JPS6141030B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an optical character reading method for reading characters, symbols, etc. that are larger than the field of view of a sensor.

In a general optical character reading device, as shown in FIG. The analog signal from the sensor 2 is converted into a binary signal by a binary conversion circuit 3 corresponding to the white of the paper and the black of the characters (hereinafter, the binary signal is referred to as a black and white pattern), and the recognition circuit 4 recognizes the characters. This is what we do. At this time, the image sensor 2 is electrically scanned in the vertical direction, and when the image sensor 2 is a two-dimensional sensor, it is electrically scanned in the horizontal direction 8, and the optical system or paper is mechanically moved to form the characters. When the image sensor 2 is a one-dimensional sensor, the optical system or paper is mechanically moved to read the characters. Note that the paper 5 is illuminated by a light source (not shown).

When reading characters optically as described above, 1
In order to capture characters two-dimensionally by the dimensional sensor, it is necessary to move the image sensor 2 in the horizontal direction 8 at a constant speed relative to the paper 5. Furthermore, if the moving speed of the image sensor 2 with respect to the paper 5 is not known, it is necessary to store black and white patterns over the entire area of the character, and in this case, the storage capacity must be increased, which is expensive.

In addition, even if a two-dimensional sensor is used, if the moving speed relative to the paper changes, as shown in Figure 2a,
This method could only be applied when the entire character 6 to be read falls within the field of view 10 of the sensor. Therefore, when a character 6 larger than the horizontal width of the field of view 10 of the sensor is to be read as shown in FIG. It is necessary to store black and white patterns in parallel to each other, and in such a case, the storage capacity needs to be larger than that when a one-dimensional sensor is used.

Furthermore, regarding the idea of using a two-dimensional sensor with a large bit configuration to process large-width characters, (i) large-sized two-dimensional sensors are expensive, and (ii) the identification process takes a lot of time. There are drawbacks such as:

For this reason, conventionally it has been difficult and economically disadvantageous to read characters larger than the field of view of the sensor.

The present invention has improved the conventional drawbacks as described above, and its purpose is to facilitate the reading and identification of characters larger than the field of view of a two-dimensional sensor. Examples will be described in detail below.

FIG. 3a is an explanatory diagram of the principle of the present invention, showing the black and white pattern when the left end of the character 6 to be read reaches the left end of the field of view 11 of the sensor, and the character 6.
The black-and-white pattern when the right end of the screen reaches the right end of the field of view 12 of the sensor is extracted, and the characters are recognized using the black-and-white patterns of these two screens.

Therefore, even if the width of the character 6 to be read is large and the field of view of the sensor is 11, 12 or more,
The left screen shown in FIG. 3b and the right screen shown in FIG. 3c can be read out regardless of the horizontal movement speed.

FIG. 4 is a block diagram of a left and right screen detection circuit of an optical character reading system according to an embodiment of the present invention. In the figure, reference numeral 3 indicates an analog signal read by a two-dimensional sensor in which photoelectric conversion elements are arranged two-dimensionally.
It is a binarization circuit that converts into values, and if a cell that detects black in each cell of two-dimensionally arranged photoelectric conversion elements is called a black cell, the output of the binarization circuit 3 is generated in each column in the vertical direction. The black cell detection circuit 13 for each column detects the presence of a black cell, and the temporary storage buffer memory 14 temporarily stores the black and white pattern. Then, in the column-by-column black cell detection circuit 13, the left-most black cell detection signal 1 detects the black cell in the left-most column.
5, and when there are consecutive black cells in several lines on the left side excluding the leftmost column, it outputs the left side character element detection signal 16, and on several lines on the right side excluding the rightmost column. When there are consecutive black cells, a right character element detection signal 17 is output;
A right-most black cell detection signal 18 is output when a black cell in the right-most column is detected. Note that the several lines mentioned here are the number of lines corresponding to the minimum width of the characters to be read.

The above-mentioned column-by-column black cell detection circuit 13 specifically has the configuration shown in FIG. 5, and this configuration will be explained for the case where the binarized signal is input in parallel (n bits in the horizontal direction) in units of rows. .

In such a case, the column-by-column black cell detection circuit 13 receives binary signals BW1 to BW1 corresponding to each column in the vertical direction.
In addition to BWn, there is a line clock CK1 synchronized with the binarized signal for each row (1 to m) in the horizontal direction, a frame clock CK2 synchronized with the processing of one screen, and a clear signal CL1 slightly delayed from the frame clock CK2.
is input, and for black cell detection,
A 1-bit register 100 and an OR circuit 101 are combined for each column, and a line clock
The data in the register 100 is updated by CK1. Further, 102 is an output register, and 103 is an AND circuit.

