JPS63223889A - Optical character reader - Google Patents

Abstract

PURPOSE:To correctly read even a price tag which has variance in print quality among characters on a line by selecting recognition results based upon respective threshold values in character units. CONSTITUTION:An image picked up by an image sensor 6 is sent to a binarization circuit with a push-button switch 19 and images which are converted into binary data by using mutually different thresholds are stored in image memories 8-1-8-N. The binarization images based upon the respective threshold levels which are stored in the memories 8-1-8-N are processed in order by a respective-character recognition processing part, whose results are stored in recognition result buffers 17-1-17-N in order. When character recognition results and character information corresponding to the respective thresholds are stored in the buffers 17-1-17-N, a digit matching processing part 18 obtains a group of character recognition results corresponding to the same character and a recognition result selection part 21 selects and outputs a final recognition result in the group.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to an optical reading device, particularly a hand-held device that reads characters on a price tag or the like by holding the scanner and touching the line where the characters are written. Regarding formula OCR (optical character reader).

[Conventional technology]

POS (Point Of) collects sales information for each item and manages inventory at supermarkets and department stores.
5ales) system is widespread. As an OCR that can be used in this POS system, there is a hand-held optical character reader that can read characters within its field of view simply by placing a scanner on a sheet of paper. No., special application 1986-
No. 255777).

FIG. 2 shows the configuration of a hand-held optical character reading device.

Reference numeral 1 denotes a scanner, which reads characters, symbols, etc. written on the paper 3 by placing the hand 2 on the paper 3 and pressing a push button switch 19. The paper 3 is, for example, a price tag on which information is written in a POS system. 4 is a light source, 5
6 is a lens system, 6 is an image sensor, and 7 is a drive circuit for the image sensor. The optical character reading device shown in Figure 2 converts the analog signal output from the image sensor into binary values of white and black in order to handle characters of various printing qualities (thick characters, thin characters, dark characters, and light characters). Threshold value when converting to value (
There are two types available. 14 and 14' are binarization circuits with different thresholds, and the binarized results are stored in the image memories 8 and 8' corresponding to the respective binarization circuits.
．． 8' stores binarized data of almost the entire field of view of the image sensor 6. FIG. 3(a) shows an explanation of binarized data by the binarization circuit 14 of the image sensor 6.

It is an image sensor with horizontal (X) Xll (Y) pixels of pxq pixels, and captures characters within its field of view.

Reference numeral 15 denotes a circuit for selecting which threshold image is to be processed in each of the character recognition parts 9 to 13, and sends the data in the image memory 8.8' to the one-digit extraction circuit 9 in response to instructions from the threshold switching determination circuit 16. When the push button switch 19 is pressed, the image memory 8 is first selected, and the image binarized by the binarization circuit 14 is recognized.

Characters and symbols are identified by the character identification circuit 13, but since the character identification circuit 13 identifies each character, it is necessary to retrieve data for one character from the image memory 8. - The digit extraction circuit 9 takes out a ta corresponding to mXq@prime equivalent, which is the processing capacity of the single character extraction circuit 1), from the screen memory 8 and stores it in the single digit memory 10. The character extraction circuit 1) takes out data equivalent to mXn pixels, which is the processing capacity of the character identification circuit 13, from the single-digit memory and stores it in the character memory 12.

In FIG. 3 (al), first the negative digit extraction circuit 9 is
The data from 1 to X=m and from Y=1 to y=q are taken out from the screen memory 8 and transferred to the negative digit memory 10 (FIG. 3 (bl)). - The digit extraction circuit 9 looks at the contents of the one-digit memory 10 and selects Y from the range including the character image (Y-1 in this example).
n lines of =1)+n-1) are transferred to the one-character memory 12 (FIG. 3(c)). - When a character is stored in the character memory 12, the character recognition circuit 13 recognizes the character. Next, the data from X=2 to X=m+L Y=1 to Y-9 is taken out from the screen memory 8 and transferred to the negative digit memory 10 (Fig. 3 (b)). It is transferred to the one-character memory 12.Then, in the same way, the position to be taken out from the screen memory 8 is shifted in order, and the one-character memory 1
0, the image including the character image is transferred to the one-character memory 12, and the character recognition circuit 13 processes it to recognize one line.

FIG. 4 shows how to determine the range to be transferred from the one-digit memory 10 to the one-character memory 12. First, horizontal OR is calculated for each row of the one-digit memory 10. Horizontal OR is to focus on one row in the horizontal direction, and if there is a black pixel in that row, it is set as 1, and if there is no black pixel, it is set as O.
This is an operation where

Now, if the black output of the sensor is expressed as 1 and the white output as O, then the result of horizontal OR is nothing but the result of taking the logical sum of each pixel in one row.

