JPH0552987B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a pattern detection device used in automatic positioning devices for articles and the like.

Conventional pattern detection devices capture a pattern image with a television camera and then convert it into a binary signal.
After storing this binary signal in a two-dimensional local memory, it is determined whether or not the patterns match by logical operation, or the binary signal is compared with a reference pattern signal previously stored in a reference pattern register. Some systems detect the degree of coincidence using a computer, or perform logical operations on the binarized signal using a computer.

However, along with converting the pattern imaging output into a binarized signal, the threshold for binarization must be optimized to account for the drift of the DC component of the imaging output, changes in the noise component, and changes in the shading brightness. However, it is difficult to automatically set this to the optimal value. In addition, the amplitude of the imaging output signal is related to the unevenness (surface condition) of the light-reflecting surface of the pattern image (sample), so in order to perform stable binarization, it is necessary to uniformly illuminate the entire sample and to stabilize it. Requires illuminance lighting device,
This is difficult. Furthermore, since binarization is performed, it is theoretically impossible to detect patterns in analog signal waveforms. Furthermore, the binary signal output (“1” or “0”) cannot be normalized, and the noise margin for processing this output data is small.

The present invention has been made in order to eliminate the various drawbacks mentioned above, and it is possible to obtain a pattern image by calculating a correlation coefficient between analog or digital input data and reference data representing a reference pattern prepared in advance. It is an object of the present invention to provide a pattern detection device that can stably detect the same or similar patterns even with variations in imaging output or analog waveforms at a high rate.

Hereinafter, one embodiment of the present invention will be described in detail with reference to the drawings.

FIG. 1 shows the basic configuration of the present invention, in which pattern input data P(x) (analog signal or digital signal) and reference data R(x) serving as a pattern detection standard are input to a correlation coefficient calculation circuit 10. guided by,
Pattern similarity is calculated by calculating the correlation coefficient f(x) based on the following equation (1).

However, K: integral range.

Note that the coefficient f(x) takes a value of +1 to 0 to -1, and "+1" indicates the maximum degree of coincidence.

Here, details of the calculation circuit 10 will be explained with reference to FIG. 2. In the first circuit 11, ∫P(x)・R(x)
The first term in the numerator of the previous formula (1) is obtained by calculation, and the second term is
The circuit 12 calculates ∫P(x)dx, and the third circuit 13 calculates ∫R(x)dx (since R(x) is fixed, it only needs to be calculated once).
The fourth circuit 14 calculates the product of the constant 1/K and the output of the third circuit 13, and the fifth circuit 15 calculates the product of the output of the second circuit 12 and the output of the fourth circuit 14. The second term in the numerator of the previous equation (1) is obtained. The sixth circuit 16 calculates the difference between the output of the first circuit 11 and the output of the fifth circuit 15, thereby obtaining the numerator of the above equation (1). The seventh circuit 17 calculates ∫P(x) ² dx, the eighth circuit 18 calculates the product of the constant 1/K and the output of the second circuit 12, and the ninth circuit 19 calculates ∫P(x) 2 dx. The product of the output of the second circuit 12 and the output of the tenth circuit 20 is calculated.
Now, the output of the seventh circuit 17 and the ninth circuit 19
The 11th circuit 21 calculates the difference between the output of the 10th
Find the 1/2 power of the output of circuit 20, and from this, the previous equation
Part of the denominator of (1) is obtained. The 12th circuit 22 calculates ∫R(x) ² dx (this only needs to be calculated once since R(x) is fixed), and the 13th circuit 2
3, the product of the output of the third circuit and the output of the fourth circuit 14 is calculated, and in the 14th circuit 24, the product of the output of the 12th circuit 22 is calculated.
Find the difference between the output of and the output of the thirteenth circuit 23,
The 15th circuit 25 outputs 1/2 of the output of the 14th circuit 24.
The remainder of the denominator in the previous equation (1) is obtained by calculating the power. The 16th circuit 26 divides the output of the 6th circuit 16 (numerator of equation 1) by the output of the 15th circuit 25, and the 17th circuit 27 divides the numerator of the 16th circuit 16 by the output of the 15th circuit 25.
By dividing the output of the 11th circuit 21 by the output of the eleventh circuit 21, the calculation of equation (1) is completed and the correlation coefficient f(x) is obtained.

