short and long line segments of the preceding and succeeding lines are arranged in a straight line to correspond to the control mark indicating the type of format, the numeral entry area and the numeral entry area. The voucher is displayed with an information mark including a plurality of pattern writing areas arranged. Also, enter the numbers and patterns on this voucher, input the voucher into an automatic reader, generate image data of control marks and information marks, calculate the coordinates of these image data, and then calculate the image of the control marks. From the data, the size of the line segments is compared in order of magnitude to calculate a binary number, a binary code is determined, a predetermined format corresponding to this binary code is read, and the standard of the information mark specified by this format is read. The coordinates are read, the calculated coordinates for the information mark and the reference coordinates are compared, and then the image data of the information mark is separated into the entered number and the image data of the entered pattern, and the image data of the entered number is set to a predetermined reference. Directly recognize the entered number by comparing it with the number pattern, and compare the image data of the entered pattern with the predetermined reference rod-shaped pattern to recognize the corresponding number of pairs. It was filled numbers on slips as the final recognition by comparing with those directly recognized numbers and corresponding recognized digits.

[Operation] Read a predetermined format from the image data of the control mark generated from the slip, compare the position of the information mark specified on the format with the position of the information mark generated from the slip, and enter if they match. It accepts patterns and numbers to be entered, recognizes the numbers directly, recognizes numbers correspondingly from the entry bar pattern, and further recognizes these numbers by comprehensively judging the direct recognition and the corresponding recognition. , The handwritten number on the slip can be read with high accuracy by the automatic reading device.

[Embodiment] An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a format diagram of a slip of the present invention. In this embodiment, a bank deposit slip will be described as an example.

In the figure, 1 is a paper such as a form of arbitrary size, a classification mark 20 as a control mark is displayed by printing or filling in the upper corner of the paper 1, and a predetermined distance is provided below the classification mark 20. Information marks 21, 22 and 23 are displayed.

The classification mark 20 identifies the type of slip, and a "payment slip" is displayed here so that the user can see at a corresponding position. The information mark 21 is the amount column, the information mark 22 is the account number column, the information mark 23 is the date column, and each information mark
At the positions corresponding to 21, 22, and 23, "payment amount", "account number", and "date" are displayed so that the user can see at a glance.

Thus, the classification mark 20 is formed by arranging two or more, for example, five line segments a, b, ... E having different lengths in succession in the horizontal direction. They are successively compared in the order of rows.

This classification mark 20 is the line segment a and 0 assumed before that.
Or the subsequent line segment b is compared to determine the size, and then the line segment b and the line segment c are compared, and similarly, the line segment d and the line segment e are compared, and a series of line segment lengths are compared. The size of is determined.

The classification mark 20 is a control mark for displaying the lengths of the five line segments a, b, ... E by changing the lengths of the preceding and succeeding lines in order, and specifying a different kind of slip "remittance slip".

Next, the information marks 21, 22, and 23 are arranged in a row in the horizontal direction, and these are provided with rectangular frameworks 31, 32, and 33, and the upper left two sides of each rectangle are displayed in bold. There is.

Under a certain classification mark 20, the coordinate value of the writing position of each information mark 21 to 23, the point (x1, y1) where the upper left two sides touch, based on the writing position (0, 0) of the classification mark 20, (X2, y
2), (x3, y3) are specified in advance.

The general shape of the information marks 21 to 23 are shown in FIGS.
This will be described in detail with reference to the drawings.

A plurality of rectangular masses are arranged in a matrix in each of the frames 31, 32, 33, and rectangular long masses 3 are respectively provided below the one side of the mass 2 in the first row so as to be separated by a predetermined distance. There is.

Each long mass 3 is divided into 6 equal parts in the length direction, and the first mass 4 and the second mass 5 to the sixth mass 9 are arranged from the side closer to the mass 2. The boundary between the first cell 4 and the second cell 5 is a thick line or 2
Separated by a heavy line.

