JP2712887B2 - Work front / back identification method and apparatus - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and an apparatus for discriminating the front and back of a work, and more particularly, to a method in which the front and back of a work sequentially fed by a component feeder such as a bowl feeder exist. The present invention relates to a method for identifying the front and back of a work for identifying whether the work is facing up or down and a device therefor.

[0002]

2. Description of the Related Art Conventionally, in a system for gripping a work such as a part using an industrial robot, a visual system including an image detection device and an image processing device is incorporated by paying attention to an advantage that positioning of the work can be made unnecessary. A system has been proposed that controls an industrial robot based on the position, posture, and the like of a workpiece obtained by a vision system ("Part supply system using vision").
Hitachi, Ltd. Production Technology Laboratory Michio Takahashi 1988
29th, Japan Society of Precision Engineering, Automatic Assembly Special Committee, 62nd Research Presentation).

In this system, the outline of a binary image area is approximated by a polygon, and data of a series of vertices is used to calculate the approximate value.
The graphic features of ten or more items such as the center of gravity, the area, the perimeter, and the like are calculated, and these many graphical features are combined by a user program to determine the type, orientation, position, and the like of the work. Therefore, if a user program is created to indicate how to combine the graphical features, then it is possible to determine the type of work, determine the orientation, and determine the gripping position without any special work. Can be.

[0004]

However, as described above, it is necessary to create a user program, and it is necessary to create a user program according to the type of work. In the current situation where small-lot production is becoming common, not only does the amount of work required to create a user program significantly increase, but also the user program changes in response to changes in the shape of the work accompanying model changes, etc. Is required and the user
There is a disadvantage that the amount of work for changing the program is increased.

[0005] Furthermore, when the work has a stable state of front and back, and there are a plurality of protrusions on the front or back surface, or as shown in FIGS. 10 (A) and 10 (B). If the projections are separated from each other when the projections are curved, it will be recognized that a plurality of workpieces exist based on the reflection images of the projections. In the latter case, FIG. As shown in (C), only both ends are obtained as reflection images, and it is recognized that a plurality of works exist based on each reflection image, so that it is impossible to accurately identify the front and back of the work. There is a disadvantage that it becomes. In particular, as shown in FIG. 11, when the work has a considerably large concave portion, it is conceivable that another work enters the concave portion. Of the reflection images existing in the area of the rectangular window (see the broken line in FIG. 11), it is impossible to identify which reflection image belongs to one work and which reflection image belongs to another work. As a result, there is a disadvantage that it is impossible to distinguish between the front and back of the work. Although FIG. 11 shows a case based on a rectangular window, a similar inconvenience occurs when a circular window is set.

[0006]

SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned problems, and eliminates the need for creating a user program for each work, and is capable of accurately identifying the front and back surfaces regardless of the shape of the work. It is an object of the present invention to provide a novel method of identifying the front and back of a work and a device therefor.

[0007]

According to a first aspect of the present invention, there is provided a method for identifying the front and back of a work, wherein the silhouette image of a teaching work is obtained based on transmitted light, and is equal to the outline of the silhouette image. A method of generating a window having an outline, extracting feature values of a plurality of teaching works based on reflected light from the teaching work in the window, and identifying front and back of the target work based on the extracted feature values It is.

In this case, discrimination between front and back is performed by selecting an area value as an initial feature value. If sufficient discrimination accuracy cannot be obtained with the area value, another feature value is selected to perform front and back discrimination. preferable. An apparatus for identifying the front and back of a work according to claim 2, wherein the image detecting means detects an image of the work.
A first work piece disposed on the opposite side of the image detection means with respect to the work;
An illumination unit, a second illumination unit disposed on the same side as the image detection unit with respect to the workpiece, and an outline of the silhouette image based on a silhouette image of the teaching work obtained by the image detection unit by light from the first illumination unit. Window generating means for generating a window having a contour line equal to the line; feature value extracting means for extracting feature values of a plurality of teaching works based on reflected light from the teaching work in the window; Target work front and back identification means for identifying the front and back of the target work based on the value.

[0009]

According to the first aspect of the present invention, a silhouette image of a teaching work is first obtained by irradiating transmitted light, and a window having a contour line equal to the contour line of the silhouette image is generated. Then, light having a direction opposite to that of the transmitted light is applied, and feature values of the plurality of teaching works are extracted based on the reflected light from the teaching work in the window. Thereafter, the front and back of the target work can be identified based on the extracted feature values.

