JPS5841542B2 - optical character reader - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an optical character reading device for reading a field in which a character at a fixed position in a reading character field and other characters have different character sets.

In recent years, in the retail industry such as department stores and supermarkets, improvements have been made to improve customer service by shortening checkout time at checkout counters, to prevent input errors caused by incorrect key-ins, and to automate accounting management, inventory management, and replenishment ordering operations using computers. To measure prices, a device has been developed and put into practical use that optically reads the characters printed on product price tags by manually scanning them and automatically inputs them into a PO8 device.

In this case, the characters printed on the price tag include, for example, the product type, item, business partner, product number, color, size, price, etc., but it is not appropriate to make the price tag large from the viewpoint of human operability and cost. Therefore, these items need to be printed over multiple digits.

When humans manually scan these multi-digit items, there is a good chance that errors will occur such as reading the same line twice, skipping something, or inputting the wrong order, so as a countermeasure to this, we can identify the items on each line. The system uses a method in which function characters are printed at a certain position in the field (for example, on the left edge), and the PO8 device checks the results of reading with an optical character reader and notifies the operator in the event of an incorrect operation. There is.

In the case of this product price tag, for ease of viewing and data processing, different character sets are used, such as numbers for items and letters or special symbols for function characters, and function characters are positioned at the left end of the field. It is one stage that is separated into two.

Figure 1 shows an example of a product price tag. Therefore, when reading the character field on the price tag, in order to correctly scan the characters with a scanner, it is required to be able to scan in either the left or right direction in a direction that is easy for the operator to operate.

In this way, manual scanning type optical character reading devices have been developed and put into practical use that read a character field with function characters printed in a fixed position, but in conventional devices, the characters being read in the field are function characters. Is there?
It was not possible to identify whether the characters represented the item contents.

For example, as shown in Figure 2, function character 1 is placed at the left end of the field, special character 2 is placed at the right end to align the number of digits, and the item content is placed between function character 1 and special character 2. When reading a field containing numbers, if the direction of movement of the scanner is from left to right, it is possible to identify that the first character detected is a function character, and the second and subsequent characters are characters that represent the item contents. However, if the direction of movement of the scanner is from right to left, the first character detected can be identified as the above special character and a number representing the item content, but the function character is only recognized as a function character after the end of the field is detected. It is not possible to identify that the character is a function character at the time the character is recognized.

Therefore, when recognizing a detected character, function characters are stored in separate character cents, such as in the identification dictionary of the character set to which only the function characters belong, and for characters representing item contents, in the identification dictionary of the character set to which they belong. It was not possible to recognize characters by using only the identification dictionary, and it was necessary to use an identification dictionary for the entire character set including both.

Therefore, in conventional manual scanning character reading devices, -
There are problems in that it takes time to recognize characters and that function characters are misread as characters representing items/contents or vice versa, and these problems cause the following problems.

In other words, in a manual scanning type optical character reading device using a surface scanner, character detection, extraction and recognition are generally performed sequentially by a single data processing circuit such as a microcomputer in order to reduce costs. However, in this case, if the next character to be read passes within the field of view of the surface scanner while the character is being recognized, the character will be skipped.

In other words, the upper limit of the movement speed of the scanner that is allowable for human manual operation is approximately (Character printing pitch) (One character recognition time) T Allo 1. s la, 15
C') If the recognition time is long, the upper limit of this speed will be low.

Alternatively, if the upper limit of speed is determined by human operability, it is necessary to reduce the types of characters to be read in order to keep the recognition time below a certain length.

Furthermore, if it is not possible to reduce the types of characters to be read, it is necessary to speed up the recognition circuit in order to shorten the recognition time, which leads to higher costs.

In addition, if a misreading between a function character and a character representing item content is entered into the PO8 device, the incorrect input can be avoided by checking, but a retry will be required.
8 The input efficiency to the device decreases.

Furthermore, in order to reduce misreading between the two, the reading accuracy must be extremely high. For example, in the case of 0CR-B font, S and 5. B and 8. D or 0 and 0 (zero), ■ and 1. T
You must avoid using similar characters such as and 7.

The above is a case of area scanning, but manual scanning type optical character reading devices using line scanners that sequentially perform character detection, extraction, and recognition have the same problems as area scanning devices. Ta.