In FIG. 3, the numbers “0 to 9” are entered in the cell 2, and the rod-shaped patterns 10 uniquely determined corresponding to these numbers are entered in the cells 1 to 4 of the long cell 3. It shows what was done.

That is, by writing one bar-shaped pattern 10 through the first to sixth cells 9 in the length direction of the long cell 3, the numeral "0" is written, and the bar-shaped pattern 10 is extended only to the first cell 4. Enter the number "5" by writing in the direction, and enter the number "1" by entering the rod-like pattern 10 only in the second cell 5 and the second cell 5 and the third cell 6.
Enter the bar pattern 10 in a straight line on the
The rod-shaped pattern 10 is linearly written on the mass 5, the third mass 6, and the fourth mass 7, and the numeral "3" is written.
A rod-shaped pattern on the third mass 6, the fourth mass 7, and the fifth mass 8.
The number "4" is displayed by entering 10 in a straight line.

The bar-shaped pattern 10 is linearly written on the first mass 4 and the second mass 5, and the numeral "6" is written, and the bar-shaped pattern 10 is linearly written on the first mass 4, the second mass 5 and the third mass 6. Number "7"
The rod-shaped pattern 10 is linearly written on the first cell 4, the second cell 5, the third cell 6, and the fourth cell 7, and the numeral “8” is given to the first cell 4, the second cell 5, and the third cell 6. 4th cell 7 and 5th cell 8
The bar-shaped pattern 10 is linearly written on the screen to display the number "9".

The cells 2 and the long cells 3 are a standard for entering numbers and the rod-shaped pattern 10, and may be displayed in cells of a color that cannot be read by the optical reading device described later.

The number from 0 to 9 is arbitrarily entered and printed in the square 2. The long square 3 is similar to the five-ball abacus, and the first square 4 expresses only the value "5". 2 squares 4 to 6
The cell 9 represents only the value of each "1". Like the abacus ball, the entry of the second square 4 expresses "1", but the entry of the third square 5 cannot express "1". However, "0" is the first
The longest rod-shaped pattern 10 that is continuous in squares 4 to 9
Fill in and express.

A pair of squares 2 and a long square 3 represent a single digit number, and in order to represent a two digit number, they are arranged a little closer to each other in the lateral direction.
The paired squares 2 and the long squares 3 are used, and the squares 2 and the long squares 3 of each pair are added closer to each other in the lateral direction as the number of digits increases.

The writer writes one of the numbers 0 to 9 in the cell 2 as necessary, and writes the above-mentioned rod-shaped pattern 10 in the long cell 3 corresponding to the entered number.

In addition, the rod-shaped pattern is drawn from the upper end to the lower end of each of the first to fourth cells 9 to 9 with a normal writing tool.

Figure 4 shows an optical reader (OCR) into which this slip is entered.
Shows the entire configuration of, 12 is a scanning device, 13 is a central processing unit (CPU), 14 is a baer memory, 15 is a random access memory (RAM), 16 and 17 are read only memory (ROM),
18 is a file device.

The scanning device 12 includes a classification mark 20 and an information mark 21 on the paper 1.
The entry data in 22 and 23 is photoelectrically converted, and the generated image data is given to the buffer memory 14. This image data is stored in RAM15 as a work memory by CUP13.
The arithmetic processing is performed together with the data from the ROMs 16 and 17 and the file device 18, and the intermediate result and the final result are stored in the RAM 15.

The final result of this calculation is sent as a numeric code to the host computer and other information processing devices via the CUP 13.

In ROM16, the number 0 that was uniquely arranged as described above
9 to 9 and the reference rod-shaped pattern are stored in advance.

Further, in the ROM 17, the reference numeral patterns 0 to 9 are 5 ×, for example.
It is stored in advance as image data on the 7 dot matrix.