Therefore, this operation is called horizontal OR. Then, as shown in FIG. 4(b), in a part where there is a character, the result of horizontal OR is black for that range. For example, if the horizontal OR becomes black from Y=13, the range to be transferred from the -digit memory to the single character memory is n pixels from Y=1), including the white above the character.

By the above processing, characters and symbols included in the field of view of the image memory 8 can be read.

The read result is sent to the threshold switching circuit 16, and if the read result is good (if there are no problems such as rejects), it is output as is. ), the threshold switching circuit 16 outputs a threshold switching signal to the threshold selection circuit 15 without outputting the read result, and the threshold selection circuit 15 outputs a threshold switching signal to the threshold selection circuit 15.
By sending the binarized image in step 4' from the image memory 8' to a partial cutout circuit 9, the binarized image is recognized in the binarization circuit 14'.

As described above, by adaptively selecting binarized images with different thresholds depending on the reading results, price tags with a wide range of print quality can be read.

[Problem that the invention seeks to solve]

In conventional optical character reading devices, thresholds are switched line by line, so the recognition results of the first threshold are used for some characters in a line, and the recognition results of the second threshold are used for other characters. It is not possible to adopt the recognition result of the threshold. Therefore, depending on the printing quality of the price tag, the correct recognition result may not be obtained.

An example is shown in FIG. The price tag says "¥123456J", and the characters ¥, l, and 2.4.5.6 are written with print quality that can be read on the first thread, and the character 3 is written on the second thread. Assume that the print quality is such that it cannot be read otherwise.The recognition result using the first threshold is shown in (a), and the recognition result using the second threshold is shown in (a).
Shown in b). (a) So what are the 30 characters? It is rejected (unreadable) as shown in . In ■), the character 3 is recognized correctly, but ¥ is rejected.
If we look at each character in line 0, where l is mispronounced as 7, we get
Correct recognition is achieved using either the first threshold or the second threshold, but in conventional optical character reading devices, the thresholds are switched on a line-by-line basis, so correct recognition results cannot be obtained.

This invention was invented to solve the problems associated with cutting (negative) thresholds on a line-by-line basis, and by selecting the recognition result of each threshold on a character-by-character basis, printing can be performed between characters in a line. The objective is to realize an optical character reading device that can correctly read price tags with varying quality.

(Structure of the Invention) In order to switch the thresholds on a character-by-character basis, it may seem like it would be sufficient to select the recognition result by comparing the recognition results of each thresh on a character-by-character basis from the beginning of the line. If there is a missing character, it is not possible to take correct action just by matching the characters.
In the present invention, the recognition results are selected after the recognition results are associated with each character using positional information within the field of view.

FIG. 1 shows an example of the configuration of the present invention.

1.2.3.4.5.6.7, IS has the same function as the part with the same symbol in FIG. 14-1 to 14-
N is an N number of binarization circuits (N is an integer such that N≧2) having different thresholds, and 8-1 to 8-N are images that store the outputs of circuits 14-1 to 14-N, respectively. It's memory. 15 is a threshold selection circuit, which selects the image memory 8-1 according to instructions from the threshold switching circuit 16'.
- 8-H is sent to each character recognition section 20.

Each character recognition section 20 recognizes each character included in the binarized image of the visual field size, and sends the character recognition result and the position information of the character within the visual field to the threshold switching circuit 16'. The threshold switching circuit 16' is a recognition result buffer 17-1 corresponding to each of the binarization circuits 14-1 to 14-N.
17-N is selected and the character recognition result and character position information are sent.

Recognition result buffers 17-1 to 17-N store character recognition results and character position information. 18 is a digit alignment processing unit, which extracts recognition results corresponding to the same character and combines them into a set based on the character position information stored in the recognition result buffers 17-1 to 17-N, and selects recognition results. Send to Department 21. The recognition result selection unit 21 selects and outputs the final recognition result according to the priority order from the character confirmation recognition results obtained corresponding to each threshold.

The number and configuration of the binarization circuit, image memory, etc. up to the recognition result buffer can be other than those shown in FIG. 1, and will be explained in the examples.

(Operation) By pressing the push button switch 19, the image sensor 6
The image captured by the binarization circuit (14-1 to 14-N)
The images that have been binarized using different thresholds are stored in image memories 8-1 to 8-N. Each character recognition section 20 recognizes each character in the field of view and outputs character recognition results and character position information, but the threshold switching circuit 16
' and the threshold selection circuit 15, the image memory 8-
The binarized images of each threshold stored in numbers 1 to 8-N are sequentially processed in each character recognition section 20, and the results are sequentially stored in recognition result buffers 17-1 to 17-N. Once the character recognition results and character position information corresponding to each thread are stored in the recognition result buffers 17-1 to 17-N, the digit alignment processing unit 18 compares the character recognition results based on the character position information. , the final recognition result is selected and outputted by the vQ recognition result selection unit 21 from among the set of character recognition results obtained by the zero digit alignment processing unit 18 which obtains a set of character recognition results for the same character.