According to a second aspect of the present invention, the teaching work is first irradiated with transmitted light by the first illuminating means, and the transmitted light is received by the image detecting means. And a window having a contour line equal to the contour line of the silhouette image is generated by the window generating means. The second illuminating means irradiates the teaching work with light in a direction opposite to the transmitted light, and transmits the light to the feature value extracting means based on an image or the like positioned in the window among the images and the like input by the image detecting means. The feature values of a plurality of teaching works are extracted. Then, the front and back of the target work can be identified by the target work front / back identification means based on the characteristic value extracted by the characteristic value extraction means.

[0011]

BRIEF DESCRIPTION OF THE DRAWINGS FIG. FIG. 8 is a view schematically showing a configuration of a component supply apparatus to which the method for identifying the front and back of a work according to the present invention is applied.
A part of the component transport truck 1a of the circulation type bowl feeder 1 is constituted by a translucent plate 1b, and the lighting devices 2 and 3 are arranged above and below the translucent plate 1b. . Then, a half mirror 3a is arranged between the translucent plate 1b and the lower illumination device 3, and a CCD is provided at a predetermined position where the light reflected by the half mirror 3a can be received.
An image detection unit 4 including a camera and the like is arranged. Further, the industrial robot 5 is provided with a suction head 5a for sucking a work. Further, an image processing unit 4a that performs a binarization process or the like by using an image signal captured from the image detection unit 4 as an input, and a robot that receives a processing result signal from the image processing unit 4a and gives an operation command to the industrial robot 5 And a controller 4b.

FIG. 9 is a front view showing a main part of the component supply device, in which components are sequentially transported on a component transport truck 1a.
Light is emitted from the illumination device 2 while being positioned on the translucent plate 1b, and guided to the image detection unit 4 via the half mirror 3a to obtain a silhouette image. Thereafter, light is emitted from the illumination device 3, and the reflected light from the component is guided to the image detecting unit 4 via the half mirror 3a to obtain a reflected image. The parts identified as being removable based on the silhouette image and the reflection image are removed by the industrial robot 5 from the translucent plate 1b, and the parts identified as being unremovable are determined. Then, it falls down the step portion following the plate 1b having translucency and is circulated as it is. It is preferable that the lighting device 2 be movable as indicated by a two-dot chain line, so that the operation of taking out parts by the industrial robot 5 is not hindered.

FIGS. 1 to 4 are flowcharts showing an embodiment of a work identification method according to the present invention. FIG. 1 shows a teaching process for a work to be identified, FIG. 2 shows a feature value selection process for work identification, and FIG. FIG. 3 shows a work identification process based on the selected feature value, and FIG. 4 shows a process associated with window creation. In the flowchart of FIG. 1, in step SP11, an image of a work in the same stable state with n samples is captured by an image detecting device such as a CCD camera, and an area value is selected as a feature value in step SP12. Calculating the variance (eg, standard deviation σ) and the average value xav,
In step SP14, THup = xav + kσ and THdn = xav-kσ (where k is a constant of about 1 to 3) are calculated to calculate the upper threshold THup and the lower threshold THdn of the area value.

Note that, from step SP11 to step SP1
A series of processes up to 4 are performed for each type of work. In the flowchart of FIG. 2, in step SP21, the variance ratio F0 of each characteristic value is calculated for the front work and the back work, and in step SP22, it is determined whether the variance ratio of the area value is smaller than a predetermined value. If it is determined that the variance ratio is smaller than the predetermined value, step S
In P23, the characteristic value having the maximum variance ratio is selected, and in step SP24, the variance and average value of the selected characteristic value are calculated. In step SP25, the upper threshold THup and the lower threshold THdn are set in the same manner as in step SP14 of FIG. calculate. Conversely, if it is determined in step SP22 that the variance ratio is not smaller than the predetermined value, the process from step SP23 to step SP25 is omitted because the area value may be adopted as the feature value.

In the flowchart of FIG. 3, in step SP31, an image of the work to be identified as the front / back side is fetched, and in step SP32, the characteristic value selected by the processing of FIG. 2 is obtained. And step SP33
In step SP33, it is determined whether or not the area value is between the upper threshold value and the lower threshold value. If it is determined in step SP33 that the area value is between the upper threshold value and the lower threshold value, the area other than the area is determined in step SP34. It is determined whether there is a feature value for which the upper threshold value and the lower threshold value have been calculated. If it is determined that there is a feature value, the upper and lower threshold values for which the corresponding feature value has been calculated are determined in step SP35. It is determined whether or not the value is between.