Note that the calculation formula (1) is not obtained from the calculation order using the circuit shown in Figure 2 as described above.
Any formula may be used to calculate the correlation coefficient. For example, a ∫P(x)・R(x)dx−1/K∫P(x)dx∫P(x)dx b {∫P(x)・R(x)dx−1/K∫P(x )dx∫R(x)dx However, if the value inside { } is negative.

c {∫P(x)・R(x)dx−1/K∫P(x)dx∫R(x)d
x} ² / [∫P(x) ² dx−1/K{∫P(x)dx} ² ] [∫P(x) ² d
x-1/K {∫P(x)dx ² ] However, if the numerator inside { } is negative.

d 〓P(x)R(x)-1/K〓P(x)〓R(x) e {〓P(x)・R(x)−1/K〓p(x)〓R(x)} However, if the value inside { } is negative, it is set to zero.

f {〓R(x)・R(x)−1/K〓P(x)〓R(x)}
² / [〓P(x) ² -1/K{〓P(x)} ² ] [〓R(x) ² -1
/K {〓P(x) ² ] However, if the numerator inside { } is negative, it is set to zero.

It can be transformed into

In addition, when reducing input data, compression is performed using various transformations such as Fourier transform, Hadamard transform, template processing (Laplacian, gradient, and other filters), data compression, filtering, etc., and the comparison between this and the reference pattern is performed. It is also possible to obtain a correlation coefficient and use this to detect a pattern in the input data.

As described above, in the pattern detection device of the present invention, since the input data is normalized, pattern detection can be performed with sufficient margin for changes in the signal amplitude corresponding to the brightness of the pattern image and the DC component corresponding to the brightness. Processing is possible. In addition, there is no need to perform binarization, and if noise affects the correlation coefficient, it is the noise area/detection area, and in principle there is some margin for noise, so pattern detection rate can be improved and stabilized. It is possible. Furthermore, since pattern detection is possible for analog signal waveforms, it is also applicable to noise recognition and printed character recognition. Further, by reducing input data or devising a calculation method, it is possible to reduce the number of calculations and shorten the pattern detection processing time.

FIG. 3 shows an example of a circuit on which the present invention is based. That is, a part of the two-dimensional input data 30 having M pieces of data in the x direction and N pieces of data in the y direction, for example, data P(x+0,
y+0) to point P(x+m, y+0), and from point (x+0, y+1) to point P(x+
m, y+1), and similarly from point P(x+0, y+i) to point P(x+m, y+i)
An m×n data group consisting of m data each up to (i=3,...n) = (x, y) is
The partial input data is extracted by the input data extraction processing circuit 31 and becomes one input of the correlation coefficient calculation circuit 32. And the above partial input data〓
Partial reference data corresponding to the range of (x, y)
(x, y) is stored in the partial reference data storage circuit 33, and this stored data is read out and becomes the other input of the calculation circuit 32. The calculation circuit 32 performs calculations based on the following equation (2), and calculates the point P (X, Y)
Obtain the correlation coefficient O (X, Y).

However, A= _n 〓 ^x=1 _o 〓 ^y=1 {〓(x, y)・〓(x, y)} B= _n 〓 ^x=1 _o 〓 ^y=1 {〓(x, y)} C= _n 〓 ^x=1 _o 〓 ^y=1 {〓(z, y)} D= _n 〓 ^x=1 _o 〓 ^y=1 {〓(x, y)} ² E= _n 〓 ^x=1 _o 〓 ^{y= 1} {〓(x, y)} ^2The calculation circuit 32 calculates the coordinates of the point P(x, y) as (1XM-m+1), (1YN-n+
1) Obtain the result of (M-m+1)(N-n+1) by calculating over the range of.
In this case, the coefficient max(X ₁ , Y ₁ ) is the maximum value
The coordinates (X ₁ , Y ₁ ) when indicating max (X ₁ , Y ₁ ) are the coordinates of the detected pattern. If the following condition is satisfied: max (X ₁ , Y ₁ ) < DC, where DC is a separately specified constant (for example, 0.5, 0.7), it is determined that the pattern cannot be detected or that no detected pattern exists. The maximum value max(X ₁ , Y ₁ ) of such coefficients (X, Y)
The detection and the comparison and determination between the maximum value max (X ₁ , Y ₁ ) and DC are performed by the determination circuit 34 at the subsequent stage of the calculation circuit 32 .

