JPH06137837A - Part attitude judgement method for visual attitude selection device - Google Patents

Abstract

PURPOSE:To provide a part attitude judgement method for visual attitude selection device capable of determining reasonably the identity number of black picture elements to judge to be the forward attitude to any operator. CONSTITUTION:Many parts mixed in forward and reverse attitudes are transferred in front of a line sensor type camera, and the number of identity picture elements is counted for each attitude. In this case, identity of feature point Q (%) = Number of identity picture elements/total number of picture elements in line image range, and a frequency distribution of frequency number N of part for each value of Q is prepared. Then, the number of identity picture element of Q for the minimum frequency between the maximum frequency on 0% side and the maximum frequency on 100% side is set at the judgment level number for transfer part in forward attitude.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for determining a component posture in a visual posture selecting device.

[0002]

2. Description of the Related Art In an automatic assembly line or the like, when a component to be fed (hereinafter referred to as a work) has directionality, its posture is determined during the feeding, and a work having a bad posture is sorted by a sorting mechanism. Try to eliminate with.

Conventionally, the work posture is discriminated using a mechanical means such as an attachment provided on the work feeding path. However, some mechanical means cannot discriminate the posture depending on the shape of the work. As the shape of the work changed, the attachment had to be changed, so changing the work was troublesome.

In recent years, an image sensor and an image processing technique for processing the image output thereof have been developed. Therefore, a moving work is imaged using a two-dimensional image sensor (in particular, a CCD camera), and an image of this image sensor is taken. A method of processing the output to determine the work posture has been proposed, but the image input time of this type of image sensor is, for example, 1/6.
Since it is as large as 0 sec and it takes a long time to process the image output, there is a problem that the work transfer speed and the work transfer interval are restricted, and the processing capacity of the drive assembly line and the like cannot be sufficiently increased. In order to provide the image processing apparatus with a timing for taking in the image data, it is necessary to separately dispose a position sensor for detecting that the work has entered the visual field of the image sensor.

Therefore, the applicant of the present invention can obtain a two-dimensional image of a moving work by using a line type CCD image sensor having a high image input speed, without using the position sensor, and determine the work posture. Japanese Patent Application No. 2-106 discloses a work posture determining device capable of accurately performing a work shape that is difficult with the mechanical means described above at high speed in a short time and without being affected by the work moving speed.
Proposed as No. 448.

FIG. 3 is a block diagram showing this work posture discriminating apparatus. In the figure, reference numeral 10 denotes a light source for illumination, which horizontally moves a work W on a work transfer path 21 of a transfer device (for example, a rotary feeder shown in FIG. 6) at a right angle to the work moving direction (front and back direction of paper surface). Irradiate from the direction. Reference numeral 30 denotes a line-type image sensor (a line-type CCD camera in this example), which is located at a position facing the light source 10 with the work conveyance path 21 interposed therebetween, and images the work W irradiated by the light source 10. , The image data is sent in the form of a signal I ′, and this image data is supplied to the object detection circuit 41 of the image processing device 40. In addition to the object detection circuit 41, the image processing device 40 includes an image memory 42 for teaching, a template memory 43, a matching memory 44, and a determination circuit (including a comparison circuit) 45.
Is equipped with. Reference numeral 50 is a teaching device. The CCD camera 30 is arranged so that the direction of the pixel array (line direction) is the direction as shown by the chain line in FIG. The image data of is output. As the CCD camera 30, for example, the number of effective pixels: 128 pixels, the clock frequency: 1 MH
z, image input time: 250 μsec is used.

The object detection circuit 41 starts transmitting the input image data from the first timing when the state in which there is no "black pixel" in the image data is changed to the state in which there is "black pixel", and the "black pixel" is output. "Black pixel" from the state that "pixel" exists
At the second timing when the state has disappeared, the transmission of the input image data is interrupted.

With this arrangement, first, the work W moving in the forward posture is imaged by the CCD camera 30, the two-dimensional image is stored in the image memory 42, and then the image data is taught from the image memory 42. The image of the work W is read out on the screen of the teaching device 50 by reading it out to the device 50. Vertically-oriented (long in Y direction) rectangular frames extending outside the image and inside the image are set on both sides of a predetermined position on the image projected on the screen, that is, a changing position where the shape changes. In this example, as shown in FIG. 5, it comprises a base W2 of the work W and a base W2 immediately thereafter. Frame 61, work W
The rectangular frame 62 is set to the base W ₁ of the base W ₂ . Although the rectangular frames 61 and 62 have different coordinates in the X direction,
The coordinates in the Y direction are the same. The image data of the set rectangular frame 61 is set as (FORWARD template) to the rectangular frame 62.
The image data (BACKWARD template) therein is stored in the template memory 43.

