JPH09274661A - Method and device for recognizing article - Google Patents

Abstract

PROBLEM TO BE SOLVED: To recognize articles even when these articles touch each other by recognizing the directions of the articles from the rotation angle of a gradation register pattern having the highest correlation value. SOLUTION: A data storage part 35 previously registers the rotation angle of a picture pattern (A or B) in a reference direction for each gradation register pattern registered in a pattern image storage part 34 and the relative position of the picture pattern to the centroid position of a bottle. An arithmetic part 36 performs a search so as to collate the real gradation pattern inside the image of a bottle 1A (or 1B) recorded in an image memory 33 with the plural gradation register patterns previously registered in the pattern image storage part 34 and the data storage part 35 and finds the gradation register pattern having the highest value most correlated to the real gradation pattern. Then, based on the gradation register pattern having the highest correlation value, the arithmetic part 36 recognizes the kind of that bottle, recognizes the direction from its rotation angle and recognizes the centroid position from an offset value.

Description

Detailed Description of the Invention

[0001]

TECHNICAL FIELD The present invention relates to a method and apparatus for recognizing articles such as bottles.

[0002]

2. Description of the Related Art Conventionally, as an article recognition method for recognizing the position and direction of the center of gravity of an article such as a bottle, as described in Japanese Patent Laid-Open No. 4-294989, geometric features of an article contour shape (binarized image) ( There is one that recognizes the position and direction of the center of gravity of the article from the center of gravity, principal axis of inertia, etc. (FIGS. 9A to 9C).

[0003]

However, the prior art has the following problems. When the articles are in contact with each other, the binarized image of the article picked up by the image pickup means is different from the one showing the contour of a single article, and the article cannot be recognized (Fig. 9 (A)-
(C)).

Therefore, the article in contact is overlooked without being recognized as the article, and the recognition efficiency is lowered.

Further, in order to avoid the above-mentioned unrecognizable contact due to the contact of the article, it is necessary to provide an article separating mechanism for releasing the contact of the article in the previous step. Alternatively, it is also necessary to provide an article recovery conveyor for recovering the article missed due to the unrecognizable state due to the contact of the article and again providing the article for recognition.

Since the article is recognized based on the contour shape of the article, it is not possible to recognize the front and back of the article or the kind of article between articles having the same contour.

An object of the present invention is to enable recognition of articles even when the articles are in contact with each other.

Another object of the present invention is to make it possible to recognize the front and back of an article and the type of article.

[0009]

According to a first aspect of the present invention, in a method for recognizing an orientation of an article, a specific pattern existing inside the contour of the article is rotated by a predetermined angle with respect to a reference direction. A plurality of different shade registration patterns are registered in advance together with their rotation angles, the image of the article is captured by the image pickup means, and the shade actual pattern inside the image of the article is stored.
By collating with the plurality of shade registration patterns registered in advance, the shade registration pattern having the highest correlation value with the shade actual pattern is obtained, and the orientation of the article is recognized from the rotation angle of the shade registration pattern having the highest correlation value. It was done like this.

The present invention according to claim 2 is the same as the invention according to claim 1, further characterized in that the pattern defining the shade registration pattern is different for each kind of article, and the shade having the highest correlation value. The product type is also recognized from the registered pattern.

According to a third aspect of the present invention, in addition to the first or second aspect of the present invention, each of the light and shade registration patterns has a rotation angle with respect to a reference direction and a correlation position with respect to an article gravity center position registered. The position of the center of gravity of the article is also recognized from the grayscale registered pattern having the highest correlation value.

The present invention described in claim 4 is the first to third aspects of the present invention.
In the present invention according to any one of, further, the article recognition information obtained from the grayscale registration pattern with the highest correlation value,
It is adapted to be transferred to an article handling device in a later step.

According to a fifth aspect of the present invention, in an article recognition device for recognizing an orientation of an article, a plurality of shade registration patterns obtained by rotating a specific pattern existing inside the contour of the article by a constant angle with respect to a reference direction. A storage unit for pre-registering the rotation angle and the image, an image memory for recording the image of the article captured by the image pickup unit, the recording data of the image memory and the registration data of the storage section, and the gray scale inside the image of the article is obtained. The pattern and the plurality of previously registered shade registration patterns are collated, the shade registration pattern having the highest correlation value with the shade actual pattern is obtained, and the orientation of the article is determined from the rotation angle of the shade registration pattern having the highest correlation value. And an arithmetic unit for recognizing.

