JPH034358B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention relates to a parts supply and assembly apparatus, and particularly to a parts supply and assembly apparatus that can be applied to all kinds of parts, regardless of the type of parts.

[Background of the invention]

In recent years, products have become more diversified and their lifespans have become shorter, and in a high-mix, low-volume production system where models are frequently changed, it is necessary to develop a production and assembly system that is compatible with this.

In this production assembly system, the parts supply section is a major barrier to fully automating the assembly process, and the generalization of so-called parts supply equipment that can be applied to all parts is an important step in promoting complete automation of the assembly process. This is an unresolved issue, and there is a strong demand for technological development in this area.

The conventional parts supply and assembly device is as described in Japanese Patent Application Laid-Open No. 56-45337, namely, as shown in FIG. A conveyance path 2 that guides the transported parts in a row, a chute 3 that leads the conveyance path products out of the bowl 1 at the outermost periphery of this conveyance path, and a position of the guided parts at the tip of this chute 3. , and a chute tip that maintains the posture and extracts parts using a handling device.

In this parts supply device, a wiper 4 and a cutout are used to arrange the parts in a single line in a conveyance path 2, to allow only parts in a specific position to pass through, and to drop parts in other positions into the bowl 1. 5 and the like, and the chute 3 is further provided with a component posture restraining member 7 for holding conveyed components that have been sorted and passed through the conveyance path 2 in the same posture.

Because of this configuration, the wiper 4 and cutout 5 provided on the conveyance path 2 and the component posture restraining member 7 provided in the chute 3 need to be redesigned every time the shape and dimensions of the component change. However, it does not respond quickly to today's high-mix, low-volume production systems, and is a major barrier to promoting production automation.

As an improvement to this, as shown in Fig. 2, instead of wipers and cutouts provided on the transport path, photoelectric switches 9A, 9B, etc. are used to detect the orientation of the parts, and only parts in a specific orientation are sent to the chute 3. drain,
For other products, use air nozzle 10, bowl 1
A parts supply device has been developed that blows the parts inside.
This device also requires a component posture restraining member 7 to maintain the same posture of the components discharged into the chute 3, and after all, the component posture restraining member 7 must be adjusted every time the shape or dimensions of the component is changed. The drawback is that it requires design changes and is not immediately compatible with today's high-mix, low-volume production systems.

In addition, the parts are discharged into the chute 3 without being restricted to a specific position or posture, and images of the parts are captured using a television camera, etc., and the posture and position of the parts are recognized by processing the signals. Attempts are being made to extract parts using a handling device such as a robot that can change the operating position, but unusable parts must be returned to the bowl manually, and full automation is not possible. There was a drawback.

[Purpose of the invention]

SUMMARY OF THE INVENTION An object of the present invention is to provide a parts supply and assembly apparatus which solves the above-mentioned conventional drawbacks and which can be readily adapted to a high-mix, low-volume production system and which also enables the simplification of an automatic parts assembly line.

[Summary of the invention]

That is, the present invention does not provide a part for restraining parts in the conveyance path, but opens the conveyance path inside the bowl,
The parts are circulated and appropriate parts are taken out while the parts are being circulated. First, the terminal end of the conveyance path is opened inside the storage part, and the parts stored in the storage part are taken out. The parts are circulated by conveying them through a conveyance path, and the parts are guided back into the storage section from the end of the conveyance path, and the position, orientation, shape, or a combination of these of the parts being circulated in this way is changed. It is detected by a detection unit such as a television camera, and the position, orientation, shape, or a combination of these is determined by the signal from this detection unit.
The judgment unit judges whether the parts are suitable or unsuitable (hereinafter referred to as suitable parts or unsuitable parts), and based on the judgment of this judgment unit, only suitable parts are taken out by the handling unit and rejected. The apparatus is characterized in that the suitable parts are guided into the storage section through the opening at the end of the conveyance path, and then placed on the conveyance path again for recirculation.

[Embodiments of the invention]

An embodiment of the present invention will be described below with reference to FIGS. 3 to 5.

FIG. 3 shows an overall overview of this embodiment. Using FIG. 3, the outline of the embodiment will first be explained as follows.

That is, the parts supply and assembly apparatus according to this embodiment has the terminal end 1 of the conveyance path 18 provided in the bowl feeder 11.
9 (see FIG. 4) is opened into the storage section 17, and a television camera 14 constituting a detection section that detects the conveyed parts 20 (see FIG. 4) is connected to the signal from this detection section (television camera 14). Therefore, there is a judging section (diagram omitted as it is a commonly used means) for judging suitable parts and unsuitable parts, and a robot 1 for extracting only suitable parts based on the judgment of this judging section.
2 and a gripping device 13, which takes out only suitable parts from the conveyed parts 20A (see FIG. 4) conveyed by the conveyance path 18, and removes unsuitable parts from the conveyance path 18. Reservoir 17 through an opening in end 19 (see FIG. 4).
The material is guided into the interior, placed on the conveyance path 18 again, and recirculated. The visual recognition device is comprised of the television camera 14 and a determining section.