The time chart of each signal in FIG. 5 is as shown in FIG. 6. However, in FIG. 6, the same symbols as in FIG. 5 indicate the same signals.

When the first row of data in the binary signals BW1 to BWn is applied to the OR circuit 101, the first row of data is transferred to the register 1 by the line clock CK1.
00, and thereafter, the logical sum circuit 101 calculates the logical sum with the data stored in the register 100 up to the mth row data, and outputs the line clock CK.
1 is set in register 100. Therefore, when a black cell exists in a column, the contents of register 100 become "1". In the columns at both ends, the content of the register 100 is directly set to the corresponding output register 102 by the frame clock CK2, so the left end black cell detection signal 15 and the right end black cell detection signal 18 are set to the corresponding output register 102.
It is set to 2. Also, the contents of the registers 100 on several lines on both sides excluding both ends are added to the AND circuit 103 on each side, and the output of this AND circuit 103 is set to the corresponding output register 102 by the frame clock CK2, so that the characters on the left side are The element detection signal 16 and the right character element detection signal 17 are set in the corresponding output registers 102. or,
In this way, after the screen data is set in the output register 102 by the frame clock CK2, the register 100 is cleared by the clear signal CL1.

Further, as described above, the binary signal of the black and white pattern stored in the temporary storage buffer memory 14 is written to the left screen buffer memory 23 while the left screen detection signal 21 is "1", and the right screen detection signal 22 is "1". 1", the left screen is written to the right screen buffer memory 24, and the left screen with the left end as a reference and the right screen with the right end as a reference are buffer memories 23 and 24, respectively.
is stored in The left screen detection signal 21 is generated by adding a signal obtained by inverting the left end black cell detection signal 15 by an inverter 15a and a left character element detection signal 16 to an AND circuit 19.
It is output from the right screen detection signal 2.
2, the right side character element detection signal 17 and a signal obtained by inverting the right end black cell detection signal 18 by the inverter 18a are applied to an AND circuit 20, and outputted from the AND circuit 20.

In the above, the relationship between the black and white pattern on the sensor and the distribution of black cells in the vertical columns can be changed from, for example, FIG. 7a to FIG. 7b to FIG. It can be changed as shown in Fig. 7c and Fig. 7d.

However, in FIGS. 7a to 7d, x1 to x4 indicate a black and white pattern based on the relationship between the field of view 10 of the sensor and the character 6, and y1 to y4 indicate a portion 25 in which no black cell exists in the vertical column. Distribution of black cells y1 to y4 in relation to the portion 26 where the black cells exist
shows.

The black and white pattern x1 in FIG. 7a is when the left black cell detection signal (LBF) 15 is "0" and the left side character element detection signal (LCF) 16 is "1", and the scanning speed is slow. Screens with the same positional relationship can be obtained continuously any number of times. Therefore, in order to extract only the screen that satisfies the above-mentioned conditions of LBF being "0" and LCF being "1" for a certain character as the left screen, the following processing is performed.

When the above conditions are met, the left screen flag (LF) is set to "1"; otherwise, it is set to "0".
Further, if the left screen flag immediately before the left screen flag to be executed is BLF, only when the complements BLF and LF of this BLF are BLF ·LF=1, the black and white pattern at that time can be extracted as the left screen.

Furthermore, the black and white pattern x3 in FIG. 7c is when the right side character element detection signal (RCF) 17 is "1" and the right edge black cell detection signal (RBF) 18 is "0", and when these conditions are satisfied In this case, the right screen flag (RF) is set to "1", and in other cases, it is set to "0".
Also, if the right screen flag immediately before the right screen flag that is executed now is BRF, the complements of this BRF, BRF and RF, are
Only when BRF・RF=1, the black and white pattern at that time can be extracted as the right screen.

For this reason, the data of the black and white pattern x1 in FIG. 7a is stored in the left screen buffer memory 23.
The black and white pattern x3 in Figure 7c is stored in
data is stored in the right screen buffer memory 24.

As described above, the present invention provides the left screen buffer memory 2.
Identification processing is performed using the black and white pattern data stored in the left and right screen buffer memory 24. This identification processing method involves extracting screen features from the left and right screens respectively, and combining these screen features to create characters. This method employs a method of identifying characters, or a method of superimposing black and white pattern data on the left and right screens and using a composite screen resulting from this superposition to identify characters.

