JPS62168287A - Calibrating method for vision sensor - Google Patents

Abstract

PURPOSE:To execute a calibration in a short time by image-picking up a matrix installed in parallel on a robot coordinate, teaching a point on its image and using it as a point on a vision coordinate. CONSTITUTION:The titled sensor is provided with a control part 8 to which a keyboard 9 can be connected as an external input, a graphic RAM 10 for displaying a digital image, an A/D 5 for A/D-converting an input from a TV camera 1, a video RAM 6 for storing an output from the A/D 5, and a composite video circuit 20 which can superposed and display an analog image from the TV camera 1 and the digital image of the graphic RAM 10, and an optical position on the analog image can be set as a coordinate on the digital image by the keyboard 9. In this way a cross cursor is set in an optical position on an image from the TV camera 1, and its coordinate can be calculated by the control part 8.

Description

DETAILED DESCRIPTION OF THE INVENTION (Industrial Application Field) The present invention relates to the position measurement function of a visual sensor as a visual function of an industrial robot. J! The present invention relates to a method of calibrating a visual sensor for converting coordinates to visual sensor coordinates.

(Prior technology) In order to perform position calculation, which is one of the most important elements required for the visual function of industrial robots, the relationship between the coordinate system of the visual sensor and the coordinate system of the robot must be defined, and both coordinate systems must be It is necessary to perform coordinate transformation between. As an example of how to achieve this, it can be easily calculated by determining the position and angle of the optical axis of the camera. However, determining the position and angle of the camera's light, 1!10, was difficult and time-consuming, as it required measurements using special measuring equipment.

In other words, as a conventional method, if the angle of the visual coordinates with respect to the robot coordinates VC, that is, the tilt angle and height of the camera, the distance to the imaging plane, etc., is known, the visual coordinate force can be converted to the robot coordinate, or the robot coordinate force can be converted from the visual coordinate force to the robot coordinate. It was possible to convert the coordinates to visual coordinates. However, this required precise measurement of the camera angle (optical axis angle), which required advanced technology and time. For example, in Japanese Patent Application Laid-Open No. 60-37007, conversion parameters are calculated by measuring the amount of movement in visual coordinates when a robot moves a certain amount multiple times. (Problems to be Solved by the Invention) The present invention provides a method for calibrating a visual sensor that can perform all calibrations in a short time and with simple operations. It is in.

Another object of the present invention is to achieve this by simple teaching on a monitor using an image projected from a TV left camera.
Another object of the present invention is to provide a visual sensor calibration method that can perform calculations with high accuracy.

(Means for Solving the Problems) Therefore, the present invention provides a control unit (CPU) to which a keyboard can be connected as an external input, and a graphic RAM for displaying digital images.

Video R that A/D converts the input sound from the TV left camera and stores it.
It has a composite video signal forming circuit that enables the combined display of analog images from the AM and TV left cameras and digital images from the graphic RAM, and any position on the analog image can be set as coordinates on the digital image using the keyboard. In a visual sensor that can
Calibration is performed by taking an image of a matrix placed parallel to the robot coordinates, teaching the points on the image, treating them as points on the visual coordinates, and calculating an approximation of the relationship between the visual coordinates and the robot coordinates based on the configuration of the points. This is a visual sensor calibration method that simulates the following.

(Embodiment 91) To explain the embodiment of the present invention with reference to the drawings, the first embodiment
The figure shows all of the visual sensors used to realize the invention - Keyboard (9) as an external input.
A control unit (CPU) (81) that can be connected to the computer and a graphic RA for displaying digital images.
MαQ, A/D (5) for A/D converting the input from the TV camera (1), a video RAM (6) for storing the output from the A/D (5), and a video RAM (6) for storing the output from the TV left camera (1). Analog image and graphic RAM (1 [J) Composite video signal circuit that enables superimposed display of digital images + 21 bits, and any position on the analog image can be set as coordinates on the digital image using the keyboard (9) It is a visual sensor.This hardware allows for 512 x 512 digital images and TV
Overlay the analog and image from left camera 1) and CR
Tα can be displayed, and the control unit (cpu)
(8) (C) It is possible to fully control the external input (keyboard) and output the cross cursor at any position on the digital screen, and by this, the cross cursor can be set at any position on the image of the TV left camera etc. The calibration method of the present invention will be described below.In general, industrial robots and visual sensors have their own coordinate systems (Figure 2).
Prepare the IXI matrix 0Q gold as shown in Figure 3, and then
The TV camera (1) is installed on the imaging surface parallel to the robot coordinates (Figure 4). When all of these images are captured by the TV camera (1), the result will be as shown in Fig. 5. Next, by operating the keyboard on this image, all points corresponding to FIG. 6 are taught using the cursor. Two straight lines are formed by this point 11 point 4 and point 21 point 3, point l.

