JPH0338789A - Pattern matching system - Google Patents

Abstract

PURPOSE:To accurately perform the inspection and recognition of a targeted pattern by correcting an image compressed in a horizontal direction at a position near to both ends of a curved surface to a normal image by performing the coordinate transformation of the image, and collating it with a reference pattern. CONSTITUTION:The image of a pattern targeted to be inspected plotted on the curved surface of a cylindrical vessel such as a bottle or a can, etc., is stored in an original image memory 3 via a TV camera 1 and a binarization circuit 2, and a position coordinate for the image pickup picture of the targeted pattern is detected with a position correction detection circuit 5. Also, a coordinate transformation circuit 4 generates an address required for the magnification of image data by applying the coordinate transformation on each part. A coordinate-transformed and magnified image is stored in an image memory 7 after transformation, after that, it is collated with the reference pattern of a reference pattern memory 6 for which position correction is applied at a positioning collation circuit 8 according to the position coordinate detected with the position correction detection circuit 5. Thereby, the accurate inspection and recognition can be performed on an image pattern even when a targeted part is shortened at the end of it.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a pattern matching method for inspecting and recognizing a target pattern extracted by an image sensor by comparing it with a standard pattern. Specifically, a pattern in which a target pattern, such as a label attached to a curved surface of a cylindrical object such as a bottle, is imaged and the target pattern is determined by comparing the binarized image with a standard pattern. This relates to a matching method.

[Conventional technology]

Conventionally, in such a pattern matching method, an image pattern obtained by imaging a target pattern such as a label attached to a curved surface of a cylindrical object such as a bottle is
When viewed from the front, even if the front part has normal dimensions, the image pattern becomes too close from the front to the end. Matching was done.

FIG. 3 is a perspective view showing a label 12 attached to the curved surface of a cylindrical object 11 such as a bottle, and FIG.
This is an image pattern when the label is viewed from the front.

In the image patterns 21 to 25 shown in FIG. 4, the letter "C" on the front is close to the normal size, but the letter r A on the edge is
'' and rB, it can be seen that they are packed together compared to the letter ``C'' on the front, and have a shape with a compressed width.

[Problem to be solved by the invention]

In the conventional pattern matching method described above, when the target is a label attached to the curved surface of a cylindrical object such as a bottle, the image pattern obtained by imaging the object is Due to this clogging, there was a problem in that accurate inspection and recognition could not be performed even if comparison was made with a standard pattern.

An object of the present invention is to solve the problems of the prior art and to provide a pattern matching method that enables accurate inspection and recognition even when the target is an image pattern that is clogged at the edges. be.

[Means to solve the problem]

In order to achieve the above object, the present invention provides a pattern for determining the target pattern by capturing an image of the target pattern drawn on a cylindrical curved surface and comparing the binarized image with a standard pattern registered in advance. In the matching method, a first memory stores an image pattern obtained by capturing an image of the target pattern drawn on a cylindrical curved surface, which becomes more closely aligned from the front to the end; By reading out from the first memory the image pattern that has become too close to fit depending on the degree of curvature of the cylindrical curved surface and converting its coordinates, the image pattern that can be obtained when the above-mentioned curved surface is a plane is read out from the first memory. a coordinate conversion means for converting a normal image pattern in which there would be a clogging into a normal image pattern and writing it into a second memory, and comparing the normal image pattern read from the second memory with the standard pattern to determine it. A verification means is provided.

[Function] According to the present invention, the coordinate conversion means transforms a clogged image pattern into a normal image pattern in which the clog is removed, and this is compared and verified with a standard pattern, so that accurate inspection and recognition are possible. This makes it possible.

〔Example〕

As already mentioned, the present invention is based on an image pattern that is obtained by imaging a target pattern drawn on a cylindrical curved surface, and that becomes closer from the front to the end.
By converting the coordinates of the image pattern (which is exactly the same size according to the degree of curvature of the cylindrical curved surface), we can obtain the image pattern that would be obtained if the curved surface was a flat surface.
Since the key point is to convert the pattern into a normal image pattern that eliminates the clogging, the principle of this conversion operation will be explained first.