If it is determined in step SP34 that there is no characteristic value, or if it is determined in step SP35 that the characteristic value is a value between the calculated upper threshold and lower threshold, Step SP
In 36, the corresponding work is processed, that is,
Identify the work as ostensible. Conversely, step SP
If it is determined in step SP35 or step SP35 that the value is not between the upper threshold and the lower threshold, the process proceeds to step S35.
In P37, the corresponding work is identified as a disturbance work, that is, a work facing down.

FIG. 4 is a flowchart showing steps SP11 and SP12 of the flowchart of FIG.
Step SP31 and Step S in the flowchart of FIG.
The process with P32 will be described in detail. In step SP41, a silhouette image of a work is fetched by the image detection unit 4 based on the light from the lighting device 2, and in step SP42, connectivity analysis is performed based on the silhouette image. In step SP43, a window is created based on the coordinates of the points forming the outline, and in step SP44, the reflected image is obtained by the image detection unit 4 based on the light from the lighting device 3. And in step SP45, only the reflection image existing inside the window is extracted, and step SP46 is performed.
In, feature values are extracted based on the extracted reflection images. As a result of performing the processing in FIG. 4, even if the silhouette image includes a hole, for example, the coordinates of the points constituting the contour of the hole cannot be obtained.

More specifically, as a result of performing the processing based on the flowchart of FIG. 1, when the distribution shown in FIG. Since there is no overlapping portion, for example, the upper threshold THup and the lower threshold THdn are set as shown in FIG. 6 for a work facing up, and the features obtained for the work to be processed in the flowchart of FIG. By determining whether or not the value is within the range between the upper threshold value THup and the lower threshold value THdn, it is possible to achieve high-precision identification of the front and back of the work. Conversely, as a result of performing the processing based on the flowchart of FIG. 1, the distribution shown in FIG. 5B is obtained for each of the front-facing work and the back-facing work (the dispersion ratio in FIG. 5A is FA, If the variance ratio in FIG. 5B is FB and FA> FB), there is a considerable overlap between the two distributions. Therefore, even if the processing in the flowchart of FIG. Identification cannot be achieved. Therefore, in this case, by performing the processing of the flowchart of FIG. 2 and selecting the characteristic value having the largest dispersion ratio, it is possible to achieve high-precision identification of the front and back of the work.

That is, when it is necessary to identify the front and back of a work, identification based on a reflection image must be performed because identification based on a silhouette image is impossible. It may exist, and if it is, the feature value will be extracted for each reflection image. However, in this example,
Since the reflection image existing inside the window based on the silhouette image is grasped as a reflection image of one work, a highly accurate feature value extracted based on all the reflection images existing inside the window is used. The front and back of the work can be identified. In addition, since it is often possible to achieve high-precision identification of the front and back of the work based on the area value among the characteristic values, it is often unnecessary to select another characteristic value based on the flowchart of FIG. Since it is only necessary to select another characteristic value based on the flowchart of FIG. 2 for only some of the workpieces, it is possible to reduce the processing load as a whole and achieve highly accurate workpiece identification.

When there are three or more types of stable states of the work, the processing of the flowchart of FIG. 2 may be performed for each of the stable states of the work. In addition, in step SP22 of the flowchart of FIG. Upper threshold T for one of the workpieces
It is possible to determine whether or not the upper threshold value THup or the lower threshold value THdn of another work exists within the range of the Hup and the lower threshold value THdn. It is possible to make design changes.

[0021]

Embodiment 2 FIG. 7 is a block diagram showing an embodiment of a work front / rear discriminating apparatus according to the present invention. , A window generation unit 9 that generates a window based on the coordinates of the points obtained by the connectivity analysis unit 8, and a work A work image extraction unit 10 for extracting an image present in the window from the reflection image, a feature value output unit 11 for obtaining and outputting a feature value of the work based on the extracted work image, and a sample number n
For all of the workpieces, a first calculation unit 12 that selects an area value among the extracted feature values to calculate a standard deviation σ and an average value xav, and a standard deviation σ calculated by the first calculation unit 12 and THup = x based on the average value xav
av + kσ and THdn = xav−kσ (where k is a constant of about 1 to 3) to calculate an upper threshold THup
And first threshold value calculating unit 13 for calculating lower threshold value THdn
A feature value determining unit 14 for determining whether or not the area value is appropriate as a feature value; and a feature value having a maximum variance ratio on condition that the area value is determined to be inappropriate as the feature value. And a second value for calculating a standard deviation σ and an average value xav of the selected feature value.
THup = xav + based on the calculation unit 16 and the standard deviation σ and the average value xav calculated by the second calculation unit 16.
kσ and THdn = xav−kσ are calculated to calculate an upper threshold THup and a lower threshold THdn.
A threshold value calculation unit 17, a feature value output unit 18 that obtains and outputs an area value and a selected feature value when appropriate based on the reflection image captured by the image detection unit 4, Among the upper threshold and the lower threshold calculated by the first threshold calculator 13 and the second threshold calculator 17, the magnitude of the upper threshold and the lower threshold corresponding to the corresponding feature value is determined, and the work to be processed is turned face up. A work front / back discriminating unit 19 that outputs a signal indicating that the work is present or a signal indicating that the work to be processed is facing backward.