〔Example〕

In FIG. 1, although each character recognition part 20 does not exist in IMi, N character recognition means are provided corresponding to each thread, and each i (1≦i≦N) character recognition means is
If it is provided between the th image memory and the i-th recognition result buffer, a configuration can be obtained in which the threshold selection circuit 15 and the threshold switching circuit 16' are omitted.

In the first configuration, different binarization results for the same image captured by the image sensor are simultaneously stored in the image memory, so that the digit alignment processing unit 18 generates character recognition results for each thread using character position information. It is now possible to match the If the time required for each character recognition part to recognize all characters in the field of view is extremely short, digit alignment processing can be performed even with a configuration in which each character recognition part sequentially obtains binarized images of different thresholds and performs recognition. be.

This is because even when the scanner 1 is moved relative to the paper 3, the position of the characters within the field of view hardly changes within an extremely short period of time. Furthermore, in order to obtain binarized images with different thresholds sequentially, it is not necessary to prepare multiple entire binarization circuits, but by changing the threshold from outside the binarization circuit, one binarization circuit can be used. It is also possible to obtain binarized images with different thresholds.

When multiple characters enter the field of view, the character X is used as character position information.
Although it is necessary to use the coordinate and Y coordinate, if only one line is included in the field of view, only the X coordinate of the character can be used as character position information.

In the embodiment of the present invention using the microprocessor shown in FIG. 6, the binarization circuit 14 has only one 0 image memory 8 whose threshold can be switched by the threshold switching signal from the microprocessor. Here, each character recognition section 20 is configured to sequentially recognize binarized images of different thresholds. Numerals 9 to 13 are examples of each character recognition portion 20, and have the same configuration as the portions with the same reference numerals in FIG. However, the - digit extraction circuit 9 sends the image to the microprocessor as the X coordinate a of the character at the coordinate when the image is extracted from the screen memory 8 to a partial memory. In this embodiment, the recognition result buffer is connected to the microprocessor. In the storage area provided in the RAM 26,
The functions of the digit alignment processing section 18 and the recognition result selection section 21 are realized by software stored in a ROM 25 connected to a microprocessor.

Flowcharts of the processing to be executed by the microprocessor when three types of thresholds are used are shown in FIGS. 7 to 9.

After the microprocessor 24 receives the X-coordinate a of the character sent from the partial extraction circuit S and the character recognition result sent from the character recognition circuit, they are stored in the RAM 26 three times by shifting the zero time and changing the threshold. For each image data obtained, the obtained X coordinate a and identification result are once stored in the RAM 26, and then read out one digit at a time by the microprocessor 24, which recognizes that the digits are the same as those stored in the ROM 25. It is checked whether the range of the X coordinate a of the character is satisfied or not, and one or more recognition results that are recognized as having the same digit are selected. If there is only one recognition result that is recognized as the same digit, that recognition result is output as the recognition result of the character of that digit, and if there are multiple recognition results, the final recognition result is output according to the predetermined priority order for the threshold. Select and output.

The digit alignment process using a microprocessor is shown in more detail in FIG. Now, the recognition result of the first thread is ■Character present, i-th (i=1.2...
・・・・I) characters and their X coordinates are respectively cJll,
xjll, the recognition result of the second thread is that there is a J character, the jth character (j = 1.2...J) and its X coordinate are respectively represented as C and X, and the third There is a character in the thread recognition result, kth, (k=1.2...
...K) character and its X coordinate as ck's;
x, ;i7, and the characters of the recognition results of the first thread, second thread, and third thread after digit alignment are 171% yp: y731 (ffi=1.2.
......L).

In the flowchart of FIG. 8, initialization is performed in steps 1 and 2. ■ is a process that indicates the end of the recognition result of each screen in advance with a special large value M. 0M is larger than the maximum value of the X coordinate that can be obtained as a result of the recognition process plus the value of W, which will be described later. (■ is the value we are looking at, the coordinates of the i-th character x j I E the coordinates of the j-th character X, +z4 the coordinates of the k-th character x, +3t7)
This is the process of finding the minimum value x, , 7. In case (2), if Xei* is the same as M used in the process (2), it is determined that the digit alignment process for all recognition results has been completed. W used in (2) indicates the width of the X coordinate that can be determined to be the same digit. Now, if x:”Sx□7+W holds, then x r”l! The processing when l<enters the digit currently being processed, that is, the processing of ■, ■ is performed.■ is the processing of registering the first character C1) as the recognition result y, +1) after digit alignment. (2) is a process of increasing the value of i since the processes of c and ++ have been completed.