The file device 18 is composed of a large-capacity memory such as a magnetic disk, and stores in advance the types of slips designated by the classification mark 20, such as deposit slips, refund slips, remittance slips, and tax payment slips.

Next, the operation of reading the slip in which the predetermined data is written by the optical reading device (OCR) will be described with reference to FIGS.

When depositing money into his account at the bank, the customer selects the deposit slip of the present invention, and fills in his account number, the date, and the amount in the deposit slip in the above-described manner.

The completed deposit slip paper 1 is scanned by the scanning device 12, the images of the classification mark 20, the information marks 21 to 23, etc. are photoelectrically converted and stored in the buffer memory 14 as image data.

CPU 13 separates classification mark 20 from buffer memory 14
Move to AM15 and decrypt it. That is, the CPU 13 first measures the length of the first line segment a, then measures the lengths of the line segments b to e, and stores them in the RAM 15.

The CPU 13 regards these measured values a to e as continuous data as shown in FIG. 5, and the respective line segments a to e are successively adjacent to each other as a and b, b and c ,. Relative to. Here, when the length of the preceding line segment is shorter than the length of the following line segment, a <b is defined as “1”, and when it is longer, b> c is defined as “0”.

The CPU 13 decodes the binary number 1011 from the classification mark 20 by comparing the sizes of the line segments a to e. And CPU13 is RAM
A storage area having a binary number 1011 as an item is provided in 15 as shown in FIG.

Next, from the image data in the buffer memory 14, the CPU1
3, the starting position of the classification mark 20 is the origin (0,0), the starting position of the amount column of the information mark 21 (x1, y1), the starting position of the account number column of the information mark 22 (x2, y2), the information mark The coordinate value of the starting position (x3, y3) of the date column of 23 is read and stored in the storage area having the binary number 1011 in the RAM 15 as an item as shown in FIG.

Thus, the CPU 13 searches the inside of the file device 18 using the binary number 1011 of the classification mark 20 as a guide and fetches the data table of FIG. 6 summarized under the code 1011.

The data table is now in the form of a deposit slip, and the entry start coordinates (X1, Y1), (X2, Y) that should be the amount column of the information mark 21, the account number column of the information mark 22, and the date column of the information mark 23
2), the reference value of (X3, Y3) etc. is read.

On the other hand, the binary numbers 1011 and x1 to y3 in the RAM 15 in FIG. 7 indicate the coordinate values actually read from the sheet 1 of the slip and the extracted binary numbers. For example, (x1, y1) relates to the information mark 21. It is the read coordinate value of the entry start point.

Next, the CPU 13 reads the coordinate values (x1, y1), (x2, y2), (x3, y3) of the information marks 21 to 23, which are predetermined values (X1, Y1), (X2, Y2). , (X3, Y3) is determined.

x1-X1 = L1 etc. L2, L3 and y1-y1 = m1 etc. m2, m3,
0 If L1, L2, L3m1, m2, m3 <k, the input paper (k is a predetermined small value) is recognized as a deposit slip, and if not, the operator or customer is instructed to re-fill. .

Next, the CPU 13 confirms the deposit slip and then the information mark 2
Move on to the recognition of the entry data in 1, 22, and 23.

The CPU 13 separates the numerical image data in the square 2 and the bar-shaped pattern image data in the long square 3 for the first digit, and the numerical image data is enlarged or reduced to 5 × 7.
Is standardized to the matrix size of.

ROM17 of standardized numerical image data and recognition dictionary
The reference candidate number patterns from are compared with each other, and the recognition candidate number and the similarity (%) are calculated in order from the number with the highest similarity, and are temporarily stored in the RAM 15.

For example, if the handwritten number in square 2 is "1", the first
Candidate number is "1", similarity is 70%, second candidate number is "7"
Then, 30% of the similarity is temporarily stored in the RAM 15.