By configuring in this way, the conveyance path 18
During the process, there is no need for special equipment to separate suitable and unsuitable parts, making it possible to have versatility in terms of shape and size of parts, and by circulating parts. Since only suitable parts are taken out, it can be fully automated and can be quickly adapted to high-mix, low-volume production systems.

Details of this example will be explained below. In Figure 3, 1
Reference numeral 1 designates a bowl feeder, and in this embodiment, above the bowl feeder 11, a television camera 14 (detection section) for monitoring conveyed parts is supported by support columns 15A and 15B. 12 is a robot; 13 is a gripping mechanism that constitutes a handling section; a television camera 14 (detection section) detects the conveyed parts; the detection signal is input to a judgment section (not shown) to determine whether the parts are suitable or unsuitable; Based on this judgment, the robot 12 and the gripping mechanism 13 are operated to take out only suitable parts. 16 is an assembly station.

FIG. 4 shows details of the bowl feeder 11 of this example. In the figure, a spiral conveyance path 18 is provided on the inner periphery of the bowl feeder 11. A conveyance path end portion 19 of this conveyance path 18 is bent inward of the bowl feeder 11 and opened. Note that 17 is a storage section, 20 is a component stored in the storage section 17, and 20A is a transport component that is transported by the transport path 18.

FIG. 5 shows another embodiment of the end portion of the conveyance path, and FIG. b is designed so that the end portion of the conveyance path 18 is terminated as it is, and the conveyed parts are returned to the conveyance path 18.
Either it is dropped upwards or it is dropped into the storage section 17.

FIG. 6 shows another embodiment of the bowl feeder. In the figure, reference numeral 24 denotes a revolving parts feeder, in which a rotating disk 25 and an alignment track 26, whose rotating axes are inclined to each other, rotate in the same direction as shown by the arrows. The parts stored on this rotating disk 25 are moved in the direction of the alignment track 26 by the centrifugal force generated by the rotation of the rotating disk 25, and the parts stored on the rotating disk 25 and the upper conveying surface 27 of the alignment track 26 are moved toward the alignment track 26. At the transfer section 28 where the parts are flush with each other, the parts are transferred onto the upper conveying surface 27 and lifted upward by the rotation of the alignment track 26. 30 is a member corresponding to the buttful 22 shown in FIG. 5a, and the unsuitable part is this member 30.
and falls onto the rotating disk 25 again.
Note that 29 is an outer wall, which is provided to prevent parts from falling outside. The television camera 14, robot 12, and gripping mechanism 13 are the same as those described in FIG. 3, and detailed description thereof will be omitted here. In this way, the revolving parts feeder has less vibration than the vibrating feeder.
This has the effect of making it easier to recognize parts on the conveyance path.

The operation of this embodiment configured as above will be explained next.

The parts 20 stored in the storage section 17 are transported by the transport path 18. The component 20A thus transported falls into the storage section 17 from the end 19 of the transport path and is placed on the transport path 18 again.

In this way, the parts 20 stored in the storage section 17 are circulated. The position, orientation, shape, or a combination thereof of the conveyed parts 20 circulating in this manner is detected by a television camera or the like of a visual recognition device, and this detection signal is input to the determination section. The determination section determines whether the transported component 20A is a suitable component or an unsuitable component based on the position, orientation, shape, or a combination thereof. That is, the detection of the type, position, and orientation of parts by this television camera 14 is based on the literature (Gerald J. Gleasor et at: A.
Modular Vision System for Sensor−
Controlled Manipulation and Inspection
Proc.of 9th Int.Sym.Ind.Robots 1979.3.13〜15
washington) can be used. The method is to extract the part area for the image by segmentation, find the area, first moment, and second moment of the extracted image as a figure, and then find the center of gravity and the direction (posture) of the principal axis of the moment. It is something. By simultaneously calculating more than ten other feature parameters, it is possible to determine the type of part. Utilizing this, the determination unit of the visual recognition device determines whether the conveyed component 20A is a suitable or unsuitable part based on the position, orientation, shape, etc., or a combination thereof.