According to a sixth aspect of the present invention, in addition to the fifth aspect of the present invention, in the storage unit, the pattern defining the shade registration pattern is different for each product type, and the calculation is performed. The section recognizes the product type from the grayscale registration pattern having the highest correlation value.

According to a seventh aspect of the present invention, in addition to the sixth aspect of the present invention, each of the density registration patterns in the storage unit also registers a rotation angle with respect to a reference direction and a correlation position with respect to an article center of gravity position. The arithmetic unit also recognizes the position of the center of gravity of the article from the grayscale registered pattern having the highest correlation value.

The present invention according to claim 8 provides claims 5 to 7.
In any one of the present inventions, the arithmetic unit transfers the article recognition information obtained from the grayscale registered pattern having the highest correlation value to an article handling device in a subsequent step.

The present invention has the following effects. Since the article is recognized by the light and shade pattern of the pattern inside the contour of the article, even when the articles are in contact with each other, the orientation and the position of the center of gravity of the article can be recognized.

Therefore, the contacting item is not overlooked by recognizing that it is the item, and the recognition efficiency is improved.

Further, since there is no recognizability due to contact of the articles, it is not necessary to provide an article separating mechanism or an article collecting conveyor.

Since the article is recognized based on the correlation value of the light and shade pattern of the pattern inside the contour of the article, it is possible to recognize the front and back of the article and the product type between the articles having the same contour.

[0021]

1 is a schematic diagram showing a first embodiment, FIG. 2 is a schematic diagram showing a handling device, FIG. 3 is a schematic diagram showing the operating principle of the handling device, and FIG. 4 is an article recognition of the present invention. FIG. 5 is a schematic diagram showing the principle, FIG. 5 is a schematic diagram showing a grayscale registration pattern in the present invention, FIG. 6 is a flow chart showing an article recognition procedure of the present invention, and FIG. 7 is a schematic diagram showing a search image of the grayscale registration pattern and a processed image of the article. FIG. 8 is a schematic diagram showing the second embodiment, and FIG. 9 is a schematic diagram showing a conventional method.

(First Embodiment) (FIGS. 1 to 7) The article handling apparatus 10 has a supply conveyor 11 and a discharge conveyor 12 as shown in FIG. Then, the article handling apparatus 10 uses the handling apparatus 20 to supply the bottles 1A of the type A and the bottles 1B of the type B that are supplied in a mixed state by the supply conveyor 11 to the center 4 of the discharge conveyor 12 for the (a) bottle 1A. It is supplied to the discharge parts 12A to 12D of the bottle with the bottom of the bottle oriented in the discharge direction.
For B, one discharge unit 1 on each side of the discharge conveyor 12
It is supplied to 2E and 12F with the bottom of the bottle oriented in the discharge direction.

At this time, the article handling apparatus 10 has the article recognition apparatus 30 in the preceding step of the handling apparatus 20, and the bottles 1A, 1B supplied by the supply conveyor 11.
The orientation and the position of the center of gravity are recognized, and this recognition information is transferred to the handling device 20 in the subsequent process. Handling device 20
Obtains the recognition information of the article recognition device 30 and sucks the center of gravity of each bottle 1A or each bottle 1B with the hand 23 described later, and moves the hand 23 so that the bottle bottoms of the bottles 1A and 1B face the discharge direction. It turns and supplies the bottle 1A to the discharge units 12A to 12D of the discharge conveyor 12 and the bottle 1B to the discharge units 12E and 12F of the discharge conveyor 12.

The handling device 20 and the article recognition device 30 will be described below. (Handling Device 20) (FIGS. 1 to 3) As shown in FIG.
One or two or more linear motors 22 are slidably provided on the lower surface of 1, and a hand 23 is provided on the linear motor 22.