This invention has been made in view of the above-mentioned points, and it is possible to distinguish between function characters and characters representing other items located at a certain position in a field by identifying only the character types of the character set to which each character belongs. It is an object of the present invention to provide an inexpensive optical character reading device that has a large number of characters, has high-precision reading performance, can be read by high-speed manual scanning, and is inexpensive.

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

For convenience of explanation, as shown in Figure 2, the left end of the field is a function character (alphabetic letter), the right end is a special character (>...) for aligning the number of digits, and between them are numbers representing the item contents. This will be explained by taking the field as an example.

FIG. 3 shows the configuration of a manual scanner, in which a scanner body 11 is provided with a window 12 for inputting a character image printed on a medium 10 to be read such as a price tag, and a light source 1 is provided inside.
3, an optical system 14, a surface scanner 15, an amplifier 16, etc. are provided.

When the light source 13 illuminates the characters on the medium 10 to be read through the window 12, the reflected light enters the optical system 14 through the window 12, and the character image formed by this incident light is formed on the optical system 14. The image is formed on the surface scanner 15 by the following.

Further, the surface scanner 15 converts the incident optical signal into an electrical signal and outputs it, and the signal corresponding to the character image obtained from the surface scanner 15 is sent to the amplifier 16 and amplified.
The signal is sent to a quantization circuit (not shown) via a signal cable 17 and converted into a digital signal.

FIG. 4a shows the planar configuration of the above-mentioned surface scanner 15, in which each mesh of vertical NR bits and horizontal NL bits has photoelectric conversion characteristics and sequentially outputs one bit at a time in the horizontal or vertical direction. This is an element that is capable of

NL and NR are used at appropriate values, or with such masks, such that the character will fit sufficiently into the area scanner, but adjacent characters in the field or characters in adjacent lines will not fit sufficiently.

In FIG. 4a, 21 is a vertical projection of a black block within the surface scanner, and 22 is a horizontal projection.

A method of creating these projections will be explained using a surface scanner that outputs serially from the upper right mesh to the lower left mesh as shown in FIG. 4, for example.

In FIG. 4, 23 is an NL-bit shift register, 24 and 25 are respective OR circuits, 26 is a flip-flop, and 27 is an NR-bit shift register. The logical sum signal 24s'Pl of the output signal 23S of the shift register 23
Shift register 23 cleared at the start of frame scanning
By inputting a 1-bit shift IL every time one mesh is output during one frame period, a vertical projection is created in the shift register 23 at the end of one frame scan.

Further, the flip-flop 26 has one row (N in the horizontal direction).
It detects a black block in a mesh string of L bits), and is reset at the start of scanning one row, and every time one mesh is output, the logical sum of 15s of the output of the surface scanner 15 and the output of this flip-flop 26 itself is 25s. If a black block is present at the end of one line scanning, a set state is entered.

If the output of the flip-flop 26 is shifted into the NR bit shift register 27 every time one row of scanning is completed, a horizontal projection is created in the shift register 27 at the end of one frame scanning.

Therefore, by using the vertical projection, it is possible to detect the moving direction of the surface scanner main body 11.

That is, first, a blank on the left or right side of a character field is detected by the presence of only white blocks or only black blocks that are considered noise in the vertical direction, and then it is determined whether the left or right side of the vertical projection is black. By detecting, for example, if the right side is black as shown in Figure 5a, it can be determined that the scanner main body 11 is moving from left to right, and if the left side is black as shown in Figure 5, it can be determined that the scanner main body 11 is moving from left to right. Scanner body 11
It can be determined that the character is moving from right to left.Next, wait until the character enters the field of view of the surface scanner 15, and as shown in FIG. If a black block corresponding to the width and height of the character fits inside the character, it is determined that the character is a character.

In this case, the scanning pattern by the surface scanner 15 is created in parallel with the creation of the above-described vertical projection and horizontal projection.
As shown in the figure, storage is controlled in a shift register.

In FIG. 6, 31 is a shift register into which the output signal of the surface scanner 15 is input, and 32 is a preprocessing process that performs, for example, noise elimination, line width normalization, character size normalization, etc., and converts it into a form that is easy to recognize. The circuit 33 is a shift register that stores the output signal from the preprocessing circuit 32.

The output of the preprocessing circuit 32 is supplied to one input terminal of an AND circuit 34, and the other input terminal of the AND circuit 34 is supplied with an input switching signal 36 via an input 35.

The output signal of this AND circuit 34 is then supplied to the shift register 33 via an OR circuit 37.

The output of this shift register 33 is sent to a character recognition circuit 39 and also to one input terminal of an AND circuit 38.