On the other hand, if x, +1 does not fit into the digit currently being processed, process (■) is performed. The process of ■ is the process of registering the symbol # that indicates that no recognition result has been obtained in the corresponding digit as y, (1. In the same way as ■ ~ ■, the process for the recognition result on the second screen ■ ~ @ , Perform processing ■ to [phase] for the recognition result on the third screen. ■ is a process that increases the number of characters 2 in the recognition result after digit alignment by 1. [phase] is the final number of characters obtained (i! , -1) are registered as variables.

In addition, although Figure 7 shows a flowchart of "selecting the final recognition result from the recognition results of all threads" after "digit matching of recognition results of all threads" is completed, digit matching is performed for each digit. It is also possible to perform selection processing.

Figure 9 shows the first and second threads for digit alignment.
Character y no 1 mi y of the recognition result of the third thread; 1. y
This is a flowchart of the process of selecting the final recognition result r from among I'' (ffi=1.2...L) according to the priority order.
Priority is given to yI Marimo y;l. That is, y;+! 7. <If it is not rejected or missing characters (y; 1 force (if recognized as a character) 71)
%, , y: l b<reject is missing characters and y, 1) If %<rejects are neither missing characters, set yI to r1. ;1) and y: R′h<reject or missing characters and y; 3b<reject even if there are missing characters i”Mofl Lt lever)': ” to r, (!
:L1. ;1琶y;1' If both toy and +21 are rejected or missing characters, set the jet to ``.

In ■■■ of Fig. 9, if y; I b< character missing (when the # symbol is set in the digit alignment process of Fig. 8) or reject (indicated by the ? symbol), then 7t'J"r, change to y deer (if the letter is missing, go to ■, in ■■■, y: Z
S<If the character is not missing or rejected, change y;z to “
, yI force (if it is a jet with a missing character, proceed to
If is a missing character jet, proceed to [phase]. [phase]
After all, proceed to y, +J! : y a''7y, +3 are both rejected or missing characters, and in this case, the jet is set to rl.

In the embodiment of FIG. 6, one
The final recognition result is obtained from the recognition results obtained each time, but the final recognition result is obtained from the majority vote of the recognition results obtained by performing recognition multiple times for one threshold. Embodiments of results are also possible.

(Effects of the Invention) According to the present invention, it is possible to realize an optical character reading device that can select recognition results based on each threshold on a character-by-character basis. Since recognition results based on different thresholds can be selected for each character, even price tags containing a mixture of characters of grade 6 print quality can be read without any problem, greatly contributing to improving the performance of optical character reading devices.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of a configuration example of the present invention, Fig. 2 is a block diagram of an optical character reading device according to the prior art, and Figs. Figures (a) to (C) in Figure 4 are illustrations of - character extraction method; Figure 5 is an illustration of an example of reading a line with variations in print quality for each character; The figure is a block diagram of an embodiment of the present invention, and FIGS. 7 to 9 are flowcharts of microprocessor processing. 1...Scanner, 2...Hand, 3...
...Paper, 4...Illumination light source, 5...
Lens system, 6... Image sensor, 7...
...Drive circuit, 8.8', 8-1.8-2.8-N... Image memory, S...- Digit extraction circuit, 10...-
Digit memory, 1)...Character extraction circuit, 12
......Character memory, 13...Character identification circuit, 14.14', 14-1.14-2.14-N.
... Binarization circuit, 15 ... Threshold selection circuit, 16 ...
・Threshold switching judgment circuit, 16'...Threshold switching circuit, 17-1.17-2.17-N...
... Recognition result buffer 18 ... Digit alignment processing section, 19 ... Push button switch, 20 ...
・Each character recognition part, 21...Recognition result selection part, 24...Microprocessor, 25...
・ROM. 26...RAM.

Claims (2)

[Claims] (1) In an optical character reading device that optically reads characters, symbols, etc., there is an image sensor that is made up of photoelectric conversion elements arranged in a planar shape and that can capture one or more lines of characters, symbols, etc. in its field of view, and this image sensor. A binarization circuit that binarizes the analog signal output from the sensor into binarized signals corresponding to a character area and a background area using a plurality of different threshold values; A character position detection means for detecting the position of a symbol, etc., an identification processing means for recognizing each character, symbol, etc. in the visual field from a binary signal, and a character position and recognition result obtained corresponding to each threshold value. A recognition result buffer for storing, and a digit alignment processing unit that performs digit alignment for each character or symbol based on the character position of the recognition results of characters, symbols, etc. obtained in response to the binarized signal based on each threshold value. and recognition result selection for selecting one final recognition result for each digit according to a predetermined threshold priority order from among the recognition results of each threshold value for each digit arranged by the digit alignment processing unit. An optical character reading device comprising: (2) In the optical character reading device according to claim 1, the field of view of the image sensor is the size of one line, and the character position detection means is means for detecting the horizontal position (X coordinate) of the character. An optical character reading device characterized in that a digit alignment processing section performs digit alignment of recognition results based on the X coordinate.