The linear motor 22 is based on the Sawyer principle and has a structure called a surface motor. The Sawyer's principle means that movement control in both XY axis directions along the platen 21 is performed by adjusting the interaction of magnetic forces. FIG. 2 and FIG.
It shows the principle. Linear motor 22 is 1
The permanent magnet (PM) and two electromagnets EMA and EMB are provided with a forcer 24.
4 can move on the lower surface of the platen 21 on the toothed plate-shaped iron core 25 having high magnetic permeability. Electromagnet EMA
Is supplied with a sinusoidal current, and an electromagnet EMB is supplied with a cosine-shaped current. Now in FIG. 3 (A),
When a cosine wave current is applied only to the electromagnet EMB, the magnetic force is doubled at the pole 3 of the electromagnet EMB and the magnetic forces at the pole 4 cancel each other out, and the forcer 24 stops at the position of FIG. 3 (A). At this time, the magnetic force of the electromagnet EMA is balanced. Next, as shown in FIG. 3 (B), when a sinusoidal current is applied only to the electromagnet EMA, the magnetic forces cancel each other out at the EMA pole 1 of the electromagnet, and are doubled at the pole 2, resulting in half the tooth width, that is, 1/4. The forcer 24 moves to the left in the figure by the pitch. As described above, by alternately passing the respective currents to the electromagnet EMB in FIG. 3C and to the electromagnet EMA in FIG. 3D, the forcer 24 sequentially moves by 1/4 pitch. As shown in Fig. 2 (C), the current of sine wave and cosine wave flowing in each electromagnet is actually divided into 40 steps and the resolution is 1/40 (mm) = 25 (μm).
Is raised to. The linear motor 22 of the actual handling device 20 includes two forcers 24A for the X axis and Y axis,
24B, which are arranged in point symmetry with respect to the center of gravity. 2 from the 4 air holes on the bottom of the platen 21
Approximately 10 (μm) is levitated from the surface of the platen 21 (iron core 25) by blowing out air of up to 3 atm to enable high speed movement.

The hand 23 is supported by the linear motor 22 so that the turning angle can be controlled, and also the bottles 1A, 1B.
A suction pad for adsorbing is provided so as to be able to move up and down.

That is, the handling device 20 attracts the bottles 1A and 1B on the supply conveyor 11 to the hand 23 by the movement operation of the linear motor 22, and then turns the hand 23 to adjust the orientation of the bottles 1A and 1B.
By the movement operation of the linear motor 22, those bottles 1A,
1B is transferred to a predetermined position of any of the discharge sections 12A to 12F of the discharge conveyor 12 and supplied.

(Item Recognition Device 30) (FIGS. 1 and 4 to 7) The item recognition device 30 is configured to have a camera 31 (imaging means) and an image processing device 32, as shown in FIG. It
Then, the image processing device 32 includes the image memory 33, the pattern image storage unit 34, the data storage unit 35, and the calculation unit 3.
6.

The camera 31 views the entire width of the supply conveyor 11, and the bottle 1 conveyed on the supply conveyor 11
Capture images of A and 1B.

The image memory 33 records the images of the bottles 1A and 1B taken in from the camera 31.

The pattern image storage unit 34 uses the specific pattern A existing inside the outline of the bottle 1A (or the specific pattern B existing inside the outline of the bottle 1B) with respect to the reference direction (for example, the discharge direction of the discharge conveyor 12). A plurality of shade registration patterns rotated by a fixed angle are registered in advance (FIG. 5). Note that the pattern defining the above-mentioned shade registration pattern registered in the pattern image storage unit 34 is different for each type of bottle 1A, 1B (A for bottle 1A, B for bottle 1B).

The data storage unit 35 stores (a) a pattern (A) for each shade registration pattern registered in the pattern image storage unit 34.
Or the rotation angle of B) with respect to the reference direction (0 degrees, 5 degrees, 10
...), and (b) the relative position (offset values X ₀ , Y ₀ ) formed by the design with respect to the center of gravity of the bottle is registered in advance (FIG. 5). In FIG. 5, the rotation angle i with respect to the reference direction
The X offset value of the grayscale registered pattern is X ₀ (i), Y
The offset value is represented by Y ₀ (i).

The calculation unit 36 stores the actual shade pattern inside the image of the bottle 1A (or 1B) recorded in the image memory 33 and a plurality of shade registration patterns registered in advance in the pattern image storage unit 34 and the data storage unit 35. Is searched so as to match with, and a grayscale registered pattern having the highest correlation value with the grayscale real pattern is obtained. Then, the calculation unit 36 recognizes the product type A (or B) of the bottle 1A (or 1B) based on the grayscale registration pattern having the highest correlation value, and (a) determines the bottle 1A (or 1B) from the rotation angle. ) Is recognized, and (b) the center of gravity of the bottle 1A (or 1B) is recognized from the offset value.