The robot 12 operates and takes out only suitable parts determined in this manner. Regarding this extraction, according to the command from the judgment unit of the visual recognition device,
Either the bowl feeder 11 can be temporarily stopped and the robot 12 can then be operated to take out the food, or the bowl feeder 11 can be rotated as it is and the robot 12 can be taken out by following the rotational speed recognized by the visual recognition device. .

On the other hand, if the parts are judged to be unsuitable by the judgment section, the robot 12 will not operate and the parts will be returned to the storage section 17 from the conveyance path end section 19 as they are. The suitable parts taken out by the robot 12 are assembled by the assembly station 16.

Note that in this embodiment, the detection unit includes a television camera 14.
is provided, processes the video signal captured by the television camera 14, recognizes and judges suitable parts and unsuitable parts, and operates the robot 12 based on the information sent from this judgment part. The above-mentioned effect does not change even if a device other than a television camera is used for detection, and the present invention is not limited to this.

That is, as shown in FIGS. 3 and 8, the parts supply and assembly apparatus of the present invention includes a robot system including a robot 12 and a robot controller 47, and a TV.
It consists of a visual recognition system consisting of a camera 14, an image processor 25, and a motor or magnet drive circuit 37 that starts and stops the feeder in response to signals from the image processor 25.

First, the robot system will be explained using FIGS. 3 and 7. 12 is an industrial robot, for example, an articulated robot having five degrees of freedom. The industrial robot 12 has a rotating base 12b that rotates around a vertical axis with respect to a base 12a, an upper arm 12c that rotates around a horizontal axis 46a, a forearm 12d that rotates around a horizontal axis 46b, and a forearm 12d that rotates around a horizontal axis 46b. The wrist 12e rotates around a horizontal axis 46c, and further rotates around an axis perpendicular to the horizontal axis 46c. This wrist 12e is equipped with a hand mechanism with fingers (chucks) 13 attached thereto. The robot control device 47 includes a control unit 49 for controlling the joint robot 12 having five degrees of freedom, and a control unit 49 for operating or moving the robot along a predetermined trajectory based on point-to-point programmed speeds. For this purpose, it is composed of a pre-programmed trajectory and a teaching unit 50 for teaching speed information.
The control unit 49 and the robot mechanism 12 constitute a position control system. This position control system counts output pulses generated by a pulse encoder PE connected to an actuator M with a counter 51 and feeds them to a control unit 49.
The microprocessor 52 detects a digital signal with a predetermined target value or a desired coordinate value given from the outside, and converts this digital signal into an analog signal by the D/A converter 53. and is configured to drive the actuator M.

The drive circuit 53 is a circuit that drives the actuator M based on a speed signal from a tachogenerator TG connected to the actuator M and an analog signal from the D/A converter 53. Serial interface 32 is for connection to teaching unit 50. The ROM 33 is a memory that stores programs for operating the robot. The RAM 34 stores information from the teaching operation performed using the teaching unit 50 or the information from the interface 3.
The operation locus of the robot's hand mechanism 13, which is the result of interpolation calculation performed by the calculation unit 36 on the operation information inputted from the image processor 52 via the image processor 5, is stored. 37 is a bus line.

Robot hand mechanism 13 stored in RAM34
The position data is read out by the microprocessor 52, coordinates are converted into rotational displacements θ ₁ , θ ₂ , ... θ ₅ detected by the counter 51, and the desired or target position (for example, input from a visual recognition system) is The robot's hand mechanism 13 is driven to the target object position.

Furthermore, in order to synchronize with the assembly station 16,
A synchronization signal from the assembly station 16 is input to the interface 35.

Although the above description of the robot system is limited to a robot having five degrees of freedom, a robot having other degrees of freedom, for example, six degrees of freedom, can be constructed in the same way. In the present invention, there are no particular limitations on the degree of freedom. However, even when assembling simple circular parts such as gears, three degrees of freedom are required.