The AND circuit 38 receives the input switching signal 3.
6 is supplied, and the AND output thereof is supplied to the shift register 33 via the OR circuit 37.

In this way, the output signal 15s from the surface scanner 15 is shifted into the shift register 31, and when a character is detected in the surface scanner 15 by both injections, the shift clock of the shift register 31 is stopped, and the shift register 31 is shifted into the shift register 31. register 3
1 can be stored.

This detected and cut out scanning pattern is sent to the preprocessing circuit 32, where it is converted into an easily recognizable form through noise elimination, line width normalization, character size normalization, etc., and then input into the shift register 33 and stored. .

The output of the shift register 33 is sent to the character recognition circuit 39 and recognized. When sending the output to the character recognition circuit 39, the input switching signal 36 is set to 1" and the output of the shift register 33 is sent to the AND circuit 38 and the OR circuit 31. The pattern can be left in the shift register 33 by passing it back as its own input via the Shift Register 33.

As a result, the previous character pattern remains in the shift register 33 until the next character is detected and extracted.

In this way, each character is detected and recognized, but even after a certain period of time (this period of time is determined by the upper limit of the allowable range of the scanner's movement speed and the distance from the adjacent field), characters are still detected. The end of field scanning is determined by detecting whether the field is blank or by detecting a blank at the edge of the field.

At this time, the character pattern detected and cut out last remains in the shift register 33.

In addition, various methods are known for character recognition, such as feature extraction, pattern matching, and strobe analysis. Therefore, if the character set to which the character belongs is limited, it is advantageous to recognize only that character set as a candidate category in order to reduce recognition time and reduce misreading.

For example, the pattern matching method is a method that calculates the degree of similarity (degree of coincidence) between an input cover pattern and a reference pattern of a candidate character, and selects the candidate character with the highest degree of similarity as the recognition result.

Figure 1 is for explaining how to recognize an input character pattern using the dictionary of the character set to which it belongs.
0 is an address counter, and 41 is a storage device in which the reference pattern (dictionary) of the function character set is stored (
A storage device 42 (hereinafter referred to as a first storage device 41) is a storage device (hereinafter referred to as a second storage device 42) in which a reference pattern (dictionary) of a character set representing item contents is stored. The output of the AND circuit 43 is supplied to one input terminal, and the output of the second storage device 42 is supplied to one input terminal of the AND circuit 44.

The dictionary switching signal 45 is supplied to the other input terminal of the AND circuit 43, and the dictionary switching signal 45 is supplied to the other input terminal of the AND circuit 44 via an inverter 46.

The outputs of these AND circuits 43 and 44 are supplied to both input terminals of an OR circuit 47, and the output of this OR circuit 47 is supplied to one input terminal of a similarity calculation circuit 48.

A character pattern signal to be read is supplied to the other input terminal of this similarity calculation circuit 48, and the similarity calculation output is a similarity comparison circuit 49 that selects the candidate character having the maximum similarity.
A storage device 50 (hereinafter referred to as the third
(referred to as a storage device 50), and are sequentially written to addresses designated by an address counter 51.

That is, the standard pattern of the function character set (
A dictionary) is stored in the first storage device 41, a reference pattern (dictionary) of a character set representing the item content is stored in the second storage device 42, and when the character to be read is a function character, the dictionary is switched. The signal 45 is set to 1'', and when the character represents the item content, the dictionary switching signal 45 is
is set to "O", a dictionary is selected, and the selected dictionary signal and character pattern signal to be read are input to the similarity calculation circuit 48 to calculate the similarity.

Then, the result is input to the similarity comparison circuit 49, and the candidate character having the maximum similarity is set as the recognition result, and this recognition result is sequentially written to the address specified by the address counter 51 of the writable and readable storage device 50. It gets crowded.

The overall control from the start of reading to the end of reading described above is shown in the flowchart of FIG.

In other words, the scanning direction is detected, the first character is detected and cut out, and if the scanning direction is from left to right in Figure 2, it is recognized by the function character cent dictionary, and if it is from right to left, it is a character set representing the item. It is recognized by the dictionary.

After detecting and cutting out the second and subsequent characters, the end of the field is detected, and when the end is detected, if the scanning direction is from left to right, the reading is completed by outputting the answers in the order of reading.

If the scanning direction is from right to left, the last character pattern detected and cut out and stored in the storage device is recognized using the dictionary of the function character set, and then the last recognized result is discarded in step A and read. Outputs the answer in reverse order and finishes reading.