Incidentally, the pattern search position (pattern center of gravity) of the shaded real pattern of the bottle 1A (or 1B) is (X
_p , Y _p ), and the grayscale real pattern is the rotation angle i.
[X offset value X ₀ (i), Y offset value Y ₀
When the density registration pattern of (i)] is matched (maximum correlation value), the center of gravity position (X, Y) of the bottle 1A (or 1B) is obtained by the following equations (1) and (2). X = X _p + X ₀ (i) (1) Y = Y _p + Y ₀ (i) (2)

Further, as shown in FIG. 7, pattern matching for matching the actual shade pattern and the registered shade pattern is performed, as shown in FIG. 7, at each point of the search image of the registered shade pattern of sizes M and N for each of a plurality of shade registered patterns. Brightness P of (m, n)
(M, n) and the brightness Bij (m) of each point (m, n) of the processed image of the grayscale real pattern of sizes M and N at each position (i, j) within the range X, Y captured by the camera 31. , N), the processed image self-dispersion and the search image self-dispersion at the position (i, j) and the mutual dispersion of the processed image and the search image at the position (i, j) are described in (3) to This is performed by obtaining the equation (5), and then obtaining the normalized correlation value f (i, j) of the following equation (6).

[0036]

[Equation 1]

[0037]

[Equation 2]

[0038]

(Equation 3)

[0039]

(Equation 4)

The normalized correlation value f (i, j) takes a value of 0 to 1, and f (i, j) = 1 means a perfect match between the actual shade pattern and the registered shade pattern. In practice, a fixed threshold value f ₀ is set in advance, and the processed image at the position (i, j) where f (i, j) exceeds f ₀ has a predetermined specific pattern (A or B). It is admitted that there is.

The article recognition procedure by the article recognition device 30 will be described below (FIGS. 4 and 6). (1) An image is captured from the camera 31 and recorded in the image memory 33 according to a timing signal corresponding to the constant feeding of the supply conveyor 11 (FIG. 4A).

(2) The grayscale registration pattern (FIG. 4B) previously registered in the pattern image storage unit 34 is searched so as to be compared with the actual grayscale pattern of the processed image recorded in the image memory 33. , A grayscale registered pattern whose normalized correlation value is higher than a threshold value, and its pattern search position (X _p ,
Y _p ) is obtained (FIG. 4 (C)). This search is performed for all of the plurality of shade registration patterns.

(3) The bottle type (A or B) is recognized from the shade registration pattern obtained in (2) above, and the bottle is detected from the rotation angle i defined in the data storage unit 35 in correspondence with this shade registration pattern. The direction of 1A (or 1B) is recognized, and the offset values X ₀ (i) and Y ₀ are also set in the data storage unit 35 corresponding to the grayscale registration pattern.
The center of gravity positions X and Y of the bottle 1A (or 1B) are recognized from (i) by the above equations (1) and (2) (FIG. 4).
(D)).

(4) The recognition information (product type, orientation, center of gravity position) obtained in (3) above is transferred to the handling device 20 in the subsequent step, and the above-mentioned handling operation is applied to the bottle 1A (or 1B), This is supplied to any of the discharge units 12A to 12F of the discharge conveyor 12.

(5) The above (1) to (4) are repeated. Therefore, according to this embodiment, there are the following effects. Since the bottle 1A (or 1B) is recognized by the shade pattern of the pattern inside the outline of the bottle 1A (or 1B), even if the bottles 1A (or 1B) are in contact with each other, the bottles 1A (or 1B) Or 1
The direction of B) and the position of the center of gravity can be recognized.

Therefore, the contacting bottle 1A (or 1B) is not overlooked by recognizing that it is the bottle 1A (or 1B), and the recognition efficiency is improved.

Further, since there is no recognition by the contact of the bottle 1A (or 1B), it is not necessary to provide a bottle separating mechanism or a bottle collecting conveyor.

Since the bottle 1A (or 1B) is recognized by the correlation value of the light and shade pattern of the picture inside the outline of the bottle 1A (or 1B), the front and back of the bottle 1A (or 1B) are recognized, It is also possible to recognize the product type between the bottles 1A (or 1B) having the same contour.