Next, the corresponding rod-shaped pattern image data is compared with the reference rod-shaped pattern from the ROM 16 of the pattern dictionary. The recognition candidate numbers and the similarities (%) are calculated in order from the corresponding numbers of the patterns having the high similarities, and are temporarily stored in the RAM 15. now,
Since the rod-shaped pattern 10 in the long square 3 is only the shortest pattern in the second square 4, the corresponding reference rod-shaped pattern and the number "1" are recognized with 100% similarity, and the first candidate number is "1". The similarity “100%” is temporarily stored in the RAM 15.

Finally, from the RAM 15, first, the candidate number “1” for the cell 2 and the similarity “70%”, and the candidate number “1” for the cell 3 and the similarity “100” are obtained.
% "Is read out and compared. Moreover, since there is no candidate number next to the candidate number" 7 "next to the cell 2 with a similarity of 30% and long cell 3, a" NULL "signal is read out and compared. .

The CPU 13 recognizes and outputs the number that matches the candidate numbers and gives the maximum value of the sum of the similarities as the data number written in the information mark 21. That is, the number "1" is output to the computer side as a code signal (0001) from 100% of "1" of the first candidate.

In the same way, the amount of information mark 21 and information mark 22
The data numbers entered in the account number column and the date column of the information mark 23 are recognized and output to the computer side.

Therefore, the type of the slip is recognized from the input classification mark 20 of the slip paper, and then the entered data numbers in the information marks 21 to 23 are recognized, the necessary data on the slip are read and sent to the computer side.

Since the handwritten numbers in the cell 2 and the handwritten stick-shaped pattern 10 in the long cell 3 are individually recognized, and these recognitions are further collated for final recognition, complete automatic recognition is possible.

Further, in the case of only direct numeral recognition, the handwritten numerals 5 and 6, 1 and 7 are often similar, but since the sticky pattern is also recognized, the corresponding sticky patterns 5 and 6 are clearly distinguished in the ROM. It is stored and stored, and can be easily distinguished only by identifying the length of the rod-shaped pattern. Similarly, the rod-shaped patterns corresponding to 7 and 1 have completely different lengths, and erroneous recognition hardly occurs.

In this way, since the logical judgment is made in triplicate, it is possible to recognize the digits with high accuracy.

Further, when determining the type of the classification mark, since the line lengths are converted into binary numbers by relative comparison, the separability for noise such as dirt and paper wrinkles is improved.

In the above embodiment, the line segment a and the line segment b of the classification mark are compared to determine the size, and then the line segment b and the line segment c are compared. It is also possible to compare 0 with the previously assumed 0, and because 0 <a, the 1st bit is always set to 1, and this 1 may be used for error checking.

By increasing the number of line segments a, b ...
The number of evolutionary bits is increased and the number of classification targets is 2 ⁴ = 16, 2 ⁵ =
32,… 2 ⁸ = 256, which can be greatly increased.

[Effects of the Invention] As described above, according to the present invention, the short and long line segments of the preceding and following lines are arranged in a straight line to correspond to the control mark indicating the type of format, the numeral entry area, and this numeral entry area. Numbers and patterns are written on a voucher that displays information marks consisting of a plurality of pattern writing areas arranged in parallel, and the voucher is input to an automatic reading device to generate image data of control marks and information marks. Then, the coordinates of these image data are calculated, and then the control mark is compared with the image data in the order of arrangement of line segments to calculate a binary number, a binary code is determined, and a binary code is determined. Read a predetermined format, read the reference coordinates of the information mark specified by this format, compare the calculated coordinates for the information mark with the reference coordinates, and then read the image data of the information mark. The input numeral is separated into image data of the input numeral and the image data of the input numeral is compared with a predetermined reference numeral pattern to directly recognize the input numeral, and the image data of the input pattern is converted into a predetermined reference rod-shaped pattern. By comparing and recognizing the numbers that form a pair and comparing these directly recognized numbers with the correspondingly recognized numbers, the entered numbers on the slip are finally recognized, so the information on the slip is highly accurate. Can be read accurately and completely automatically, and mistakes and omissions can be easily checked, and a slip reading method that does not require any intervention or correction by an operator can be realized.

