JPH0212710B2 - - Google Patents

Description

[Detailed Description of the Invention] [Technical Field of the Invention] The present invention relates to a visual sensor for a robot, and more specifically, it identifies the shape of a target object, that is, a workpiece, measures the position, inclination, etc. of the identified workpiece, and
The present invention relates to a visual sensor for a robot that has a function of transmitting information such as a grasping position and an assembly position to a robot.

[Prior art]

Conventionally, in so-called factory automation, where robots perform assembly work, transportation work, etc., in order for robots to perform work at high speed and accuracy, it is necessary to use high-precision, high-performance visual sensors that serve as the robot's eyes. was necessary and its realization was desired.

[Summary of the invention]

This invention was made in view of the above points, and is a method for extracting the overall shape characteristics of one or two or more separate workpieces within the field of view from image data received by a camera. , overall identification means for identifying which of the registered workpieces the workpiece within the field of view corresponds to by searching data regarding the overall shape of multiple types of workpieces registered in advance; and identification by the overall identification means. an overall measuring means for transmitting information regarding the position and inclination of the entire workpiece to the robot by measuring the coordinates of two predetermined points on the workpiece;
For the desired partial region of the workpiece identified by the overall identification means, extract local features based on data regarding the local shape taught in advance to determine if there is a shape that matches the taught local shape. a partial identification means for identifying whether or not the local shape is identified by the partial identification means, and a part for transmitting information regarding the local center of gravity position and inclination to the robot by measuring the coordinates of a predetermined point regarding the local shape identified by the partial identification means. By equipping the robot with measurement means, we can dramatically expand the capabilities of robots.
The object of the present invention is to provide a visual sensor for a robot that can add a high degree of adaptability to each task. Hereinafter, teaching in advance data regarding the overall or partial shape of the workpiece will be referred to as teaching.

[Embodiments of the invention]

An embodiment of the present invention will be described below based on the drawings.

FIG. 1 is a system configuration diagram showing an embodiment of the present invention. In FIG. 1, 1a and 1b are television cameras, 2 is an image input device, and the television camera 1
The input image signals from a and 1b are binarized, and processing such as arc point removal is performed. Incidentally, the erect point removal is a process of removing an erect point by determining that the information of one erect point is the same as the information of the pixels surrounding that point. 3 is an image processing device that generates a run length from the output of the image input device 2; 4
is an image memory that stores the entire image, which in this embodiment is 256×256 bits. In other words, the number of pixels is 256×256. 5 is a central processing unit (CPU), 6 is a storage device, 7 is a GPIB interface, 8 is an RS232C interface, 9 is a monitor interface, 10 is a light pen interface, 11 is a visual sensor controller, and 5 to 6 are the visual sensor controller. 10. 12 is a robot controller, 13 is a cassette magnetic tape device, 14 is a monitor using a television tube display device, 1
A light pen 5 is used to perform various sensor operations such as menu selection and teaching.

The monitor 14 displays the CPU based on the input information.
The results processed by step 5 are displayed. The cassette magnetic tape device 13 is used for long-term storage of teaching data.

First, overall identification of the target object, that is, the workpiece, will be explained. This is done by extracting global features. In other words, by showing the target workpieces in advance, the overall characteristics of multiple workpieces can be determined such as the workpiece area, circumference length, outer circumference length, number of holes, etc.
Calculate the total circumference, that is, the total circumference including the circumference of the hole if there is one, the area of the hole, the maximum and minimum length from the center of gravity to the outer circumference, compactness, etc., and then calculate the A indicated by the light pen. , calculate the coordinates of point B2 based on the secondary moment specific to the workpiece, and calculate 1
Each work from n to n is memorized. Note that compactness is expressed as (perimeter) ² /area, and the more complex the shape, the larger this value becomes.

Next, when the work actually flows or when the robot goes to see the work, camera 1
The above-mentioned overall characteristics are calculated using the data input from a and 1b, and the number of the work is determined by searching from the data registered in advance.

Next, measure the position and inclination of the workpiece. This calculates the coordinates of two predetermined points A and B on the workpiece based on the workpiece's inherent second-order moment, and calculates the angle or inclination that the vector from point A to point B makes with the x-axis. Send the coordinates and the above angle to the robot.