After the above preparations, the postures of the works W sequentially transferred from the work supply line onto the work transfer device 21 are determined. The image data output by the object detection circuit 41 is stored in the matching memory 44. The matching memory 44 receives the image data of one line sequentially sent by the object detection circuit 41 and stores the matching image data capable of judging the matching / mismatch with the template. The determination circuit 45 compares the template stored in the template memory 43 with the matching image data captured by the matching memory 44 at each timing when the object detection circuit 41 outputs the image data for one line. A difference in the number of matching “black pixels” is detected, and if this difference is less than or equal to a threshold value, it is determined to be matching (matching). That is, first, the matching image data stored in the matching memory 44 is FOR
The operation of comparing with the WARD template is performed, and if they match, then the matching
BACKW the matching image data in the memory 44
Compared to the ARD template, the decision circuit 45 determines that the order of matching is first, the FORWARD template, then the BA
When it is in the order of the CKWARD template, it is determined that the posture of the work W is the forward posture, and a pass signal is generated.

Specifically, in order to prevent deterioration of the determination accuracy due to noise, for example, the width of each of the rectangular frames 61 and 62 is set to 5 dots, and as shown in FIG. 5, n pieces (in this example, n =
5) The FORWARD template is stored in the template memory 51 including the line memories 51 _{M1 to} 51 _M5 . Similarly, although not shown, the above BACKWARD template is stored in the other n line memories.
The matching memory 52 is a first-in first-out (FIFO) memory having line memories 52 _{M1 to} 52 _M5 of the same size as the template memory 51. These line memories have a memory capacity for storing image data for one scanning line of the CCD camera 30. 54 is CC
It is a pulse generation circuit that gives a timing for reading data for one pixel of the D camera 30. Matching memory 5
Each time the line memories 52 _{M1 to} 52 _M5 of No. 2 receive the pulse of the pulse generation circuit 54, the data of one pixel stored first is sent to the memory of the preceding stage and the one pixel of the above described is sent out by the memory of the following stage. Stores the data of. Judgment circuit 53
Is the line memory 5 in the matching memory 52.
2 _M1 data is template memory 51 _M1 data, line memory 52 _M2 data is template memory 51 _M2 data and line memory 52 _M3 data is template memory 51 _M3 data, line memory 52 _M4 data into template memory 51 _M4 data and line memory 52 _M5 data into template memory
The data is compared with the data in the memory 51 _M5 .

With such a hardware configuration, the collating operation can be performed while the image data for one scanning line is being captured from the CCD camera 30,
The determination processing time can be shortened. In the work posture determination device of FIGS. 4 and 5, the order of change of image data in a direction perpendicular to the work movement direction (X direction), that is, in the direction Y (height direction) (from large to small, from small to small). Since the work posture is determined from (large), there is an advantage that the work posture can be determined without waiting for the rear end of the work W to be detected.

However, in the above-mentioned conventional example, the rectangular frames 61 and 62 of the work W in the normal posture and the reverse posture are set, the number of black pixels at this position is counted, and the black pixels of the black pixel are changed depending on the traveling direction of the work W. It is determined whether the direction is the forward direction or the backward direction by determining whether the larger one is first or later. In any case, the rectangular frames 61 and 62 are set by the operator of the teaching device 50. While actually flowing the work W in the forward direction and the backward direction, while observing the position near the front end portion in the forward direction and near the front end portion in the backward direction, the positions in which the forward and backward directions can be reliably recognized. Since the positions of the rectangular frames 61 and 62 are determined, the operation is troublesome in any case.