FIG. 8 is an explanatory diagram of the principle of an identification processing method for identifying characters by extracting left and right screen features, and black and white pattern data on the left screen 30 and right screen 31 obtained by the left and right screen detection circuit of FIG. The logical sum of the black cells of each line in the horizontal direction is taken, and the area where a specified number or more of consecutive lines with black cells exist is defined as a vertical character area 33, and the other areas are defined as blank areas 32 and 34. Then, using the black and white pattern data in the vertical character area 33 on the left or right screen,
This extracts the features of the left or right screen.

For example, suppose one screen is composed of a black and white pattern of n×m bits (n bits in the horizontal direction and m bits in the vertical direction).
When the area up to 2 is the vertical character area 33,
First, the column features viewed vertically are extracted using the black and white patterns in the two leftmost columns, then the column features of the two black and white patterns shifted one column to the right are extracted, and in the same way, the column features using the rightmost column are extracted. This method extracts up to (n-1) column features, and such extraction is performed for the left and right screens.

FIG. 9 is a block diagram of the identification processing section according to the identification processing method explained in FIG.
6, a column feature extractor 35 constitutes a sequential circuit in combination with a latch circuit 38, a screen feature extractor 37 constitutes a sequential circuit in combination with a latch circuit 38, and a left screen code memory 39 and a right screen code on the input side. memory 4
0, and a left and right screen combining section 41 having

The column feature extraction unit 35 receives two columns of black and white pattern data 42 in 2-bit parallel from the left screen buffer memory or the right screen buffer memory m times corresponding to the number of bits in the vertical direction.
2 and the contents of latch circuit 36, and the output is latched into latch circuit 36. Therefore, the contents of the latch circuit 36 change sequentially, for example, as shown in the state transition diagram of FIG. . In Figure 10, if both 2-bit parallel inputs are white in the initial state a, the initial state a remains, but if both 2-bit parallel inputs are black, the state transitions to state b, and then the 2-bit parallel inputs are black. Transition to state c when both parallel inputs are white, then transition to state d when both 2-bit parallel inputs are black, then transition to state e when both 2-bit parallel inputs are white, and then transition to state e when both 2-bit parallel inputs are white. to 2
If both bit parallel inputs are black, the state transitions to state f, and if the 2-bit input is the final stage, state f is used as the column feature code 43 by the screen feature extractor 37.
is input.

The screen feature extractor 37 receives the column feature code 43 of the left screen or right screen (n-1) times from the latch circuit 36 of the column feature extractor 53, and uses the column feature code 43 and the contents of the latch circuit 38 as input. The output is latched into the latch circuit 38. Therefore, the contents of the latch circuit 38 are, for example, the first
The state changes sequentially as shown in the state transition diagram in FIG. It is inputted in a different way.

Note that, regardless of whether the screen feature extraction unit 37 extracts the features of the left screen or the right screen first, the screen features are extracted after the output from the left screen buffer memory or the right screen buffer memory to the column feature extraction unit 35 is started. Since the timing until the screen feature is obtained by the extraction unit 37 is known, the signals 45 and 46 for screen feature extraction are sent to the left screen code memory 39 and the right screen code memory 40 in synchronization with the timing at which the screen feature is obtained. can be added to. That is, the latch circuit 36 of the column feature extraction section 35 receives a clock synchronized with each shift of each cell in the vertical direction, and a clear signal in synchronization with the shift in the horizontal direction. Furthermore, a clock is input to the latch circuit of the screen feature extraction section 37 in synchronization with the horizontal shift and before the clear signal of the column feature extraction section 35, and the output of the column feature extraction section 35 is taken in and the right edge processing is performed. After that, a clear signal is input. Therefore, as mentioned above, the signals 45 and 46 are stored in the memories 39 and 4.
Can be added to 0.

In Figure 11, in the initial state a', the input a'-1
If there is an input b'-1, the state transitions to state b', and if there is an input b'-1, the state transitions to c', and even if there is only an input c'-1 that overlaps after this state c', it does not transition and becomes the final stage,
The state c' is written as a screen characteristic code 44 into the left screen code memory 39 or the right screen code memory 40. However, the screen feature is a code given to a part or the entire shape of a character to be read, and a plurality of codes are given to one character. There are also cases where different characters are given the same code. In this way, the screen feature code 44 is extracted at the state transitions shown as A, B, C, etc. in FIG.