The feature value discriminating unit 14 discriminates whether the area value is appropriate as a feature value by discriminating the magnitude of the variance ratio and the predetermined value. The determination may be made as to whether at least one of the threshold and the lower threshold is located within the range between the upper threshold and the lower threshold of the other work, and the determination is performed based on other values. You may.

Further, it is needless to say that the first calculation unit 12 and the second calculation unit 16 may be used in combination, and the first threshold value calculation unit 13 and the second threshold value calculation unit 17 may be used in combination. The operation of the front / back identification device for a work having the above configuration is as follows. First, a silhouette image and a reflection image of a work of number n of samples are taken in by the image detection unit 4 for each of the front-facing work and the back-facing work, and the connectivity analysis is performed by the connectivity analysis unit 8 based on the silhouette image, and the outline is obtained. The coordinates of points constituting the line are obtained, and a window is generated by the window generating unit 9 based on the coordinates of the points obtained by the connectivity analyzing unit 8. Then, among the reflection images of the work taken in from the image detection unit 4, an image existing in the window is extracted by the work image extraction unit 10, and the feature value output unit 11 outputs the characteristic of the work based on the extracted reflection image. Obtain and output the value. Thereafter, the first calculator 12 calculates the standard deviation σ and the average value xav for the area value for each type of work.
Is calculated, and based on the calculated standard deviation σ and the average value xav, the first threshold value calculation unit 13 calculates THup = xav
The calculation of + kσ and THdn = xav-kσ is performed to calculate the upper threshold THup and the lower threshold THdn.
Then, the feature value determination unit 14 determines whether the area value is appropriate as the feature value.

When the characteristic value judging section 14 judges that the area value is appropriate as the characteristic value, the image detecting section 4 takes in the reflection image of the work to be identified, and the characteristic value output section 11 presents the reflected image in the window. The work determination unit 19 obtains and outputs an area value based on the reflected image, and outputs the characteristic value output from the characteristic value output unit 18 to the first threshold value calculation unit 1.
Then, a signal indicating that the workpiece to be processed is facing up or a signal indicating that the workpiece to be processed is facing down is output by determining the magnitude of the upper threshold and the lower threshold calculated in step 3. That is, when the output feature value is within the range of the corresponding upper threshold value and lower limit threshold value, a signal indicating that the processing target work is facing up is output, and when the feature value is outside the above range, the processing is performed. A signal indicating that the target work is facing down is output.

Conversely, if the characteristic value discriminating section 14 determines that the area value is not appropriate as the characteristic value, the characteristic value selecting section 15 selects the characteristic value having the maximum variance ratio and the second calculating section. 16 to calculate the standard deviation σ and the average value xav for the selected feature value,
The calculation of THup = xav + kσ and THdn = xav-kσ is performed based on the standard deviation σ and the average value xav calculated by the second calculation unit 16 to calculate the upper threshold THup and the lower threshold THdn. afterwards,
The image of the work to be identified is captured by the image detection unit 4, the area value and the value of the feature value selected by the feature value selection unit 15 are obtained and output by the feature value output unit 18, and the feature value output is performed by the work determination unit 19. The size of the feature value output from the unit 18 and the upper and lower thresholds calculated by the first threshold calculator 13 and the upper and lower thresholds calculated by the second threshold calculator 17 are determined to determine the size of the work to be processed. Outputs a signal indicating that the workpiece is facing up or a signal indicating that the workpiece to be processed is facing down. That is, when the output feature value is within the range of the corresponding upper threshold value and lower limit threshold value, a signal indicating that the processing target work is facing up is output, and when the feature value is outside the above range, the processing is performed. A signal indicating that the target work is facing down is output.