That is, the data sent to the robot is the work number based on the above-mentioned overall characteristics indicating what the work is, the coordinates of point A, and the inclination. For example, when a robot grasps a workpiece, it can move from point A to
By specifying whether to grab the workpiece at mm, the robot can grasp the workpiece by tilting its hand because the tilt is known. If there are multiple workpieces within the field of view, scan horizontally from the upper left corner, that is, scan horizontally from the upper left corner of the monitor 14 screen, and measure the first workpiece that hits, for example, if it is unregistered. If the work is registered, it will send an "unregistered error" to the robot, measure the next work, and if the work is already registered, send the work number and A.
The coordinates of the point and the angle formed by the vector from point A to point B with the X axis are calculated and sent to the robot. Furthermore, if there is a workpiece, the workpiece is subsequently measured and the data is sent to the robot. Then, the robot selects the workpiece to be gripped based on the workpiece number transmitted by the visual sensor.

Note that the sensor side and the robot side have separate coordinates, so there is an offset between the two coordinates.
This offset is taught to the robot in advance through calibration, and the sensor measurement results are corrected on the robot side. Furthermore, the sensor can now measure the number of millimeters corresponding to one pixel by scaling.

FIG. 2 is an explanatory diagram for explaining the measurement of the position and inclination of the workpiece, where 20 indicates the workpiece, G indicates the position of the center of gravity, and the arrow indicates the minimum direction of the secondary moment specific to the workpiece.

Next, partial identification of a workpiece will be explained. FIG. 3 is an explanatory diagram for explaining local feature extraction, and FIG. 4 is an explanatory diagram showing an example of an identification figure. Figure 3,
In FIG. 4, 21 is a monitor screen, and 22 is a window, which can be arbitrarily set with a light pen, and indicates that information is to be processed only within this window, that is, indicates a processing range. 23 is a hole drilled in the workpiece. This part identification is used when inserting something into the hole 23 or inserting a screw.

When specifying to find the hole 23 within the window 22, a circle with the specified radius within the window 22 is found based on the taught data, and the coordinates of its center of gravity are measured.

For the ellipse 24, by specifying the major axis _Ll and the minor axis _Ls , the angle or inclination of the arrow 25 connecting the specified ellipse's center of gravity G and the longest point with the x-axis is measured. .

As for the rectangle 26, by specifying the long side _Ll and the short side _Ls , the center of gravity G and the inclination are measured in the same way as for the ellipse.

As for the groove 27, by assuming a square shape and specifying the width W and depth L, the center of gravity G of the specified groove can be determined.
The slope is the arrow 25 in the outward direction from the center of gravity G.
It is determined by the angle between a and the x-axis.

For the corner 28, by specifying the lengths L ₁ and L ₂ , the corner LC, and the angle θ, the position of the center of gravity G of the specified corner is determined, and the inclination is determined by the arrow 25b indicating the half angle of the corner and the x-axis. Determined by the angle formed.

As described above, the part that matches the command is identified, and its center of gravity position and inclination are transmitted to the robot. If there is no match, an error will be returned.

The two cameras 1a and 1b are used by arbitrarily switching between them according to commands from the robot controller 12. For example, one unit performs overall identification and the other unit performs local identification. However, one machine cannot do both at the same time.

〔Effect of the invention〕

As described above, according to the present invention, by providing the whole recognition function and the partial recognition function, a highly accurate and highly functional visual sensor can be obtained, dramatically expanding the capabilities of robots, and improving the robot's ability to perform various tasks. A high degree of adaptability can be added.

[Brief explanation of drawings]

FIG. 1 is a system configuration diagram showing an embodiment of the present invention, FIG. 2 is an explanatory diagram for explaining the measurement of the position and inclination of a workpiece in overall identification, and FIG. 3 is an explanatory diagram for explaining local feature extraction. Explanatory diagram, FIG. 4 is an explanatory diagram showing an example of an identification figure. In the figure, 1a and 1b are cameras, 2 is an image input device, 3 is an image processing device, 4 is an image memory, and 5 is a
A CPU, 6 a storage device, 12 a robot controller, 13 a cassette magnetic tape device, 14 a monitor, and 15 a light pen. In addition, in the figures, the same reference numerals indicate the same or corresponding parts.

Claims (1)

[Claims] 1. Extract the overall features of the shape of one workpiece or two or more separately existing workpieces within the field of view from the image data received by the camera, and Overall identification means for identifying which of the registered works the workpiece within the field of view corresponds to by searching data regarding the overall shape of the workpiece; An overall measurement means that transmits information regarding the position and inclination of the entire workpiece to the robot by measuring the coordinates of two points; and a local shape taught in advance for a desired partial area of the workpiece identified by the overall identification means. a partial identification means for identifying whether or not a shape matching the taught local shape exists by extracting local features based on data regarding the local shape; and regarding the local shape identified by the partial identification means; 1. A visual sensor for a robot, comprising: partial measurement means for transmitting information regarding a local center of gravity position and inclination to the robot by measuring the coordinates of a predetermined point.