Next, visual recognition, which is another component of the present invention, will be explained using FIG. That is, TV
The video signal captured by the camera 14 is converted into a digital signal by the A/D converter 44, and is stored in the image memory 41a composed of RAM.
One screen imaged is stored. Then, the microprocessor (B) 38 reads out the screen stored in the RAM 41a, compares the binary levels of each picture element and the phosphorous picture element, and if they show the same level,
It is assumed that the RAM (segmented image memory) 41b is connected, and a specific same level is assigned to separate and identify each connected area.
to memory. That is, the microprocessor (B)
38 examines the connection relationship of the binary pixelization pattern,
The pattern area corresponding to each part is segmented, and each segmented pattern is stored in the RAM 41b. Next, the microprocessor (B) 38 uses the calculation unit 43 to calculate the area, first moment, and second moment for each segmented pattern, that is, for each region with a specific label, according to the program stored in the ROM 39. Calculate, RAM (data memory) 41c
to be memorized. Furthermore, the microprocessor (B) 38
The calculation unit 43 calculates the center of gravity of the pattern for each part and the direction (posture) of the moment principal axis from the area, primary moment, and secondary moment stored in the RAM 41c.
and store it in the RAM 41c as well. The microprocessor 38 simultaneously processes over ten other characteristic parameters (for example, the length in the major axis direction indicating the longest length, the length in the minor axis direction indicating the shortest length, the major axis and the minor axis relative to the above-mentioned moment principal axis). angle of inclination)
is calculated by the arithmetic unit 43 to determine the type of the component, and based on the determined result, it is determined whether a suitable component is an unsuitable component the next time the robot assembles the robot. If the component is determined to be suitable, the microprocessor (B) 38 sends a stop signal to the motor or magnet drive circuit 37 via the parallel I/F 40 to stop feeding the component to the feeder. The microprocessor (B) 38 newly captures the image from the TV camera 14 into the RAM 41, reads out this captured screen, segments it into a pattern for each part, and stores it in the RAM 41b. Then, the center of gravity and the direction (posture) of the principal moment axis of the pattern for each part are determined, and the results are sent to the robot controller 2 via the interface 42.
Feedback to 4. The microprocessor (A) 52 corrects the horizontal position coordinates and horizontal orientation information stored in the RAM 34 using the feedback values, and controls the robot 12 via the D/A converter 53. . The robot 12 is stopped in the feeder with the hand 13, grabs the part judged to be suitable, moves to the position of the part to be assembled at the assembly station 16, and assembles it.The robot 12 then determines whether the part is suitable or unsuitable. When doing
It is also possible to match the center of gravity of the pattern obtained as described above and the direction of the principal axis of moment with the center of gravity of the dictionary pattern and the direction of the principal axis of moment, and to judge based on the degree of agreement between the two.

It is also possible to approximate the contour of the pattern by several straight lines. That is, one embodiment of a pattern matching device for implementing this is shown in FIG.

Here, 100 is a control unit that controls the flow of necessary processing for each unit described below, and 101 is
102 is the same image memory; 103 is the same preprocessing unit; 104 is the same area information memory; 105 is the contour extraction unit related to the contour extraction processing means; 106 is the same contour extraction unit; memory, 107
is a line differentiation processing unit related to the line differentiation processing means, 108
109 is the same line differentiation information memory, 109 is the same dictionary pattern memory, 110 is the area feature extraction unit related to the area feature extraction processing means, 111 is the same area feature memory, and 112 is the common area related to the pattern matching processing means. The area extraction unit 113 is the same common area memory,
114 is a recognition processing unit.

First, the input image data is input to the input control unit 1.
After being binarized with a predetermined threshold value of 01, it is input and stored in the image memory 102 as binary image data.

The image data written in the image memory 102 is
First, the preprocessing unit 103 determines which area on the image is to be the target object area, and then the position information of the representative point of that area is stored in the area information memory 104.
is memorized.

Next, the contour extracting unit 105 sequentially searches for contour pixels in the specified area based on the position information of the representative point, and stores the center coordinate values in the contour memory 106.

Based on the center coordinate values, the line segmentation processing unit 107 approximates the point sequence representing the center of the contour pixel to a line segment within a specified tolerance, and the end points of the line segment, that is, each of the corresponding vertices that approximates the area as a polygon. The coordinates are sequentially stored in the line segmentation information memory 108. The line division information is transferred to the dictionary pattern memory 109 when registering the dictionary pattern.

The region feature extraction unit 110 also extracts features of the specified region (for example, area, center of gravity, principal axis of inertia, maximum length, perimeter, etc.) into the image memory 102 and the region information memory 10.
4. Obtain from the contents of the line segmentation information memory 108 and store it in the area feature memory 111. When registering a dictionary pattern, its characteristic information is also stored in the dictionary pattern memory 109.

Furthermore, the common area extracting unit 112 extracts the currently input pattern (stored in the line segmentation information memory 108) and the dictionary pattern (stored in the dictionary pattern memory 109) so that the characteristics of both After performing coordinate transformation on the input pattern side so that they match (for example, so that the center of gravity coordinates match, or the principal axis of inertia directions match, etc.),
The common area (product figure) of both patterns is similarly determined as a vertex sequence and stored in the common area memory 113.

Finally, the recognition processing unit 114 calculates the area of the common area (product figure), calculates the ratio of this to the area of the dictionary pattern or the input pattern, and if the value exceeds a predetermined value, both patterns are It is determined that there is a match and the recognition result is output. if,
If the two patterns do not match and there are a plurality of dictionary patterns, the matching process with the second and third dictionary patterns is repeated in the same way. Note that the recognition processing unit 114 also outputs position and orientation information of the input pattern based on the relative positions of both matching patterns before conversion.