More specifically, when the distributions shown in FIG. 5A are obtained for the front-facing work and the back-facing work based on the n number of samples, a portion where the two distributions overlap each other is obtained. Since there is no existing work, for example, the upper limit threshold THup and the lower limit threshold THdn are set as shown in FIG. 6 with respect to the area value of the work facing up, and the characteristic values obtained for the work to be processed are the upper threshold THup and the lower limit threshold. THdn
By discriminating whether it is within the range, it is possible to achieve high-precision identification of the front and back of the work. Conversely, when the distribution shown in FIG. 5B is obtained for each of the front-facing work and the back-facing work, there is a considerable overlap between the two distributions. Identification cannot be achieved. Therefore, in this case, by selecting the feature value having the largest variance ratio, it is possible to achieve high-precision identification of the front and back of the work.

That is, in order to identify the front and back of a work, since identification based on a silhouette image is impossible, identification based on a reflection image must be performed. However, a plurality of reflection images exist for one work. In such a case, the feature value is extracted for each reflection image as it is. However, in this example,
Since the reflection image existing inside the window based on the silhouette image is grasped as a reflection image of one work, a highly accurate feature value extracted based on all the reflection images existing inside the window is used. The front and back of the work can be identified. Also, among the feature values, it is often possible to achieve high-precision discrimination of the front and back of the work based on the area value. Therefore, it is often unnecessary to select a feature value other than the area value. Only the characteristic value other than the area value needs to be selected, so that the processing load can be reduced as a whole, and highly accurate workpiece identification can be achieved.

[0028]

As described above, according to the first aspect of the present invention, there is no need to create a user program for each type of work, and the amount of work for creating and changing the program can be greatly reduced. This has a specific effect that a high-precision front / back discrimination can be achieved for a work.

According to the second aspect of the present invention, there is no need to create a user program for each type of work, so that the amount of work for creating and changing the program can be greatly reduced, and all the works have high-precision front and back surfaces. This has a specific effect that identification can be achieved.

[Brief description of the drawings]

FIG. 1 is a flowchart showing a teaching process based on a sample work in one embodiment of a method for identifying the front and back of a work according to the present invention.

FIG. 2 is a flowchart illustrating a feature value selection process for identifying the front and back of a work in the method of identifying front and back of a work according to the present invention;

FIG. 3 is a flowchart illustrating a work identification process based on a selected feature value in the work identification method according to the present invention.

FIG. 4 is a flowchart showing a process associated with window creation in the work front / back identification method of the present invention.

FIG. 5 is a diagram schematically showing propriety of a feature value;

FIG. 6 is a schematic diagram showing an upper threshold and a lower threshold of a feature value.

FIG. 7 is a block diagram showing an embodiment of a work front / back identification apparatus according to the present invention;

FIG. 8 is a diagram schematically showing a configuration of a component supply system to which a method for identifying front and back of a work according to the present invention is applied.

FIG. 9 is a front view showing a main part of the component supply device.

FIG. 10 is a diagram illustrating a reflection image obtained based on a curved work.

FIG. 11 is a diagram illustrating a disadvantage when a rectangular window is set corresponding to a work having a concave shape.

[Explanation of symbols]

2, 3 lighting device 4 image detection unit 8 connectivity analysis unit 9 window generation unit 10 work image extraction unit
11 feature value output unit 12 first calculation unit 13 first threshold value calculation unit 14
Feature value determination unit 15 Feature value selection unit 16 Second calculation unit 17 Second threshold value calculation unit 18 Feature value output unit 19 Work front / back determination unit

──────────────────────────────────────────────────続 き Continuation of the front page (51) Int.Cl. ⁶ Identification number Agency reference number FI Technical display location G01B 11/00 G01B 11/00 Z 11/26 11/26 H

Claims (2)

(57) [Claims] A silhouette image of a teaching work is obtained based on transmitted light, a window having an outline equal to the outline of the silhouette image is generated, and a plurality of windows are generated based on reflected light from the teaching work in the window. A feature value of the teaching work, and identifying the front and back of the target work based on the extracted feature value. 2. An image detecting means (4) for detecting an image of a work, a first lighting means (2) arranged on the opposite side of the image detecting means (4) with respect to the work, and an image detecting means (4) with respect to the work. ), And a silhouette image based on the silhouette image of the teaching work obtained by the image detecting means (4) obtained by the light from the second illuminating means (3) and the first illuminating means (2). Window generating means (8, 9) for generating a window having a contour line equal to the contour line; and feature value extraction for extracting feature values of a plurality of teaching works based on reflected light from the teaching work in the window. Means (10, 11) and means (12, 13, 14, 15, 16, 17, 17, 1) for identifying the front and back of the target work based on the extracted characteristic values.
8, 19).