Therefore, the accuracy of digit recognition is dramatically improved, and the calculation of statistical data and the like using various slips is speeded up. In that case, it is possible to improve the immediacy and unmanned, and to reduce the cost of banking business.

[Brief description of drawings]

FIG. 1 is an overall view of the slip sheet of the present invention, FIG. 2 is an enlarged view of the information mark of the present invention, FIG. 3 is a view of the information mark of FIG. 2 in which numbers and rod-shaped patterns are entered, and FIG. FIG. 5 is a configuration diagram of an optical reading device, FIG. 5 is a diagram showing a method of decoding a control mark,
Figure 6 shows the memory configuration of the file device, and Figure 7 shows the RAM.
It is a diagram showing a recording state of. 1 ... Paper, 2 ... Mass, 3 ... Long mass, 4-9 ... 1st to 6th mass, 10 ... Rod pattern, 12 ... Scanning device, 13 ... CPU, 14 ... Buffer Memory, 15 …… RAM, 16 ・ 17 …… ROM, 18 …… File device, 20 …… Classification mark, 21-23 …… Information mark.

Claims (1)

[Claims] 1. A control mark indicating a type of format by arranging short and long line segments of the preceding and succeeding lines in a straight line, and a numeral entry area and a plurality of pattern entry areas arranged corresponding to the numeral entry area. Numbers and a pattern corresponding to the numbers are written on a voucher displaying the following information mark, and the pattern writing area is formed from long squares that are equally divided into six in the length direction, and each short square that is equally divided into six is obtained. From the side closer to the number entry area, the first square, the second square to the sixth square,
Use the sum of these numbers to represent the number 5 by writing the rod-shaped pattern on the first cell independently, and to extend and retract the rod-shaped pattern on the second to sixth cells to represent the numbers 1 to 5. The patterns corresponding to the numbers are written so as to represent the numbers 1 to 10, image data of the control mark and the information mark are generated, coordinates of these image data are calculated, and then the image of the control mark is generated. From the data, the size of the line segment is compared in order to calculate the binary number, and the binary code is determined.
A predetermined format corresponding to this binary code is read, the reference coordinates of the information mark specified by this format are read, the difference between the calculated coordinates and the reference coordinates for the information mark is calculated, and this difference is calculated. The slip reading is continued only when it reaches the preset allowable range, and then the image data of the information mark is separated into the input numeral and the image data of the input pattern, and the image data of the input numeral is predetermined. Directly recognize the entered number by comparing it with the reference numeral pattern, and compare the image data of the pattern written in the six equally divided short squares with a predetermined reference rod-like pattern and recognize the representative number correspondingly. Then, in the method of reading a slip in which these directly recognized numbers are compared with correspondingly recognized numbers to recognize the entered numbers on the slip, Over data and numbers 0, 1, 2, 3,
6, 7, 8, 9 are compared with the reference numeral patterns, and the similarity of each reference numeral pattern to the image data is checked, the degree of similarity is calculated, and the pattern written in the short squares is calculated. Image data and the reference rod-shaped patterns predetermined corresponding to the numbers 0, 1, 2, 3, 4, 5, 6, 7, 8, and 9, respectively, are compared, and each of the reference rod-shaped patterns has the short mass. The similarity between each of the patterns entered in the image data is checked, the similarity is calculated, and the similarity for each of the reference numeral patterns and the similarity for each of the reference rod-shaped patterns corresponding to the respective numerals are respectively indicated by respective numerals. 0,
The sums of 1, 2, 3, 4, 5, 6, 7, 8, and 9 are added to obtain the respective sums, and of these sums, the numbers corresponding to both reference patterns having the maximum sum of the similarities are given to the slips. How to read the slip that is finally recognized as the number entered above.