JPS5842498B2 - Kosuhanbetsouchi - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a number discriminating device used to carry out the ``number recognition method'' of Japanese Patent Application No. 115028/1983.

The method described in Japanese Patent Application No. 115028/1983 is designed to recognize each ring-shaped pattern as one pattern as shown in FIG.

The principle of the above invention will be explained based on FIG.

A new letter A is drawn at the first grid point A of each intersection of the scanning line and the clock pulse (hereinafter referred to as a grid point) by sequentially scanning the ring pattern shown in 1 in Figure 2 two-dimensionally. The same letter A for each grid point in the generated pattern
When the character A is generated, the character A branches left and right and then recombines, and the character A appears at the end point of the pattern.

The umbrella-shaped pattern shown in 2 in Figure 2 is scanned in the same way, and a new character B is generated at the point A, and the same character B is generated at each lattice point in the pattern. The character B appears at the end of the pattern.

The cotton-like pattern shown at 3 in FIG. 2 is scanned in the same manner to generate new characters C1D and E at each of the points A, C, and C, respectively.

The letter C generated at point A and the letter spelling generated at the white point make it recognized that the pattern of C and the pattern of D are continuous at point two, and the letter C is replaced with the letter letter C.

Next, the letter E generated at point C is the pattern of D and E at point H.
Recognize that the pattern is continuous and replace the character with the character E.

In this way, the letter E will appear at the end of the cotton pattern.

In this way, the area to be inspected is sequentially scanned two-dimensionally, new characters that are different from each other are generated at the scanning end point of each pattern, and the continuity between the patterns of each character is checked sequentially. In this case, one character can be replaced with the other character, and one character can be generated in one pattern at the end point of pattern scanning, and the number can be recognized.

Further, the number of particles can be determined by the number of characters generated.

Then, if you assign a letter for each particle to each particle and memorize the area, diameter, circumference, etc. of each pattern in correspondence with the letter, you can later use the particle size distribution and circumference distribution of the entire particle. It is convenient for quantitative measurement of etc.

To generate characters like the one above, do the following:

That is, as shown in FIG. 3, lattice points are arranged at equal intervals on a two-dimensional electronic scanning line, and four lattice points M1. ■M2. ■M3. ■Extract M4 and group information (
sensor), and one point M4 among the four grid points is set as a reference.

Note that 0 points and 0 points are grid points to be stored.

For example, the number of scanning lines in one screen frame is 256 (vertical sampling), and the sampling frequency is 5 clock pulses.
There are 256 sampling points (horizontal sampling) on the 11 MHz scanning line, and the total number of grid points is 256×256.

Note that the number of characters is 256 if expressed in 8-bit binary.

Characters used in one pattern can be reused by refreshing them when the scanning of that pattern is completed.

By the way, the sensor scans the screen sequentially starting from the top left corner and ending at the bottom right corner.

The information in the sensor is a binary signal 1 m, which will be described later.
When it is 2, it is represented by (), and when it is character information, it is represented by 3 m4 1 M2 ().

Note that as the sensor 3M4- moves, characters in the character memory are generated, replaced,
Disappearance (hereinafter collectively referred to as allocation) is processed according to the following rules based on the state of the sensor information ( ) of the binary l m2 signal.

O 1, the information in the sensor of the binary signal is ( ) or 0 By the way, the replacement character correspondence table in Table 2 is a table showing whether or not each character is replaced with respect to each of the 256 types of characters.

For example, in the replacement character correspondence table, when B is written as a replacement character for a used character A, this means that the used character needs to be replaced with the replacement character B.

Write this situation as A=>B. In the replacement character correspondence table of Table 2, for example, when the replacement character is written as O for the character B used, the character B does not need to be replaced.

As a result of the comparison, if the replacement character is 0, the character written in M2 is placed in the reference point, ie, M4.

Also, as a result of the matching, if there is a replacement character, for example, if B is written as a symbol, the character A written in M2 is replaced with B, and the character B is set as the reference point, that is, M4.
Put it in.

I IV) For binary signal sensor information (): 1 Check the characters written in M2 with the replacement character correspondence table in Table 2.

If a replacement character is written for the character M2 in the replacement character correspondence table, after that replacement, M2 and M
3 is compared, and M2\M3 is flf.This indicates the combination of left and right patterns to which different characters are assigned, and in that case, a replacement character is generated at the position corresponding to the replaced character M2.

At that time, for convenience, M3+M2, for example, A, >B, is written in the replacement character correspondence table of Table 2.

Then, the character written in M2, for example, the character B, is placed at the reference point, that is, the pigeon.

0 V) When the sensor information () is a binary signal: Check the characters written in IM□ with the replacement character correspondence table in Table 2.

As a result of the matching, if there is no replacement character, that is, if O is written, then the character written in M□ is set as the reference point, that is, M
Put it in 4.

Further, as a result of the matching, when there is a replacement character, for example, when C+B is shown as a symbol, the character C written in M1 is replaced with B, and the character B is placed at the reference point, ie, M4.

0 VI) For binary signal sensor information (): 1 M at the reference point, that is, M4, without checking with the replacement character correspondence table.
Enter the characters that were in 3.

1 ■) When the information in the sensor is a binary signal (): 1 Similarly, it is standard, that is, the character that was in M2 is entered in M4.

I Vl) When binary signal sensor internal information (): 0 Similarly, place O at the reference point, that is, the pigeon.

Then, compare the characters written in M2 and M3 and M
If it is 2\M3, a replacement character is generated at the position corresponding to the character M3.

At that time, for convenience, the replacement character correspondence table in Table 2 includes M3;>M
2 For example, write A=>B.

Next, concrete examples will be described regarding four model patterns shown in FIG. 4.

It is assumed that the binary signal M4 of the sensor is sequentially scanned starting from the grid point (1, 1), and ends at the grid point (256, 256).

The above 2rn1 is applied to each grid point in the four model patterns.
Based on the state of the sensor internal information () of the rn2 value signal, characters are assigned according to the character allocation determined based on the 3 m4 standard.

As a result, the character arrangement becomes as shown in FIG.

There are the following two methods for recognizing the number of characters from the arrangement of characters in a pattern.

That is, the first number recognition method is to check the discontinuity of the characters used.

First, as the sensor moves, characters are allocated and the characters used in the scanning line are registered in a used character registration table (not shown).

Then, the usage status of the characters used in the nlyn-th two scanning lines is registered.

At the end of scanning line n, compare the characters registered in the n-1st scanning line and the n-th scanning line, register them in the n-1st scanning line, and register them in the nth scanning line. Find the missing characters.

That is, the found character is a character that disappears in the n-th scanning line, and the disappeared character may be generated by character replacement or by termination of a pattern.

Therefore, by comparing the annihilation character with the replacement character correspondence table in Table 2, if the character has not been replaced by another character, it can be determined by the annihilation character that the pattern has ended on the n-1th scanning line. .

Regarding the end of a pattern, end information is obtained at each of the locations A, C, C, and 2 in FIG. 4, and one pattern can be recognized as one by the end information.

The total number of screens can also be counted by counting the end information.

In short, the end information of the pattern can be obtained at each of the locations A, B, C, and B. This information is obtained by comparing the characters appearing in the current scanning line ○
The character registered in CNM but not registered in CNM is compared with the replacement character correspondence table (Table 2), and if it is not replaced with another character, it means that the pattern to which that character is assigned has ended. do.

For example, in FIG. 4, when PNM has scan number 6 and CNM has scan number 7, the characters that are in PNM but not in CNM are F and H.

When comparing the characters F and H with the replacement character correspondence table, F becomes D.
It can be seen that H is substituted with B, respectively.

Therefore, each character indicates that the assigned pattern is not yet finished.

Next, considering PNM 7 and CNM8, B is the corresponding character and when B is checked against the correspondence table, it is found that B has not been replaced with any other character.

Therefore, if you can obtain information that the pattern B ends at scanning line 7 and store the area of the pattern allocated in B in another memory at the same time as the end, you can calculate the area distribution of multiple patterns at the end of one frame. Information can be provided to the desired software program.

The area of the B pattern can be determined by the total number of times B characters are assigned to each grid point.

If A is replaced with B, the number of A may be added to the number of B at the time of replacement.

Further, when B is finished, the characters used in the B pattern, such as B, A, E, and H, can be reused by clearing the corresponding replacement character correspondence table.

The second number recognition method is a method described in Japanese Patent Application No. 48-58716, which removes the 11 count generated from the negative pattern.

That is, in the ring-shaped pattern shown in ■ in Fig. 4,
The binary signal matrix for counting +1 described in Japanese Patent Application No. 1987-58716 has the symbol t and -1
These are the binary signals Ma I-IJ lux, symbols E and E, which count .

The sensor matrix of the character information in the symbols E and H is M2\M3 in the case of the symbol E, and M2=M3 in the case of the symbol H.

The sensor matrices of character information for symbols E and F are information indicating that the pattern containing M2 and the pattern containing M3 are continuous.

However, the character information sensor I shown in symbols E and F is
-IJ Lux, different from each other.

That is, the sensor matrix in the symbol H is generated when the sensor matrix branches off and then recombines like the ring-shaped pattern shown in 1 in FIG. This occurs when characters that are different at the beginning of a pattern are combined, as in a pattern.

As described above, information indicating pattern continuity includes information generated by branching and combining of patterns shown at 1 in FIG. 2, and information generated only by the combination shown at 3 in FIG. 2.

Therefore, the 11 count generated from the negative pattern is generated by pattern branching and combination, so in order to recognize the ring pattern as one, check M22M3 of the character information matrix and if M2\M3, 11 If it is M2-M3, count 11 and C
Just handle it as you see fit.

In short, this invention allows one pattern to be recognized as one even when the shape of the pattern is complex and ring-shaped, so the number of patterns, particle size distribution, area distribution, length, etc. can be easily determined. can.

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a number discriminating device for realizing the method of the invention will be described below with reference to the drawings.

In the present invention, as shown in FIG.
The video signal 4 is applied to a comparator circuit 5 to be compared with a reference voltage 6, and an AND circuit 6'
generates a binary signal which takes each intersection of the scanning line and the clock pulse as a grid point and becomes "1" when an image exists at the grid point, and "0" when there is no image, and converts this into an image signal 7.
As shown in FIG. 6, this is added to the shift register 8 for serial-to-parallel conversion, the speed of which is reduced to H, and then written to the pixel memory 9.

In this embodiment, the capacity of this pixel memory 9 is 25 in the horizontal direction.
6 bits, vertical force direction 256 bits, total 65.536 bits.

The horizontal synchronization signal of the TV camera 3 has a cycle of 64.5 μs as shown in FIG.

In this embodiment, the effective screen horizontal scanning time required for image processing is 51.2 μs.

Therefore, if the clock pulse for writing the image signal 7 to the pixel memory 9 is set to 5 MHz, 2
56 pieces of image information are obtained.

- Force Since the vertical direction is 60 Hz synchronous 2:1 interlace, one rusk consists of 52,571 scanning lines at 1/60 second.

By making this compatible with 256-bit memory in the vertical direction,
Only one frame is written to the image memory 9.

The horizontal clock is 5 MHz, and the memory access time must be 200 ns or less, but since current semiconductor memories are insufficient in speed, this binary active code is serial-parallel converted by the shift register 8 as described above. and reduce the speed to H.

The horizontal memory configuration of the pixel memory is 32×8 bits, so that an 8-bit image signal 7 can be input in parallel.