FIG. 4 shows an embodiment of the present invention,
The same parts as in FIG. 3 are given the same reference numerals, and the explanation thereof will be omitted, and the different parts will be explained below. The M×N data group of the two-dimensional input data 30 is reduced by the data reduction circuit 41 so that it becomes 1/I in the x direction, 1/J in the y direction, and 1/IJ as a whole. For example, if I=4 and J=4, 16 points P(k, l) in the two-dimensional input data 30 as shown in FIG.
Each data of P (k, l) is reduced to one piece of data Q (k, l).
(1, 1) is as shown in the following equation, k=1~I, l=
The average value of each point data of 1 to I is calculated.

Q(1, 1)=1/I×J _I 〓 ^k=1 _J 〓 ^l=1 P(k, l) Note that among the reduced data group shown in
The reduced data Q(k, l) of (k, l) is Q(α, β)=1/I×J _I 〓 ^k=1 _J 〓 ^l=1 P{I(α−1)+k, j( β-1)+l)

A part of the reduced data 42 obtained in this way is cut out by the reduced partial data cutting processing circuit 43 and becomes one input of the partial data correlation coefficient calculation circuit 44, and is input to the reduced partial reference data storage circuit 45.
The reduced portion reference data read out from the calculation circuit 44 becomes the other input of the calculation circuit 44. This calculation circuit 44
performs calculations similar to those described above with reference to FIG. 3 to generate a correlation coefficient matrix. This calculation result is led to the determination circuit 46 to determine the approximate position of the pattern coordinates of the input data, so the input data cutout processing circuit 31 is controlled according to this calculation result, and the pattern coordinates are centered on the input data 30. Partial data within a predetermined range are extracted and processed. It has been experimentally confirmed that the position of the detection pattern can be detected by such processing.

Therefore, according to the above embodiment, the calculation circuit 4
Since the number of calculations in steps 4 and 32 is significantly reduced compared to the circuit configuration shown in FIG. 3, high-speed detection processing is possible. That is, the details are as follows. (m×n) of input data (M×N)
In the case of Figure 3, in which all combinations of correlation coefficients with detection standard data are obtained by cutting out a data group, the number of times the product is calculated is m x n x (M - m + 1) (N - n
+1) The square calculation circuit is m×n×(M-m+1)(N-
n+1) The number of calculations for the sum is 2×m×n×(M-m+1)(N
-n+1).

On the other hand, the input data (M×N) is reduced to cut out partial data, the correlation coefficient is calculated between this and the reduced partial reference data, and based on this calculation result, several parts of the input data are cut out. In the case of this example in which the correlation coefficient is calculated between the data and the partial reference data, the number of calculations for the product and square when reduced is m/I x n/J x (M-m+1/I) (N -m+1/J), and the number of calculations for the sum in the case of reduction is 2×m/I×n/J×(M-m+1/I) (N-m+1/J).In the end, in the circuit shown in Figure 3, Compared to the number of calculations, it is reduced to 1/I ² J ² . However, the number of calculations in this case is to calculate only the first term A in the numerator, B in the second term numerator, and D in the denominator of the previous formula (2), and calculate the other terms C and E.
is calculated when the reference data (pattern) is stored (registered) and the result is used, and this calculation is not included in the number of calculations.