Next, FIG. 10 is a preprocessing flowchart, and FIG. 3 is a format diagram of area information.

The preprocessing involves the input control unit 101 to the area information memory 104 described above, and sequentially examines the density gradation of each pixel (“0”, “1” in the case of a binary image), and selects the same gradation. A group of pixels with different tones is extracted as one region, and a meaning is assigned to each region based on the inclusion relationship between each region (determining candidates for object regions and hole regions), and a label for each region,
Create a data table of representative targets, areas, and inclusion relationships. In other words, this processing serves to cut out the target object from the image.

As shown in FIG. 10, a typical processing example is
Starting from the upper left pixel on the image, all pixels are scanned in a raster pattern, and while determining the connection relationship between pixels of the same pitch,
Connected pixels of the same gradation are given the same label, and a collection of pixels with the same label is regarded as a region.

In this case, each region is given a different label, and the region with the largest area and touching the periphery of the image is determined to be the background region. Then, an area that is bordered by and included in the background area and whose area is within a certain range is selected as a candidate for a target object area, and is used as a subsequent processing target area. Further, a region included in the region considered as a candidate for the target object region is set as a candidate for the hole region.

The start address of each area is the address of the first pixel found when the image is scanned in the raster direction. This address is registered as the coordinates of the area representative point, and the table format of other area information is as shown in FIG. 11, for example.

Next, the contour extraction process is performed by the contour extraction section 1 described above.
05. Concerning the contour memory 106, several examples of the processing method have been reported, which compress information by expressing the coordinate series of the contour with respect to region information in a two-dimensional array. may also be used.

The process starts with the above-mentioned area representative point as the starting point, and searches for adjacent pixels one after another based on 4-connectivity or 8-connectivity for pixels at the boundary of the area, going around the area once. The process ends at the representative point position. The output result may be in the coordinate system of contour pixels.

Information is further compressed by expressing the contour coordinate series of the target area obtained by this contour extraction process by line segment approximation so that it falls within a predetermined approximation error.

This line differentiation processing is carried out by the above-mentioned line differentiation processing unit 107,
The line segmentation information memory 108 is involved, and the result is a coordinate string of both end points of each line segment. That is, it is expressed by a series of vertices obtained by approximating the area to a polygon. Any method may be used for this line segmentation processing, but as one that specifically satisfies the requirements for high-speed processing,
An example is shown below.

The process of approximating a series of continuous reading points to a line segment is generally done by first determining the maximum approximation error ε _nax , and then adjusting the parallel lines so that the continuous reading point sequence can be accommodated as much as possible within a parallel line with a constant width determined by this error value. One possible method is to find the position and direction and determine an approximate line segment from the center line.

Here, A is a set of input point sequence data (output of contour extraction processing), and B is an approximate line segment (a set of points on a line segment).
Expressed as Havsdorff-Euclidean Distance between A and B, H(A, B)
is the Euclidean distance between points p ₁ and p ₂
Distance) is ‖p ₁ - p ₂ ‖, then H (A, B) = max [max min‖p ₁ - p ₂ ‖, p ₁ B
p ₂ A max min‖p ₁ −p ₂ ‖]p ₁ A p ₂ B.

Therefore, the line segmentation process described above means performing line segment approximation so that H(A, B)≦ε _nax is satisfied for the maximum (specified) approximation error ε _nax , and the error This is a line segment approximation method that minimizes the cumulative value of squares.

In reality, in view of the amount of calculation, the approximation process described above is not performed faithfully, but instead a method of searching for the end point of the approximate line segment from the input point sequence (as shown in Figure 12, which will be described later), or An approximation method is used in which calculation is performed using the Hausdorff distance, for example, by substituting an error only in the y-axis direction. Another possible method is to find the approximate line segment by searching for the shortest path within a band-shaped area formed by a contour pixel array (one pixel is a square area).

Next, the line segmentation process will be specifically explained based on the line segmentation process flowchart of FIG. 12 and the line segmentation process explanatory diagram of FIG. 13.

One of the outline point sequences is determined as the starting point of the approximation process, and this point is registered as the first approximate line segment vertex (401).

Then, as shown in Figure 12b, the next contour point is
Assuming a circle with radius ε centered on p ₁ , upper tangent angle TA2 and lower tangent angle TA1 are determined from the starting point (402). Similarly, as shown in Figure 13b, draw a circle with radius ε centering on the contour points to be connected one after another, draw connections, and calculate the maximum value of TA1 among them.
Set MAXTA1 and the minimum TA2 value to MINTA2 (403).

If MAXTA1>MINTA2 holds true, it is determined that one line segment approximation cannot be performed any more (405), and the EPN-th point is determined and registered as the end point of the line segment (406).

To identify candidate points that can be the end point, after drawing a tangent and updating MAXTA1 and MINTA2, it is recognized whether the CA of each point is sandwiched between MAXTA1 and MINTA2, and if it is, the candidate point is (404) If the EPN-th end point is not the last data, the line segment approximation process is continued using that point again as the starting point of line segmentation (407).

In this way, the outline of the binary image as shown in Fig. 13a, after passing through the steps b and c in Fig. 13, is
The contour can be divided into predetermined lines as shown in FIG.

Through the above-mentioned series of processes, the area to be recognized is extracted from the input image, the area information is expressed as a line segmentation suitable for subsequent recognition processing, and the data is output as polygonal vertex coordinate series data. do. This data is stored in the aforementioned line segmentation information memory 108, and checked against a dictionary pattern previously created using a similar procedure to determine if it matches.

This matching process is performed after the two patterns (polygonal figures) are positioned by matching their respective characteristic features (centroid coordinates, principal axis of inertia direction, etc.).

The area (pattern) feature extraction process involves the above-mentioned area feature extraction unit 110 and area feature memory 111, and takes into account that the outline of the area is segmented into line segments, and extracts the area from both end points of the line segment to the coordinate axes. A trapezoid formed by the drawn perpendicular line and the coordinate axes is used as a processing unit, and the processing procedure shown in the region feature extraction processing flowchart of FIG. 14 is performed.

The basic calculation formula is as shown below.

The vertex coordinate string of the approximate line segment is (X ₁ , Y ₁ ), (X ₂ ,
Y ₂ ), ..., (X, Y), and the vertex coordinates are arranged in a closed loop on the plane,
Let (X ₁ , Y ₁ )=(X, Y).

(1) Area of line differentiation model A = _b 〓 ^I=2 1/2 (X _I −X _I-1 ) × (Y _I-1 + Y _I ) (2) First moment M _1X = _b 〓 〓 ^{I= 2} 1/6 (X ₁ −X _I-1 ) × (Y _I-1 + Y _I ) × {Y _I-1 + Y _I ² /
(Y _I-1 +Y _I )} M _1Y = _b 〓 〓 ^I=2 1/6 (X _I-1 +X _I ) × (Y _I-1 −Y _I ) × {X _I-1 +X _I ² /
(X _I-1 +X _I )} (3) Second moment M _2X = _b 〓 〓 ^I=2 1/12(X _I −X _I-1 )×(Y _I-1 +Y _I )×(Y _I ² _-1 +Y _I ²
) M _2Y = _b 〓 〓 ^I=2 1/12(Y _I-1 −Y _I )×(X _I-1 +X _I )×(X _I ² _-1 +X _I ²
) (4) _Synergistic moment M _XY D _YI = _{X I-1 + 1/3 (} X _{I −} ) [When X _I-1 < _{X I} ] D _YI = X _I + 1/3 (X _I-1 − _{X I} ) + 1/3 ₍ X _{I-1 − -1} + _{Y I )} [When ^X _{I -1 X I} ] _Then , _M Y _I-1 + Y _I ² /
(Y _I-1 + Y _I ) _} ×D _YI (5) _X coordinate XO, Y coordinate YO of centroid Secondary moment and multiplicative moment in the system M ₂₀ = M _2Y −M _1Y ×M _1Y /A M ₀₂ = M _2X −M _1X ×M _1X /A M ₁₁ = M _XY −M _1X ×M _1Y /A (7) Spindle tilt angle ANG=1/2×tan ^-1 {2×M ₁₁ /(M ₂₀ −M ₀₂ )} (8) Secondary moment with respect to the spindle M _02Z = M ₂₀ + M ₀₂ −M _20Z The major axis and minor axis of the main axis are determined by calculating M _20Z and M _02Z .

Finally, the input pattern whose centroid and principal axes have been obtained by line segmentation is used to match the centroid and principal axes with the already registered dictionary patterns, as shown in the pattern matching explanatory diagram in Figure 15. , coordinate transformation is performed on the input pattern side, matching is performed, and an area where both patterns overlap is extracted as a similar line segmented figure. a is image input (binarization), b is segmentation contour extraction, c is line segment approximation, centroid, principal axis determination, d is dictionary pattern group, e is current dictionary pattern, f is centroid, principal axis alignment shows.