The two straight lines formed by point 2 and point 49 and point 3 are each considered to be a reference axis expressing the robot coordinate axis.

Expressing each reference axis as a straight line on all visual coordinates, Y = a,
Reference qqHY = aj x + b3 (point 1 1 point 4) Reference axis 3Y = al x + b4 (point 2
One point 3) Reference axis 4 (al ”'-a4 + bl
~b4 is a constant) Here, if the reference axis 1.2 is the Y@ of the robot and the reference axis 3.4 is parallel to the X axis of the robot coordinates, each reference axis is on the entire robot coordinates. When expressed as a straight line, X=A reference axis I X=B reference axis 2 Y=C reference axis 3 Y=D reference axis 4 (A, B, C, and D are constants). This is based on the assumption that the matrix is placed parallel to the robot coordinates.

Now, the point (X++7+) and the point (Xw +
7z) as a value on the robot coordinates is as follows.

In order to calculate the difference Δxl in the X direction between two points on the robot coordinates, the reference axis 1.corresponds to X1IIJ on the robot coordinates.
Use 2. Here, this reference axis can be approximated by changing the inclination linearly on the visual coordinates, then the reference axis passing through the point (Xl, yl) is defined as Y'' from the reference axis 1.2.
a5x ten b5 (a5 + bs is a constant) and table =
It can be expressed as a6 X+1)a (aa + b6 is a constant) (Figure 7).

Therefore, according to the approximation above, △X is It can be expressed as becomes.

Next, the difference △Y in the Y direction between the two points on the robot coordinates is determined by the reference axis f, which passes through the point (xt, yt) in the same way from the reference axis 3.4.
Y = a7 xtbq (a + b7 is a constant),
Reference axis 'rY w aax+b passing through point (xt, yt) k
8 (aB and bs are constants) (Figure 8).

Therefore, similarly to △X becomes.

With these ΔX and ΔY, the difference between two points on the visual coordinates can be treated as a difference on the robot coordinates.

In the present invention, approximations are made on the assumption that this reference axis changes linearly. Therefore, if the parallelism between the two coordinate systems is extremely poor (the rotation is large), the calculation error may become large. In that case, the sample matrix should be
As shown in the figure, one with a large number of criteria such as 2×2.3×3 may be selected.

(Effect of the invention) With the main BJ4, the task of measuring the angle of the camera and the distance to the imaging surface as in the past is no longer necessary, so the visual sensor can be taught more easily and in a shorter time. became. However, since calibration is the most important element that affects measurement results, if the teaching method is difficult and complicated, there is a high possibility that teaching errors will occur and errors will occur in the measurement results. According to this method, the possibility of teaching errors is low and the calculation accuracy is high, so the method provides a visual sensor calibration method with very high reliability of measurement results.

[Brief explanation of drawings]

Fig. 1 is a block diagram showing the hardware of a visual sensor that implements the entire vision sensor method according to the present invention, and Figs. 2 to 9 are explanatory diagrams of the embodiment method of the present invention. Show. 1...TV left camera 5...A/D6...Video RAM 8...Control unit (cpU)9...
Keyboard 10... Graphic RAM 15.
...CRT 20...Composite video signal forming circuit agent Patent attorney Jun Kawauchi 2ill; Cao 1;11 2nd paper 3 Figure 4 Figure 16 Original 72, 7e 17 Camera drawing Figure 5 19 Figure procedure correction sergeant (self-motivated) February 24, 1985 Director General of the Metropolitan Police Department Michibu Uga 1, Indication of the incident Patent Application No. 98c2 of 1988 2, Name of the invention Visual sensor calibration method 3, Make amendments Relationship with patent case Patent applicant name (
519) Fuji Koshi 4 Co., Ltd. Residence of agent: 251 World Trade Center Building, 2-4-1 Hamamatsucho, Minato-ku, Tokyo Details column for detailed explanation of Bara's invention, drawing 7, contents of amendment (71) +1411, page 7, line 1, 1 coordinate table” is corrected to “coordinates”. (72)Amend Figure 1 of the drawings as per the attached sheet. Tow 1 diagram

Claims (1)

[Claims] A control unit (CPU) to which a keyboard can be connected as an external input, a graphic RAM for displaying digital images, a video RAM that A/D converts input from the TV camera and stores it, and analog images and graphics from the TV camera. A visual sensor is equipped with a composite video signal forming circuit that enables overlapping display of digital images in RAM, and is capable of setting the position of a position on an analog image as coordinates on the digital image using a keyboard. Calibration is performed by capturing an image of a matrix placed parallel to the image, teaching points on the image, treating them as points on the visual coordinates, and approximately calculating the relationship between the visual coordinates and the robot coordinates based on the configuration of the points. A method for calibrating a visual sensor, characterized by: