JP2896043B2 - Image recognition device and component supply device using the same - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for recognizing a posture of a recognition object, and a component for taking out a component having the same posture from a group of many parts distributed in various postures and sequentially supplying the same to another device. It is applied to a supply device and the like.

[0002]

2. Description of the Related Art Various methods for detecting the attitude of an object by optical means have been proposed. In Japanese Patent Application Laid-Open No. 4-141325, as an example, a predetermined position of a parts transport track of a parts feeder is formed of a light-transmitting plate, and an illuminating device is provided above and below, and an image is detected at one of the upper and lower parts. A component supply device for disposing the respective means is disclosed. This device recognizes the posture and position of a part by taking in a silhouette image of the part conveyed on a light-transmitting plate, or a threshold for binarization of a part whose posture cannot be specified by a silhouette image. The values are recognized in correspondence with the reflected light image, and the attitude of various components is recognized.

[0003]

In such a component supply apparatus, the binarized area of an image sometimes does not match the actual one due to the influence of fluctuations and aging of illumination light, scattered light incident from the outside, and the like. In some cases, parts with different postures were supplied. As an example of a method for coping with this, there is a method described in Japanese Patent Application Laid-Open No. 4-141384. That is, the calibration jig is positioned on the light-transmitting plate, a hole with a known area drilled in the jig is recognized, and a threshold value is set so that the binarized area of the hole becomes equal to the known area. This is a method in which the binarized area of the image does not significantly deviate from the actual body even if the illumination light changes with time. The present invention provides a component supply device that does not use such a special jig and that responds to fluctuations in lighting conditions in a short time.

[0004]

According to the present invention, an illuminating means for irradiating an object to be recognized with light, an imaging means for imaging a reflected light image from the object, and a sample having the same shape as the object are placed in a specific posture. Is provided with a sample holding unit. And
The feature detection means for binarizing the reflected light image captured by the imaging means and detecting the area of the image, and the threshold value for performing the binarization, are set in advance by the binarized area of the reflected light image from the sample. Threshold value setting means for setting the area to be equal to the prescribed area value; and a threshold value for causing the image pickup means to take in the reflected light image of the sample at a predetermined cycle and to cause the threshold value setting means to reset the threshold value. Value updating means, and determining whether or not the area of the reflected light image of the object detected by the feature detecting means is within an allowable tolerance of the specified area value, whereby the attitude of the object is the same as the sample. Recognize whether or not.

According to the present invention, a plurality of components are placed,
A recognition table having a sample area for mounting a sample having the same shape as a part in a specific posture on a part of itself, a transfer unit for taking out a part on the recognition table at a predetermined position, and an illuminating unit for irradiating light to the recognition table And an image pickup means for picking up a reflected light image therefrom. Then, the reflected light image captured by the imaging unit is binarized, the individual component images in the imaging screen are distinguished from each other, and a feature detecting unit that detects the area and the direction of each component image, and a threshold for performing the binarization. Threshold value setting means for setting the value so that the binarized area of the reflected light image of the sample is equal to a predetermined area value, and taking the reflected light image of the sample into the imaging means at a predetermined cycle, The threshold setting means is provided with a threshold updating means for resetting the threshold, and among the component images detected by the feature detecting means, those having an area within an allowable tolerance of a specified area value are selected, The transfer means picks up the component in accordance with the position and orientation of the corresponding component, and transfers the component in a predetermined position.

[0006]

An image of a component sample held in a specific posture is captured from a predetermined direction at a fixed cycle, and the image area of the captured sample matches a specified area to be taken by the component image in that posture. , The threshold value for binarization is updated every period. Then, the component image is recognized with the updated threshold value, and only the component having the same posture as the sample is extracted to a predetermined position.

[0007]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the image recognition apparatus of the present invention and a component supply apparatus using the same will be described. FIG. 1 is an explanatory diagram showing a main configuration of a component supply device according to the present embodiment. The component supply device 1 includes a recognition table 2, a robot 3, an illumination unit 4, an imaging unit 5, an image recognition unit 6, and a control unit 7. It is configured.

The recognition table 2 has a rectangular cylindrical frame 21.
And a protrusion 22 provided at the same height on the inner surface of the frame 21.
And is attached to the robot base 31 via a swinging mechanism (not shown). The light-transmitting plate 23 has the same area as the horizontal cross section of the inner space of the frame 21, is fixed in an inner lid shape slightly below the upper end surface of the frame 21, and has a component P of the same shape on its upper surface. It is randomly distributed from a hopper (not shown) or the like. FIG. 3 shows this state in a plan view. At one corner on the upper surface of the light transmitting plate 23, a partition plate 231 having a light transmitting property and having an L-shaped cross section is provided.
And a space surrounded by this and the inner wall surface of the frame 21 is a sample area 230.

The illuminating means 4 is supported on the left and right inner surfaces as shown in FIG. 1 at a position further lower than the projection 22 of the frame 21, and irradiates light from below to uniformly illuminate the light transmitting plate 23. ing. Reference numeral 41 denotes a reflection plate for reflecting the light of the illumination means 4 upward, and at the same time, has a function of blocking light downward.
The imaging means 5 is held upward just below the light transmitting plate 23,
The optical axis passes through the center of the light transmitting plate 23 vertically, and is arranged at a focal length such that the entire light transmitting plate can be taken into the field of view.

The robot 3 is a horizontal articulated robot, and takes out a part P by mounting a chuck 31 as an end effector and supplies it to another device (not shown), or assembles it directly to a work positioned on a conveyor. . The chuck 31 is moved to an arbitrary position on the transparent plate of the recognition table 2 by the turning operation of the robot arm,
Given the vertical movement and the θ rotation about the vertical movement axis, the part P is lifted with an arbitrary θ angle.

The image recognizing unit 6 includes a feature detecting unit 61, a threshold setting unit 62, and a threshold updating unit 63. The characteristic detecting means 61 is provided for the illumination means 4 captured by the imaging means 5.
Is binarized, the images of the individual components P in the imaging screen are distinguished (labeled), and the area and direction of each component image are detected (feature extraction). Since the labeling and the feature extraction are not the gist of the present invention, detailed description will be omitted. The threshold value setting means 62 includes the feature detecting means 6
A threshold value for performing binarization in 1 is set in the reflected light image of the component P (hereinafter, referred to as a component Ps) placed in a specific posture in the sample area 230 described above. The threshold value is set so that the quantified area and a previously calculated area of a surface where the component P in that posture contacts the light transmitting plate 23 (hereinafter, referred to as a specified area) are equal. The threshold value updating means 63 captures the reflected light image of the component Ps in the sample area at a predetermined cycle.
And the threshold setting means 62 resets the threshold.

The control section 7 selects one of the component images detected by the feature detecting means 61 whose binarized area is within the above-mentioned allowable tolerance of the specified area value. The center of gravity position and orientation are detected. The control unit 7 causes the robot 3 to sequentially pick up the component P and supply it to another device (not shown) according to the position and orientation of the selected component in the recognition table 2, or directly on the transport conveyor. It can be incorporated into a positioned work.

The operation of this embodiment will be described with reference to the drawings. FIG. 2 is an enlarged perspective view of a component P used in the present embodiment. As shown in FIG. 3, the component P is randomly distributed from a hopper (not shown) on the light transmitting plate 23 of the recognition table 2. I have. Also, sample area 2 of the recognition table
A component Ps in which the component P is in a specific posture (the posture in FIG. 2) is placed in 30. The posture sample Ps is restrained on the light transmitting plate 23 by a clamp mechanism or the like (not shown).

The component group on the light-transmitting plate and the posture sample Ps are
The light is illuminated from below by the illumination means 4 and its reflection plate 41 via the light transmitting plate 23, and reflected light from them is taken into the imaging means 5. FIG. 4 is a diagram showing the captured reflected light image. The captured image information is sent to the feature detecting means 61 of the image recognition unit 6 and labeling is performed. That is, since there are a plurality of recognition target images in one screen, the individual images are distinguished before the analysis of the shape is started, and the process proceeds to the extraction of the feature of the shape for each distinguished image.

When the feature detecting means 61 performs the feature extraction of the image, first, the threshold value setting means 62 sets the feature detecting means 6.
In order to set the threshold for binarization in 1,
As shown in the upper part of FIG. 5, a density histogram of an image in the sample area 230 is created by setting the vertical axis to the number of pixels N and the horizontal axis to density D. Then, a density value at which the cumulative distribution from the smaller density value (the area occupied by the posture sample image Vs) is equal to a predetermined area calculated in advance is found, and this is set as a threshold value d1.

Here, an image V1 in the area labeled L1 will be described as an example. The feature detecting means 61
With the threshold value d1 set by the threshold value setting means 62, the grayscale image in the L1 area is binarized, and the smaller binarized area V1 is calculated. Then, when the area V1 is within the above-mentioned tolerance of the specified area, the control unit 7 determines that the component P in the area L1 is in the same posture as the posture sample Ps. The feature detecting means 61 calculates a binarized area for other similarly labeled areas, and the control section 7 performs a posture discrimination. Detect position and orientation. In this way, the control unit 7 obtains the position and orientation data relating to the component having the same orientation as the orientation sample.

The control unit 7 positions the chuck 31 of the robot 3 just above the component P in the L1 region, for example, according to the obtained position and orientation of the same posture component in the recognition table 2. Then, the chuck 31 is lowered onto the light transmitting plate 23 while rotating the chuck 31 by a necessary amount θ according to the orientation data so that the tip of the chuck 31 sandwiches the web portion of the component P, thereby clamping the component P. Further, the control unit 7 raises the chuck 31 holding the component P and causes the robot 3 to supply the same to another device (not shown) in a predetermined direction, or directly incorporates the workpiece into the work positioned on the transport conveyor. Or let it go.