If both patterns are obtained from the same part under the same imaging conditions, the area of the overlapping region determined by matching is the same as that when the object is represented on a digital image partitioned into a mesh. It is expected that the area will match the area of the dictionary pattern (input pattern) within an error range that takes into account errors and errors due to line segment approximation. When this area ratio exceeds a predetermined value (for example, 95%), it is determined that the outline (pattern) of the target part is present.

These processes are carried out by the common area extraction unit 11 described above.
2, the common area memory 113, the recognition processing unit 114, the line segmentation information memory 108, and the dictionary pattern memory 109 are involved.

FIG. 16 is a flowchart of line segmentation pattern matching.

Here, first, the outline of the area (inner contour,
Compare the input pattern, which is a linear representation of the input pattern (including both the outer contour), with one of the already registered dictionary patterns, and calculate the x-coordinate direction, y-coordinate direction between the centroids of both patterns.
Calculate the distance in the coordinate direction (801).

Next, all vertices of the input pattern are translated by the above distance so that the centroid of the input pattern matches the centroid of the dictionary pattern (802).

Also, select the major or minor axes of the principal axes of inertia of both patterns, determine the principal axes that reach the outer periphery at approximately the same distance from the centroid as corresponding principal axes, and calculate the angle between the principal axes of these two corresponding patterns. (803).

Furthermore, all the vertices of the input pattern are rotated by the above angle with the centroid as the rotation center, so that the principal axes of both patterns coincide (804), and the common area of both patterns, that is, the two polygons, is Find the product figure of the rectangles (which is also a polygon) (805).

Once the product figure is determined in this way, its area is calculated and the ratio to the area of the dictionary pattern (or input pattern) that has already been determined is determined (806).

Compare the area ratio with a predetermined judgment threshold (807),
If the area ratio shows a value equal to or higher than the threshold value, it is determined that both patterns match and a message indicating that recognition has been completed is displayed.In addition, the input calculated from the type of dictionary pattern that matched and the relative position of the dictionary pattern is displayed. Outputs pattern position and orientation information (808).
If the area ratio is smaller than the determination threshold, it is determined whether the next dictionary pattern exists (909).

If the matching process with all dictionary patterns is completed and there is no next dictionary pattern, a message indicating that recognition is impossible is output and the process ends (810). If you have the following dictionary pattern,
Select that pattern (811) and repeat the matching process from the beginning.

Next, an example of a processing method for creating a product figure (common area) of a line segmented pattern and a dictionary pattern will be described below.

In order for the point P(X, Y) on the two-dimensional plane to exist on the right half plane of the line segment P ₁ P ₂ , the following equation must be positive. That is, P ₁ = (X ₁ , Y ₁ )P ₂
= (X ₂ , Y ₂ ), then ₁₂ (P) = -X Y 1 X ₁ Y ₁ 1 X ₂ Y ₂ 1 Here, the following function is defined.

g ₁₂ (P) = 1 [ ₁₂ (P) > 0] 0 [ ₁₂ (P) = 0] -1 [ ₁₂ (P) < 0] According to this function, the line segments P ₁ P ₂ , P ₃ P ₄ The condition for intersection is expressed by the following equation.

g ₁₂ (P ₃ ) = −g ₁₂ (P ₄ ) g ₃₄ (P ₁ ) = −g ₃₄ (P ₂ ) Furthermore, in order to check the presence or absence of intersection points as a whole, consider polygons F and F′, and Find the VS matrix and SV matrix that indicate the left/right determination of the opponent's vertex with respect to the minute.

VS (i, j) = g _i , _i+1 (P _j ) SV (i, j) = g _j , _j+1 (P _i ) Using this relationship, the intersection between line segments is determined by the following equation. conduct.

C _ij = | {SV (i, j) − SV (i, j + 1)} × {VS (i + 1, j) − VS (i,
j)}| The judgment conditions are as follows.

C _ij =4: [P _i P _i+1 ] and [P _j P _j+1 ] intersect.

2: The vertex overlaps the line segment.

1: The vertices overlap.

0: Do not intersect.

Based on the above judgment formula, the presence or absence of a vertex of the product figure is checked, and if its existence is confirmed, it is found by calculating the intersection of binary lines. The obtained product figure is also represented by a series of vertex coordinates, similar to the line segmentation model, and the area of this polygon is calculated and the degree of pattern matching is evaluated using the following equation.

Pattern matching rate = area of product figure / area of standard pattern × 100 [%] As described above, the above embodiment does not require large-scale dedicated hardware to perform recognition with the same accuracy, and only the vertices can be used. Since data is processed, high-speed processing is possible.