Note that it is also possible to control the part of the input data corresponding to the reduced partial data for which the correlation coefficient calculation result for the reduced partial data has the maximum value to be the starting position of the calculation of the correlation coefficient with the reference data. good.
Alternatively, calculate the center of gravity of the values around the maximum value of the above calculation results, and set the part of the input data corresponding to the reduced partial data whose calculation result is this center of gravity value as the starting position for calculating the correlation coefficient with the reference data. may be controlled.

In order to further reduce the number of calculations, as shown in FIG. Of the new calculation target range (the part surrounded by the dotted line) that has been moved by ΔX and ±ΔY, a new calculation can be made by the amount of the movement, and the new calculation can be added to or subtracted from the previous calculation term. In other words, the x-direction coordinate of the previous calculation range is x ₁
~x ₂ , and the y-direction coordinates are represented by y ₂ ~ _{y 1} , the calculation formula for a new calculation range moved by +ΔX from the previous calculation range A, as shown in Figure 7a, is ∫x ₂ +ΔX x ₁ +ΔX ∫y ₁ y ₂ P(x)dxdy・FG＋H However, F=∫x ₂ x ₁ ∫y ₁ y ₂ P(x)dxdy (previous calculation term) As shown in Figure 7b, the previous calculation range is
New calculation range moved from A′ by, for example, +ΔY
The calculation formula for B′ is x ₂ x ₁ ∫y ₁ +ΔY y ₂ +ΔYP(x)dxdy=F−G′+H′ However, G′=∫x ₂ x ₁ ∫y ₂ +ΔY y ₂ P(x)dxdy H′ =∫x ₂ x ₁ ∫y ₁ +ΔY y ₁ P(x)dxdy.

As described above, the present invention is a pattern detection device that can stably and at a high rate detect the same or similar pattern with respect to fluctuations in pattern imaging output or analog waveforms, and can shorten the pattern detection time. can be provided.

[Brief explanation of the drawing]

Fig. 1 is a diagram for explaining the basic configuration, Fig. 2 is a block diagram showing Fig. 1 in detail, Fig. 3 is a block diagram showing an example of the circuit that is the premise of the present invention, and Fig. 4 is a diagram showing the present invention. 5 and 6 are diagrams shown to explain the operation of FIG. 4, and FIGS. 7a and 7b are diagrams shown to explain the operation of the second embodiment. It is. 10, 22, 32, 44... Correlation coefficient calculation circuit, 34, 46... Judgment circuit, 31, 43... Data extraction processing circuit, 41... Data reduction circuit,
33, 45...Reference data storage circuit.

Claims (1)

[Claims] 1. A first cutting circuit that cuts out a part of two-dimensional pattern input data having shading, and partial data P(x) cut out by the first cutting circuit and prepared in advance. Partial reference data R(x)
and a correlation coefficient f(x) having a value in the range of +1 to 0 to -1. However, K: a first correlation coefficient calculation circuit that calculates an integral range, and a first determination circuit that determines input data whose pattern most closely matches the reference data from the output of this first correlation coefficient calculation circuit. a data reduction circuit that calculates the average value of the input data and reduces the input data; and a second extraction circuit that cuts out a part of the data reduced by the data reduction circuit and outputs the reduced partial data. and a second correlation coefficient calculation circuit that calculates the correlation coefficient between the reduced partial data extracted by the second extraction circuit and the reduced reference data prepared in advance according to the above equation; A pattern detection device comprising: a second determination circuit that controls the extraction position of the first extraction circuit according to the calculation result of the second correlation coefficient calculation circuit. 2. The first correlation coefficient calculation circuit sets a portion of the input data corresponding to the reduced partial data for which the calculation result of the second correlation coefficient calculation circuit has a maximum value as a correlation coefficient calculation start position. A pattern detection device according to claim 1, characterized in that: 3. The first correlation coefficient calculation circuit calculates the centroid of values around the maximum value of the calculation result of the second correlation coefficient calculation circuit, and the calculation result corresponds to the reduced partial data having this centroid value. 2. The pattern detection device according to claim 1, wherein a part of the input data is used as a correlation coefficient calculation start position.