(Second Embodiment) (FIG. 8) The second embodiment differs from the first embodiment only in that a handling device 40 consisting of a robot 41 is used instead of the above handling device 20. . The handling device 40 makes it possible to move the hand 42 of the robot 41 in the XY directions similarly to the handling device 20 described above, makes it possible to control the turning angle, and freely move up and down the suction pad for sucking the bottles 1A and 1B. Prepare As a result, the handling device 40 also obtains the recognition information of the article recognition device 30 and applies the same handling operation as in the first embodiment to the bottles 1A and 1B, similarly to the handling device 20 described above.

The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration of the present invention is not limited to this embodiment, and the design may be changed without departing from the scope of the present invention. The present invention is also included in the present invention.

[0051]

As described above, according to the present invention, even when articles are in contact with each other, the articles can be recognized.

Further, according to the present invention, it is possible to recognize the front and back of an article and the type of article.

[Brief description of drawings]

FIG. 1 is a schematic diagram showing a first embodiment.

FIG. 2 is a schematic diagram showing a handling device.

FIG. 3 is a schematic view showing the operating principle of the handling device.

FIG. 4 is a schematic view showing an article recognition principle of the present invention.

FIG. 5 is a schematic diagram showing a grayscale registration pattern according to the present invention.

FIG. 6 is a flowchart showing an article recognition procedure of the present invention.

FIG. 7 is a schematic diagram showing a search image of a grayscale registered pattern and a processed image of an article.

FIG. 8 is a schematic diagram showing a second embodiment.

FIG. 9 is a schematic view showing a conventional method.

[Explanation of symbols]

 1A, 1B Bottle (article) 20, 40 Handling device 30 Article recognition device 31 Camera (imaging means) 33 Image memory 34 Pattern image storage unit 35 Data storage unit 36 Arithmetic unit

Claims (8)

[Claims] 1. An article recognition method for recognizing the orientation of an article, wherein a plurality of shade registration patterns obtained by rotating a specific pattern existing inside the contour of the article by a constant angle with respect to a reference direction are registered in advance together with the rotation angles thereof. The image of the article is captured by the image pickup means, and the actual grayscale pattern inside the image of the article is collated with the plurality of previously registered grayscale registration patterns, and the grayscale registration having the highest correlation value with the grayscale real pattern is registered. An article recognition method, characterized in that a pattern is obtained and the orientation of the article is recognized from the rotation angle of the shaded pattern having the highest correlation value. 2. A pattern for defining the shade registration pattern,
The article recognition method according to claim 1, wherein the article classification is different for each article type, and the article type is also recognized from the grayscale registration pattern having the highest correlation value. 3. The grayscale registration pattern is registered with the rotation angle with respect to the reference direction as well as the correlation position with respect to the article gravity center position, and the article gravity center position is also recognized from the grayscale registration pattern having the highest correlation value. 2. The article recognition method described in 2. 4. The article recognition method according to claim 1, wherein the article recognition information obtained from the grayscale registration pattern having the highest correlation value is transferred to an article handling device in a subsequent step. 5. An article recognition device for recognizing the orientation of an article, wherein a plurality of shade registration patterns obtained by rotating a specific pattern existing inside the contour of the article by a constant angle with respect to a reference direction are registered in advance together with the rotation angles thereof. A storage unit, an image memory for recording an image of an article captured by the image pickup unit, a record data of the image memory and registration data of the storage unit are obtained, and a shade real pattern inside the image of the article is registered in advance. A plurality of shade registration patterns are collated, a shade registration pattern having the highest correlation value to the shade actual pattern is obtained, and an arithmetic unit for recognizing the orientation of the article from the rotation angle of the shade registration pattern having the highest correlation value is provided. An article recognition device characterized by the following. 6. In the storage unit, the pattern defining the grayscale registration pattern is different for each product type, and the computing unit also recognizes the product type from the grayscale registration pattern with the highest correlation value. The article recognition device according to claim 5. 7. In the storage unit, each of the grayscale registration patterns is also registered with a rotation angle with respect to a reference direction as well as a correlation position with respect to an article gravity center position, and the computing unit calculates from the grayscale registration pattern having the highest correlation value to the article gravity center position. 7. The article recognition device according to claim 6, which also recognizes. 8. The article recognition device according to claim 5, wherein the calculation unit transfers the article recognition information obtained from the grayscale registered pattern having the highest correlation value to an article handling apparatus in a subsequent step.