Note that since an approximation process called line segmentation is performed, it is necessary to confirm the effect on recognition accuracy, but as a result of recognition experiments, the positional accuracy of the center of gravity is approximately 0.2 pixels, and the calculation accuracy of the principal axis of inertia is approximately Good results have been obtained at around 0.015 radian.

In this way, when assembling the part using the visual recognition device in the feeder, if the part is determined to be suitable, the robot 12 grabs it and assembles it, and if it is determined to be an unsuitable part, the feeder is activated. The parts return to the feeder and one of them is determined to be a suitable part and sent out. Therefore, if several types of parts are placed in the feeder, even if they are delivered in different assembly orders, they will be delivered one after another, so the robot can assemble the parts in the assembly order. becomes possible.

〔Effect of the invention〕

As described in detail above, according to the present invention, the terminal end of the conveyance path is opened inward of the storage section to circulate the parts, and the circulating conveyed parts are detected by the detection section, and the detection section detects the circulating components by using the signal from the detection section. Since the judgment section determines suitable and unsuitable parts and takes out only suitable parts, there is no need for a special device on the conveyance path to separate suitable and unsuitable parts.
In addition, since the end of the conveyance path is opened inside the storage and parts are circulated, there is no need for the conventional parts position restraining member, and as a result, it can be applied to all kinds of parts, and the parts feeding device The versatility of has been expanded, and it has become possible to obtain a parts supply device that is immediately compatible with high-mix, low-volume production systems.

Furthermore, even if multiple types of parts are mixed into the device of the present invention, these parts will form a jumbled line and circulate through the conveyance path, so the judgment section will find necessary and suitable parts from this part line. For example, in the assembly station of an automatic assembly line that assembles multiple types of parts at one station, conventionally, as many parts supply devices as there are parts were required, but with one part supply device, parts can be taken out sequentially. The effects are significant, such as the ability to supply these parts using equipment and the simplification of automatic parts assembly lines.

[Brief explanation of the drawing]

1 and 2 are perspective views showing a conventional component supply device. 3 to 6 show an embodiment of the present invention, in which FIG. 3 is a schematic diagram showing the overall configuration of the component supply device, and FIG. 4 is a perspective view showing details of the ball feeder. FIGS. 5a and 5b are partially enlarged perspective views of other embodiments of the conveyance path end portion, and FIG. 6a is another embodiment,
FIG. 6b is a cross-sectional view of the rotary parts feeder; FIG. 7 is a schematic diagram showing the configuration of the robot control device shown in FIGS. 3 and 6; FIG. FIG. 9 is a block diagram of an embodiment of the pattern recognition device according to the present invention, and FIG.
Figure 0 is the preprocessing flowchart, Figure 11 is the format of the area information, and Figure 12 is the
Flowchart of same-line differentiation processing, FIG. 13 is an explanatory diagram of the same-line differentiation process, FIG. 14 is a flowchart of same-area feature extraction processing, FIG. 15 is an illustration of same-line pattern matching, and FIG. 16 is an illustration of same-line differentiation process. This is a flowchart of differentiation pattern matching processing. 11... Bowl feeder, 12... Robot,
13...Gripping device (hand mechanism), 14...TV camera, 16...Assembling station, 17...Storage section, 18...Transportation path, 19...Terminal section.

Claims (1)

[Claims] 1. Consisting of a storage section for storing parts and a conveyance path for transporting the parts stored in the storage section in a row,
Images are taken of the bowl feeder with the end of the conveyance path opening inward into the storage section and the parts conveyed to the conveyance path, and it is recognized whether or not the parts are suitable for assembly, as well as the position and direction of the suitable parts. and a visual recognition means to take out the suitable part recognized by the visual recognition means by correcting the take-out position data of the robot based on the position and direction data of the suitable part, and take it out from the target installed at another location. and a robot that assembles the parts to be assembled, the robot takes out only suitable parts suitable for assembly from the parts conveyed by the conveyance path of the bowl feeder, and the bowl feeder removes unsuitable parts to the end of the conveyance path. is guided into the storage section through the opening, placed on the conveyance path again,
A parts supply and assembly device characterized in that it is configured to recirculate. 2. The parts supply and assembly apparatus according to claim 1, wherein the robot stops transporting the parts in the bowl feeder when the robot lifts the parts. 3. The parts supply and assembly apparatus according to claim 2, wherein the robot has at least three degrees of freedom. 4. A patent claim characterized in that the above-mentioned visual recognition device is configured to segment a pattern for each part and determine the center of gravity of the segmented pattern for each part and the direction of the principal axis of moment. The parts supply and assembly device according to item 1. 5. The parts supply and assembly apparatus according to claim 1, wherein the bowl feeder is a rotating parts feeder.