FIG. 5 is an explanatory diagram showing such a clogged image pattern and a normal image pattern.

In the figure, numeral 11 is a cylindrical object (for example, a bottle or a cup), which is shown as viewed from above. 12 is a label attached to the curved surface of the cylindrical object 11. The label 12 includes A, B.

Character patterns 21 to 25 named C, D, and E are printed.

Letters A and B printed on the label 12 attached to the curved surface
, C, D, E (21 to 25) are imaged with the character C in front.
It is 3. The normal image pattern obtained when the label 12 is peeled off from the curved surface, flattened, and imaged from the front is 41.
It is. The above-mentioned conversion operation corresponds to enlarging this partially clogged image pattern 13 to a normal image pattern 41.

This enlargement is compared to the letters A and E on the edges compared to the letters C on the front.
Because the edges are so clogged, a normal image pattern cannot be obtained for the characters at the edges unless the magnification is increased. In other words, it is necessary to change the magnification rate depending on the location of the label. Below, we will explain how to change it.

On the upper surface of the cylindrical object 11, viewed from the center point O, there are n angles that include both ends of the label 12, and each angle is defined as θ1. The length of the Rahel part with respect to the angle θ1 is uniformly l, but when viewed from the front, the end shrinks somewhat. This means that the dimensions correspond to the .

Fig. 6 is an explanatory diagram showing how to quantitatively determine the enlargement ratio when increasing the enlargement ratio at the edge of the label in this way, and Fig. 6A is an explanatory diagram showing how the enlargement ratio should be determined quantitatively. It is an enlarged view of the main part M of.

Please refer to FIG. 6 and FIG. 6A. 14 is the X coordinate axis on the clogged image, and 42 is the X coordinate axis on the normal image. Each pattern 2 drawn on the curved surface label 12 of the cylindrical object 11
1 to 25, the image data 13 is divided into n parts corresponding to each part divided into n parts by a sufficiently small circular arc l (center angle θl). Each part 31 to 46 of the image data 13 divided into n
width l.

is found as follows.

Considering the arc l with a sufficiently small central angle θ1 to be direct, the width j2m' of the m-th divided portion laterally from the center point C seen from the imaging direction of the cylinder is 42m' = fficos θm
--(1). From Figure 6, θm and i are j2=rθ1 (r is the radius of the cylinder)...
(IA) θm−π−(θ2+θ3) ・
...<2) θ2. θ3 is θ2 = 1/2 (π-θ1〉...
...(3) θ3−π−(π/2+(m−1)θ1)
......(4) In equation (2) above, (3), (
4) Substituting the formula, θm − (m-1/2) θ1
・・・・・・(5) In the above formula (1>), (IAL
Substituting equation (5), 12m”=rθ1 cos (m
1/2) θ1 = (6). Each divided portion nm" is determined in advance, and the image data 13 is divided as shown in 31 to 46 in FIG. 5.

Next, each divided portion is summarized and coordinates are transformed using the coefficients determined for each divided portion. The coefficient is determined as follows.

The captured image data has a width of 1 m.
'Since it is reduced to the width, each part is f/f'
Multiplying the image returns it to the original normal image. From Fig. 6, the coefficient km of the mth part to the right from the center C is km=l/l'...
...(7) Substituting the above (IA) and equation (6) into the above (7) 2, km = 1/cos (m - 1/2) θ
1...(8).

From the above, you should understand how much the magnification rate (coefficient) should be.

FIG. 1 is a block diagram showing one embodiment of the present invention. In the figure, 1 is a TV camera, 2 is a binarization circuit, and 3 is a TV camera.
is an original image memory, 4 is a coordinate conversion circuit, 5 is a position correction detection circuit, 6 is a standard pattern memory, 7 is an image memory for storing image data after coordinate conversion, 8 is an alignment matching circuit, 9 is a CPU, 10 is a lotus.

The basic operation of the embodiment shown in FIG. 1 is as follows.

First, the standard pattern must be loaded into the standard pattern memory 6, so correct images 51 to 55 as shown in FIG. 3
This is then read out and stored in the standard pattern memory 6 as a standard pattern.