A method has also been developed in which one window 70 as shown in FIG. 7 is used to determine whether the work is in the forward or reverse direction. In other words, the work W reaches the front of the line sensor in a positive posture, and the operator sets the portion where the shape changes the most while operating the teaching device. As shown by A in FIG. 1 with 128 dots
An explanation will be given of a case where the window 70 is constituted by three continuous lines and the normal posture of the work W is judged by this. In the state of "1", no black pixel appears in the window 70, and the work progresses. In the state of "2", the black pixel at the tip is obtained on the rightmost line, and further "1"
When going a distance between "2" and "2", the state becomes "3", and the black pixel appears on the rightmost and middle lines as shown in the figure. All black pixels appear, but this time is not yet the time for judging the normal attitude, but in this conventional example, when the window 70 detects the work W in the state of "5", the normal attitude is judged. There is. That is, as shown on the leftmost side of FIG. 8B, if the posture of the work W is judged at the position "5" in the template memory, the work W is being transferred and the window 70 When the number of black pixels reaches the position corresponding to the position of "5", the number of black pixels is counted by the CCD camera 30 (step 71), and the number of black pixels stored in the template memory is already set. Therefore, these are compared (step 72), and if the number of coincidences is larger than the number of determination levels, it is determined in step 73 that it is the positive direction, and if it is smaller, it is determined that it is the different direction in step 74. The predetermined reject means is operated to remove it from the predetermined transfer path.

In the teaching method as shown in FIG. 7, the operator also operates the teaching device 50 while the operator moves the window 70 in the forward direction so that the front end of the work W is at the position of this window, that is, immediately before the line sensor. After reaching the position "5", the template memory counts the number of black pixels at this position when the position "5" is reached. It is necessary to perform a precise operation to judge that.

As described above, assuming that the distance from the detection of the tip of the work in the normal orientation to the arrival at the position of the window is S, the number of coincidences of black pixels at this position is a predetermined number or more. If so, it is determined that the posture is in the positive direction, but conventionally, how to set this set value depends on the setting by the operator's thought and error, that is, the experience. Therefore, the probability of the forward posture supplied to the next process by the operator varies.

[0016]

SUMMARY OF THE INVENTION The present invention has been made in view of the above-described problems, and rationally determines the number of black pixel coincidences for determining that the posture of the operator is in the positive direction for any operator. An object of the present invention is to provide a component posture determination method in a visual posture selection device capable of performing the above.

[0017]

The above object is to arrange a line sensor type camera and a light source with the part sandwiched in the direction orthogonal to the transfer direction of the part, and to provide a line shape by the line sensor type camera. Visual posture selection in which the image pickup range is set to a predetermined position from the front end of the transfer component, and the posture of the transfer component is determined based on the number of matching pixels of black pixels and white pixels in the line-shaped image pickup range with respect to the forward direction posture. In the component orientation determination method in the apparatus, a large number of components mixed in a forward orientation and a backward orientation are moved in front of the line sensor type camera to count the number of matching pixels, and at this time, the degree of matching of feature points Q ( %) = The number of coincident pixels / the total number of pixels in the line-shaped imaging range, and a frequency distribution of the frequency N of the component for each value of Q is created to obtain the maximum frequency on the 0% side and the maximum frequency on the 100% side. Between the poles Parts attitude determination method in the visual orientation and sorting device is characterized in that a determined number of levels for the transfer part of the positive orientation matches the number of pixels of the Q for the frequency is accomplished by.

[0018]

The operator inserts a large amount of parts for setting a set value, which should be determined to be a large number of parts in the normal orientation, into a conveyor, for example, a rotary parts feeder. Then, if this rotary parts feeder is driven, the number of coincidences of black pixels is sequentially counted by the line sensor until immediately before the line sensor type camera, and this is the number of black pixels in the window preset in the computer, That is, as described in the above-mentioned prior art, in the case of three lines, if one line has 128 bits, the ratio of the counted number of coincident pixels to the number of 128 × 3 pixels is calculated to obtain a ratio or By multiplying this by 100, it is possible to count one percentage as the degree of coincidence of the feature points. Therefore, the horizontal axis represents the degree of coincidence, and the vertical axis represents the frequency distribution of the number of workpieces against this, and the maximum value on the 0% side and the maximum value on the 100% side are calculated by the computer, and the minimum value between them is calculated. By reading the value of Q for, the set value of the determination level can be automatically determined and has no relation to the skill level of the operator.

[0019]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A component attitude determining method in a visual attitude selecting apparatus according to an embodiment of the present invention will be described below with reference to the drawings.

As the visual posture selection device, a rotary parts feeder as shown in FIG. 6 is applied. That is,
The work W is a film body, but the window is made up of three pieces as described in the above-mentioned prior art, and the number of works is now 1000. The carrying capacity of this rotary parts feeder is 200 pieces per minute.