The write address at this time is generated by the write address generator 10 using the horizontal synchronization signal H of the TV camera 3, the vertical synchronization signal (2), and the clock pulse Cp.

The image information from the pixel memory 9 is read and outputted in 8-bit parallel by the salt address generator 11.

This signal is subjected to parallel-to-serial conversion using a shift register 12 to match the original image signal 7 in time series.

The output of the shift register 12 is m4 of the m sensor in FIG.
It is input to the register 13 for m as follows.

As shown in FIG. 6, the information of the m matrix of the m register 13 and the M matrix of the M register 14, which will be described later, is input to the computer 15, and as explained in FIG. 3 and Tables 1 and 2, Based on this information, it is determined which character to assign to the information point of m4, and M4 is output.

M4 is input into the character storage memory 16 for one scan.

It is assumed that this memory 16 has a capacity of, for example, 256×8 bits.

Therefore, its output is the address of the input character plus 25
6, that is, the character from one scan ago, and thus becomes the character M2.

This M2 is input to the M register 14 and simultaneously input to the 8-bit register 17.

Further, the M4 output from the computer 15 is input to an 8-bit register 18.

A control signal for moving the output from the computer 15 to the right by one grid point causes registers 18 and 17 to output M31M1, respectively.

This M39M1 is input to the M register 14.

In addition, M32M12M2 is connected to the m detection circuit 19, 20°21
Depending on whether there is a character signal, set it to 1 if there is a character signal, or set it to O if there is no character signal, and calculate m31 m11 m2 and input them to the m register 13.

In the computer 15, depending on the state of M1-2M22M3, characters are assigned to M4 according to Tables 1 and 2.

Furthermore, each time one horizontal scanning line ends, it is determined whether character end information is obtained, as described above.

As described above, in the present invention, different characters are assigned to each of the many patterns in an image, the number of times these characters occur is counted separately for each character, and the area group of each pattern is can be found.

The area distribution of many patterns in the image can be determined from this area group, and if the pattern can be approximated to a circular shape, the particle size distribution can also be determined from this area group.

[Brief explanation of the drawing]

Figures 1 and 2 are explanatory diagrams of patterns discriminated by the device of the present invention, Figure 3 is an explanatory diagram of the arrangement of sensors and character memory memory, and Figure 4 is an explanatory diagram showing the state in which characters are assigned to patterns. , FIG. 5, and FIG. 6 are respectively of the present invention C.
The circuit diagram showing the main parts, and FIGS. 7 and 8 are explanatory diagrams thereof. 1... Image, 2... Optical system, 3...
...TV camera, 4...Video signal, 5...
...Comparison circuit, 6...Reference voltage, 6 (...
...AND circuit, 7... Image signal, 8...
...Shift register, 9...Pixel memory,
10...Write address generator, 11...
...read address generator, 12...shift register, 13...m disk, 14...
...M disc, 15...Computer, 1
6...-Memory for storing characters for scanning-17, 18
...8-bit register, 20.21...
・m detection circuit. 19°

Claims (1)

[Claims] 1 Scan the area to be inspected two-dimensionally and scan the area at predetermined intervals 2
A mechanism that generates value information, a pixel memory that sequentially stores this binary information, and a pixel memory that receives four pieces of binary information stored at four adjacent points on the upper, lower, left, and right sides of this memory as one set of information. A binary register that receives character information, a character register that receives character information, and a character register that receives and receives output from the binary register and character register, determines the contents of the four binary information, and determines the contents of the four binary information. A mechanism for allocating characters in the character register to value information, and a mechanism for sequentially assigning character information on the completed scan line and character information on the previous scan line each time each scan line ends. A mechanism for comparing and determining whether a character on the previous scan line but not on the completed scan line has been replaced with another character, and generating end information when the character has not been replaced. A number determining device consisting of a mechanism and a mechanism that adds this end information.