Next, the image of the pattern to be inspected drawn on the curved surface of a cylindrical container such as a bottle or can is stored in the original image memory 3 through the TV camera 1 and binarization circuit 2. From the stored image, the position correction detection circuit 5 detects the position coordinates of the target pattern with respect to the imaging screen using a normal method. Further, as described above, the coordinate transformation circuit 4 enlarges the image data stored in the original image memory 3 and divided into n parts by coordinate transformation, using a predetermined coefficient (enlargement ratio). Generate the necessary address 0 for. The coordinate-converted and enlarged image is stored in the converted image memory 7, and then stored in the standard pattern memory 6 in which the position is corrected in accordance with the position coordinates detected by the position correction detection circuit 5 in the alignment matching circuit 8. Check against the standard pattern.

The coordinate transformation by the coordinate transformation circuit 4 is performed by setting the coefficient km and the constant Xmo (see FIG. 6) described above to the coordinate transformation circuit 4 for each divided image portion by the CPU 9, so that the coordinate xm before transformation is The converted coordinates Xm are generated using the following equation (9), and each divided image portion is enlarged by krr1 times. The converted image data is stored in the converted image memory 7.
is stored in

Xm −km−xm×Xmo --
(9) From Fig. 6, the above equation (9) is expressed as: Xm = km (xm - xmo) + (xc + (m - 1)
42)-km-xm-km-xm, +(xc+(m-1
)f)...(10) Therefore, Xmo for the right image part is Xm
,=-km-xmo+xc+(m-LL+!-
(11) xc...Center coordinates mi! =km°
xm-kn+° xmo+xc-ml, and Xmo becomes Xm, =-km-xm, +xc-ml.

The thus converted image in the image memory 7 is aligned with the standard patterns 51 to 55 in FIG. 2 stored in the standard pattern memory 6 in the alignment matching circuit 8.
They are collated, inspected, and recognized. At this time, the comparison is performed after the position of the standard pattern is corrected in accordance with the position of the target pattern on the imaging screen that has already been detected by the position correction detection circuit 5.

2 [Effects of the Invention] According to the present invention, in an apparatus for inspecting and recognizing a pattern (for example, characters on a label) drawn on a curved surface of a cylindrical object such as a bottle or a can, a pattern drawn in a horizontal direction near both ends of the curved surface can be detected. There is an advantage that accurate inspection and recognition of the target pattern is possible by correcting the compressed image into a regular image by image coordinate transformation and comparing it with a standard pattern.

[Brief explanation of drawings]

Fig. 1 is a block diagram showing an embodiment of the present invention, Fig. 2 is an explanatory drawing showing a standard pattern, and Fig. 3 is a perspective view showing a label affixed to the curved surface of a cylindrical object such as a bottle. figure,
Fig. 4 is an explanatory diagram showing the image pattern when the label is viewed from the front, Fig. 5 is an explanatory diagram showing the clogged image pattern and the normal image pattern, and Fig. 6 is an explanatory diagram showing the image pattern when the label is viewed from the front. FIG. 6A is an explanatory diagram showing how to determine the magnification ratio when obtaining a normal image, and FIG. 6A is an enlarged view of the main part M;
It is. Explanation of symbols 3 ■...TV camera, 2...Binarization circuit, 3...Original image memory, 4...Coordinate conversion circuit, 5...
Position correction detection circuit, 6... Standard pattern memory, 7.
... Image memory after conversion, 8... Registration verification circuit,
9...CPU, 10...Bus

Claims (1)

[Claims] 1) In a pattern matching method that determines the target pattern by capturing an image of a target pattern drawn on a cylindrical curved surface and comparing the binarized image with a standard pattern registered in advance, a first memory for storing a closely packed image pattern obtained by imaging the target pattern drawn on the surface, which becomes closer from the front to the end; Accordingly, by reading out the image pattern that has become too crowded from the first memory and converting its coordinates, the image pattern that has become too crowded that would be obtained when the curved surface is a plane is resolved. a coordinate conversion means for converting the normal image pattern into a normal image pattern and writing it into a second memory; and a matching means for comparing and determining the normal image pattern read from the second memory with the standard pattern. Features a pattern matching method.