Under the above operating conditions, it is possible to simply put 1000 workpieces into the rotary parts feeder.
When the work passes in front of the line sensor type camera 30, the number of matching pixels of the work corresponding to the number of pixels in the window (128 × 3) is calculated, and these ratios are multiplied by 100 to obtain the feature point. The degree of agreement is expressed as a percentage. Also, as shown in FIG. 1, this percentage is
Suppose you want to count every 10%. 0 in this embodiment
% -10% is 0, 10% -20% is 0, 20% -3
0% is 296, 30% -40% is 188, 40%
~ 50% is 38, 50-60% is 0, 60% -70
% Is 0, 70% -80% is 3, 80% -90% is 2
The data of 0 pieces, 460 pieces of 90% to 100% can be stored in the computer, and in some cases, this may be displayed on the display. When it is displayed on the display, the minimum value of the bar graph is clearly recognized when Q is 60%, so it can be seen that this may be used as the determination level number. In addition, it is natural that the maximum value is 460 at 90% to 100%, but the maximum value of 296 is observed at 20% to 30%. Of course, although the number of matching pixels is different from that in the positive direction, it is clear that a certain number of black pixels are counted. Therefore, depending on the shape of the work, it is possible that this maximum value becomes even lower or higher, but in any case, the maximum value on the 0% side is when the work in the reverse posture passes just before the window. It is counted.

In the above, the degree of coincidence Q ₀ of the feature points every 10%
Was calculated by a computer, and the frequency distribution of this was taken, but if the range of this Q is made smaller, for example, if it is divided into 2%, a more accurate frequency distribution will be obtained, and finally, in FIG. It should be represented as a curve as shown. In this case, if Q ₀ of the minimum point becomes the determination level, it becomes an ideal value, but if it is larger than this, the number of workpieces in the forward direction supplied to the next process decreases, but a more accurate determination is made. It is clear that

Although the embodiments of the present invention have been described above, of course, various modifications can be made based on the technical idea of the present invention.

For example, in the above embodiment, a large amount of the work W for which the determination level is required is thrown into the bowl and is conveyed by the rotary parts feeder to create the frequency distribution as shown in FIG. The apparatus is not limited to this, and for example, a belt conveyor may be used to arrange the line sensor type camera 30 and the light source in a direction perpendicular to the conveying direction.

[0025]

As described above, according to the component posture determining method in the visual posture selecting apparatus of the present invention, it is possible to easily and constantly determine that the forward posture is constant regardless of the skill level of the operator. The determination level of can be set.

[Brief description of drawings]

FIG. 1 is a frequency distribution for explaining a component posture determination method in a visual posture selection device according to an embodiment of the present invention.

FIG. 2 is a graph that approximates an ideal frequency distribution.

FIG. 3 is a block diagram of a conventional component attitude determination device.

FIG. 4 is a block diagram of a main part in FIG.

5 is a front view showing an example of a work whose posture is determined by the apparatus of FIG.

FIG. 6 is a cross-sectional view showing an example of a transfer device applied to the same device.

FIG. 7A is a schematic view showing a teaching method of another conventional component posture selecting apparatus, and B is a similar schematic view.

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁵ Identification code Internal reference number FI technical display location G06F 15/62 400 9287-5L

Claims (1)

[Claims] 1. A line sensor type camera and a light source are arranged with the part sandwiched in a direction orthogonal to a transfer direction of the part, and a line-shaped imaging range by the line sensor type camera is set to a predetermined value from a front end of the transfer part. In the component orientation determination method in the visual orientation selection device, the orientation of the transferred component is determined based on the number of matching pixels of the black pixels and the white pixels in the line-shaped imaging range with respect to the orientation in the forward direction. A large number of parts mixed in posture and reverse posture are moved in front of the line sensor type camera to count the number of matching pixels, and at this time, the degree of matching of feature points Q (%) = the number of matching pixels / line-shaped imaging With the total number of pixels in the range, a frequency distribution of the frequency N of parts for each value of Q is created, and the matching pixel of Q with respect to the minimum frequency between the maximum frequency on the 0% side and the maximum frequency on the 100% side is created. Positive number A component posture determination method in a visual posture selection device, wherein the number of determination levels for the posture transfer component is set.