In this way, the series of reading is completed. In addition, although the case where a surface scanner was used was explained in the said Example, it is applicable not only when the surface scanner is used but also when the line scanner is used.

For example, as shown in FIG. 9, a roller 61 is installed in a position that does not obstruct the field of view of the line scanner 60, and the rotational force of this roller 61 is transmitted to the rotor of a generator 63 by a gear 62.
The moving direction of the scanner main body 11 can be detected based on the positive and negative voltages generated by the sensor 3.

Further, character detection can be carried out by storing the output signal of the line scanner in a storage device every time the gear 62 rotates by a certain angle, and at the same time creating vertical and horizontal projections in the same manner as in the previous embodiment.

As a method of storing the output signal of the line scanner in a storage device every time the gear 62 rotates by a certain angle, for example,
There is a method in which a hole is drilled with a constant pinch on the same circumference of 2, the passage through the hole is detected by a lamp and a solar cell (not shown), and the data is stored in a storage device at this timing.

Furthermore, in the above embodiment, an example was explained in which only the characters at the left end of the field are included in a character set different from the other characters, but if the two characters at the left end are function characters,
A similar implementation can be achieved by preparing a storage device for two characters of detected and extracted character patterns.

The same applies to the case of three or more characters.

Also, if the characters are at both ends of the field, the same can be achieved by simply identifying the first character using the function character set dictionary.

Further, in the case of a feature extraction method, the character pattern to be stored may be an extracted feature pattern.

As described above, according to the present invention, there is a function to detect the scanning direction, a function to store the detected character until the next character is detected, and a function to select the type of character to be identified (candidate category) according to the input character. It has a function to sequentially memorize the recognition results, and if it is unknown which character set the character belongs to at the time of recognizing the character, it is temporarily stored in the dictionary of the character set to which it is most likely to belong. At the same time as recognition, the detected cutout and preprocessed patterns are stored until the end of field scanning, and for characters recognized in a dictionary with the wrong character set, the pattern is memorized after field scanning is completed. By re-recognizing each character using a correct character cent dictionary, it becomes possible to recognize each character by referring only to the dictionary of the character set to which it belongs, and from these points, (1) the recognition time is shortened.

(2) The upper limit of the movement speed of the scanner is increased.

(3) Misidentification between different character sets can be prevented.

(4) Each character set can have many types.

(5) It is possible to reduce the speed of the recognition circuit, that is, to simplify the circuit and reduce the cost.

Various effects can be obtained.

[Brief explanation of the drawing]

The drawings are for explaining one embodiment of the present invention. Fig. 1 shows an example of a product price tag, Fig. 2 shows an example of a field, and Fig. 3 shows an example of a product price tag. FIGS. 4a and 4b are cross-sectional views showing the structure of a scanning character reading device; FIG. b is an explanatory diagram of the method for detecting the moving direction of the scanner, Fig. 6 is a control block diagram for performing detection extraction, preprocessing, and storage, and Fig. 7 is an explanatory diagram of the input character pattern in the dictionary of the character set to which it belongs. FIG. 8 is a flow chart for explaining reading control, and FIG. 9 is a sectional view showing the configuration of a manual scanning type character reading device using a line scanner. 11... Scanner body, 15... Surface scanner, 21...
・Vertical projection of the black block, 22... Horizontal projection of the black block, 41... Storage device (first storage device) in which the reference pattern of the function character is stored;
42...Storage device (second storage device) in which a reference pattern representing item content is stored, 48...Similarity calculation circuit, 49...Similarity comparison circuit, 50...Reading result A storage device for storing (third storage device).

Claims (1)

[Claims] 1. A scanner that scans a field and detects characters within the field, means for detecting the scanning direction of the scanner, means for detecting and cutting out character patterns obtained from the scanner character by character, and means for storing at least one character of the character pattern detected and cropped by the means until the next character pattern is detected and cropped; and means for selecting a candidate category based on the detected scanning direction and using the stored character pattern in the selected candidate category. a means of recognizing
means for storing the recognition result of the means until the field scanning by the scanner is completed, and a candidate category for at least one character in a specific digit according to the detected scanning direction when the field scanning by the scanner is completed; What is claimed is: 1. An optical character reading device comprising means for switching and re-recognizing the recognition results and replacing the stored contents of the corresponding recognition results with the re-recognized recognition results.