When all the parts having the same posture as the posture sample Ps on the light transmitting plate 23 are removed by the robot 3,
The control unit 7 drives a swinging mechanism (not shown) to eject a component group in another posture, and newly supplies a component to the hopper. The control unit 7 drives the threshold value updating means 63 every component supply cycle or every several component supply cycles, and causes the threshold value setting means 62 to reset an appropriate threshold value. For example, as shown in the lower part of FIG. 5, when a deviation occurs in the density histogram of the sample area, the threshold value setting means 62 (in this case, the overall image density is high) Similarly, a threshold value d2 at which the area occupied by the posture sample image Vs is equal to a predetermined area calculated in advance is set. The threshold value d2 is a density value shifted from the threshold value d1 shown in the upper part of FIG. In this manner, the component supply device 1 always supplies components in a specific posture by image recognition using an appropriate threshold.

In this embodiment, the illuminating means and the imaging means are fixed below the recognition table to capture the reflected light image of the component.
Although the discriminated component is taken out from above the recognition table by the robot, the illumination unit and the imaging unit are attached to the robot arm to take a reflected light image from above the component, or the imaging unit is attached to the robot arm to attach the imaging unit to the robot arm. It may be configured to capture a silhouette image from above.
Further, in the present embodiment, a description has been given of an example in which a component having a specific posture is extracted from a large number of component groups. However, the image recognition device of the present invention is not limited to the present embodiment, and a recognition target, imaging means, and illumination means If the positional relationship with is constant, the present invention can be applied to an apparatus that determines the posture of each article.

[0020]

According to the present invention, an image of a component sample held in a specific posture is captured at a constant period, and the area of the captured image is matched with a specified area of the component in that posture.
Since the threshold value for binarization is updated in each cycle, even if the gray level of the image changes due to a change over time of the illumination means or a change in external illumination light, the same posture as the sample is obtained. There is little mistake in selecting parts.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram illustrating a configuration of a component supply device according to an embodiment of the present invention.

FIG. 2 is a perspective view of components used in the embodiment.

FIG. 3 is a perspective view illustrating a state of a component group on a recognition table.

FIG. 4 is a diagram illustrating a reflected light image of the component group in FIG. 3 captured by an imaging unit.

FIG. 5 is an explanatory diagram showing a threshold setting / updating method.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Component supply apparatus 2 Recognition table 3 Horizontal articulated robot (transfer means) 4 Illumination means 5 Imaging means 6 Image recognition part 61 Feature detection means 62 Threshold setting means 63 Threshold update means 7 Control part P parts V parts Reflected light image

──────────────────────────────────────────────────の Continued on the front page (51) Int.Cl. ⁶ Identification code FI G06T 1/00 B65G 47/90 Z 7/00 G06F 15/62 400 7/60 15/64 400J // B65G 47/90 15 / 70 355 (58) Fields surveyed (Int. Cl. ⁶ , DB name) G01B 11/00-11/30

Claims (2)

(57) [Claims] 1. An apparatus for imaging an object to be recognized and recognizing a posture of the object, an illuminating unit for irradiating the object with light from a predetermined direction, and a reflected light image from the object. Imaging means for imaging the object from a predetermined direction; a sample holding unit for holding a sample having the same shape as the object in a specific posture; and binarizing the reflected light image captured by the imaging means to obtain an area of the image. A threshold value for performing binarization in the feature detection unit is set such that a binarized area of a reflected light image from the sample is equal to a predetermined area value. A threshold setting unit; and a threshold updating unit that causes the imaging unit to capture the reflected light image of the sample at a predetermined cycle, and causes the threshold setting unit to reset the threshold. Detection means By determining whether the area of the reflected light image of the object to be detected is within the tolerance of the specified area value, it is possible to recognize whether or not the posture of the object is the same as the sample. Image recognition device. 2. A component supply apparatus for recognizing a plurality of components as images and supplying only a component in a specific posture to a predetermined position, wherein a plurality of components are placed, and the component is mounted on a part of itself. A recognition table having a sample area for placing a sample having the same shape as that of the recognition table, a transfer unit for taking out a component on the recognition table to a predetermined position, and irradiating the recognition table with light from a predetermined direction. Illuminating means, imaging means for imaging the reflected light image from the recognition table from a predetermined direction, binarizing the reflected light image captured by the imaging means, distinguishing individual component images in the imaging screen, A feature detecting means for detecting an area and an orientation of each component image; and a threshold value for performing binarization in the feature detecting means, wherein a binarized area of the reflected light image of the sample is equal to an area value defined in advance. Threshold setting means for setting the threshold value so that the reflected light image of the sample is taken into the imaging means at a predetermined cycle, and the threshold setting means resets the threshold value. An updating unit, among the component images detected by the feature detecting unit, selecting an image whose area is within an allowable tolerance of the specified area value, and assigning a component to the transfer unit according to the position and orientation of the corresponding component. And a control means for picking up the sheet and transferring the sheet in a predetermined position.