Therefore, characters can be identified by combining screen feature codes as shown in FIG. 12. For example, when the left screen feature code L is A and the right screen feature code R is B, "5" If the right screen characteristic code R is C, it can be identified as "9". Further, when the left screen characteristic code L is B or C, if the right screen characteristic code R is C, it can be identified as "3".

The left and right screen synthesis unit 41 performs identification by combining the screen characteristics of both the left and right screens, as explained in FIG. The screen characteristic code 44 input into the right screen code memory 40 is accessed by the signal 46 and input as the lower address, and each address information is combined and identified, and the identification result 47
Output.

In the above explanation, in order to extract the screen features of the left and right screens, column features are extracted using two vertical lines, and screen features are extracted using the column features obtained in this way. More than two lines can also be used to extract features. or,
Using the same circuit as when extracting column features using multiple horizontal lines instead of column features,
It is also possible to extract screen features by extracting row features and using the row features thus obtained.

In addition, as an identification processing method, a method was explained in which screen features are extracted from the black-and-white patterns of the left and right screens, and then the screen features are combined to identify characters. It is also possible to apply a method to identify the characters by extracting the screen characteristics of the composite screen and identifying the characters.
A block diagram of the embodiment in this case is shown in FIG.

In FIG. 13, 50 is a left screen buffer memory, 51 is a right screen buffer memory, 52 and 53 are column number registers, and 54 is a dissimilarity degree for calculating the difference in the bit configuration of columns in the vertical direction (hereinafter referred to as dissimilarity degree). 55 is a minimum value register in which the minimum difference value is set; 56 is a column number register that stores the column number giving the minimum value; and 57 is a minimum difference comparison circuit.

The column number register 52 is set with the rightmost column number of the character pattern on the left screen set in the left screen buffer memory 50, and the column number register 53 is set with the column number of the character pattern on the left screen set in the left screen buffer memory 50. All column numbers are set sequentially.

According to the contents of the column number register 52, the bit pattern of the rightmost column number in the character pattern on the right screen is read out from the left screen buffer memory 50, and according to the contents of the column number register 53, the bit pattern of the rightmost column number is read out from the left screen buffer memory 50. The bit patterns of the numbers are read out and input to the difference calculation circuit 54, respectively. When the calculation circuit 54 performs the first difference calculation, the minimum value register 55
The degree of dissimilarity is set to . and calculation circuit 5
When the second and subsequent dissimilarity calculations are performed in step 4, the comparison circuit 57 compares the dissimilarity obtained by the calculation circuit 54 with the dissimilarity set in the minimum value register 55. , calculation circuit 54
When the degree of difference obtained is smaller, the minimum value register 55 is updated to that degree of difference, and the column number with the smaller degree of difference is set in the register 56.

By calculating and comparing the degree of difference with respect to the bit pattern of the rightmost column number of the right screen in this way, the degree of difference with the bit pattern of the rightmost column number of the left screen is stored in the column number register 56. The minimum column number on the right screen will be set.

When character identification is performed by combining the contents of the buffer memories 50 and 51, the left screen buffer memory 50
From then on, the column numbers are sequentially incremented from column number 1 to read out the black and white pattern. This process is repeated until the column number matches the contents of the column number register 52.

After the black-and-white pattern on the left screen is read up to the column of the contents of the column number register 52, the black-and-white pattern on the right screen is sequentially read from the contents of the column number register 56 up to the column number n.

In the example shown in FIG. 14, after the black and white data from the leftmost column to the rightmost column 70 is read out on the left screen (a), the degree of difference from the bit pattern in the column 70 is determined on the right screen (b). The black and white data from the smallest column 71 to the rightmost column is read and characters are identified.

Furthermore, in the example shown in FIG. 15, since the characters are contained within one screen each of the left screen (a) and the right screen (b), it is only necessary to overlap columns 72 and 73, and in reality the left screen The pattern is the same as that of black and white data.

As explained above, the present invention provides a black and white pattern when the left end of a character, symbol, etc. reaches the left end of the field of view of the sensor, and when the right end of a character, symbol, etc. reaches the right end of the field of view of the sensor. are stored in the left screen buffer memory and right screen buffer memory, and the left,
The left screen feature code and the right screen feature code are extracted based on the contents stored in the right screen buffer memory, and the left screen feature code (in the embodiment, the screen feature code stored in the left screen code memory 39) and the right screen feature are extracted. Since characters, symbols, etc. are identified by a combination with a code (in the embodiment, the screen characteristic code stored in the right screen code memory 40), if the characters, symbols, etc. are larger than the field of view of the sensor. It also has the advantage of being able to reliably recognize characters, symbols, etc. Furthermore, the present invention recognizes characters, symbols, etc. using only the black and white pattern when the left end and right end of the character, symbol, etc. reach the left end and right end of the sensor. Even when the relative movement speed between the paper on which the characters, symbols, etc. are written and the sensor is not constant, there is an advantage that the characters, symbols, etc. can be reliably recognized with a small storage capacity. Therefore, if the present invention is applied to an apparatus using a hand scanner, the effect will be very large.

Furthermore, since the sensor used in the present invention is a two-dimensional sensor, the advantage of being able to accurately capture the phase difference of characters, symbols, etc. is maintained.

[Brief explanation of the drawing]

Figure 1 is an explanatory diagram of a general optical character reading device.
Figures 2a and b are illustrations of the relationship between characters and the field of view of the sensor, Figures 3a to c are illustrations of the principle of the optical reading system of the present invention, and Figure 4 is an illustration of an embodiment of the invention. 5 is a block diagram of the column black cell detection circuit, FIG. 6 is a time chart of signals applied to the column black cell detection circuit, and FIGS. 7a to 7d are block diagrams of the left and right screen detection circuits. An explanatory diagram of the relationship between the black and white pattern and the distribution of black cells, Fig. 8 is an explanatory diagram of the principle of the identification processing method performed by extracting the left and right screen features, and Fig. 9 is an illustration of the identification processing using the identification processing method explained in Fig. 8. 10 is a state transition diagram of column features in the identification processing section of FIG. 9. FIG. 11 is a state transition diagram of image features in the identification processing section of FIG.
Figure 2 is an explanatory diagram of the process of combining and identifying screen features.
FIG. 13 is a block diagram of an embodiment in which an identification processing method is applied that extracts screen features of a composite screen, and FIGS. 14 and 15 are explanatory diagrams of specific examples of creating a composite screen. . 1 is an optical system, 2 is an image sensor, 3 is a binarization circuit, 4 is an identification circuit, 5 is paper, 6 is a character, 1
0, 11, 12 are sensor fields of view, 13 is a black cell detection circuit for each column, 14 is a temporary memory buffer memory, 15
a, 18a are inverters, 19, 20 are logic circuits, 23 is a left screen buffer memory, 24 is a right screen buffer memory, 35 is a column feature extractor, 36 is a latch circuit, 37 is a screen feature extractor, 38 is a latch circuit, 39 is a right screen code memory, 40 is a left screen code memory, 41 is a left and right screen combination unit, 50 is a left screen buffer memory, 51 is a right screen buffer memory, 52 and 53 are column number registers, 54 is a difference calculation circuit, and 55 is a minimum A value register 56 is a column number register, and 57 is a minimum value comparison circuit.

Claims (1)

[Scope of Claims] 1. While a sheet of paper on which characters, symbols, etc. are written and a sensor in which photoelectric conversion elements are arranged two-dimensionally are moved relative to each other at an arbitrary speed, the characters, symbols, etc. are recorded by the sensor. In an optical character reading method in which characters, symbols, etc. are read by scanning and processing signals read by the sensor, the left end of the characters, symbols, etc. is located within the field of view of the sensor based on the output signal of the sensor. a left edge detection means for detecting that the right edge of the character, symbol, etc. has reached the right edge of the field of view of the sensor based on the output signal of the sensor; means for creating a black and white pattern of an image formed by the sensor by binarizing the output signal of the sensor; and a left end detecting means for detecting the left end of the character, symbol, etc. a left screen buffer memory for storing a black-and-white pattern created by the black-and-white pattern creation means when it is detected that the part has reached the left end of the field of view of the sensor; a right screen buffer memory that stores the black and white pattern created by the black and white pattern creation means when it is detected that the camera has reached the right end of the field of view of the sensor; A left screen feature code extracting means for extracting a code, and a right screen feature code extracting means for extracting a right screen feature code based on the stored content of the right screen buffer memory, the left screen feature code extracting means and the right screen feature. An optical character reading method characterized in that the characters, symbols, etc. are identified by a combination of a left screen characteristic code and a right screen characteristic code extracted by a code extraction means.