JPS62289705A - Visual system for finding three-dimensional inclination of body plane - Google Patents

Abstract

PURPOSE:To improve programming accuracy by calculating three-dimensional coordinate values of respective points from a two-dimensional image obtained by a visual sensor by a light cutting method, and projecting a light and shade image obtained from its height information on the axes of an ellipse which is inertially equivalent to it. CONSTITUTION:The image of a part which contains a light cutting line when an object body (not shown in figure) is irradiated with slit light is picked up by a television camera (visual sensor 51) to obtain two-dimensional image values. Three-dimensional coordinate values of respective points are calculated from the coordinate values by using a converting means 53 and light and shade information on the two-dimensional image is calculated from its height information. Then, an approximate projecting means 55 approximates the distribution of the light and shade information to an ellipse which is inertially equivalent to it to project the light and shade information on its major and minor axes. Then, an arithmetic means 57 finds the inclination of the body from the projection value. Then, the three-dimensional inclination of the body is accurately obtained.

Description

[Detailed Description of the Invention] 3. Detailed Description of the Invention [Technical Field of the Invention] This invention is based on a method of cutting light obtained from a visual senna.
The dimensional image is processed by a computer, etc., and the original 3-dimensional image is
It relates to a visual system that determines the three-dimensional inclination of a certain object plane included in a dimensional world (this is called a "scene").

[Technical background of the invention and its problems]

When using robots to automate FjU'A processing and assembly work currently performed by humans, high I
q visual sensor system is required.

For example, let us assume that we perform some kind of work with the tip of a robot perpendicularly facing the plane of a certain object.

Here, it is assumed that the object plane is not precisely positioned by other means, but that the three-dimensional orientation of the object plane is not constant. In this case, in order to apply the robot tip to the object plane at I i , the three-dimensional inclination of the object plane must be known and the robot must be moved accordingly. The visual system uses a television camera to reflect the scene in which the object plane is placed, processes the captured two-dimensional image with a computer, etc. to understand the scene, and calculates the three-dimensional inclination of the opposing object plane. It is a system that plans the following.

Visual systems suitable for such uses are currently being actively researched (for example, Jimon, April 1982-1689).

In this type of visual system that is conventionally known, the above-mentioned slope is calculated using the following processing method.

A plane that passes through the point (X (1, Y (1, Zg)) of the scene coordinate system and whose normal vector is the unit vector 〒-<A, B, C) is given by the following equation (2).

A (X-X g) +B (Y-Yg
10C (Z-Zq) =○ ・・・■
A2 +B2 +C2=1 From this formula (■), the following formula (■) is derived.

...■ Information on n points on the object plane (X+
, Y+, Z+) ~ (Xn, Yn, Zn
) and substitute them into equation (■) to obtain n simultaneous equations. After finding A, B, and C by applying the least squares method to these n simultaneous equations, add the small square equation to AB obtained from formula (■) and the equation by eliminating △ from formula (■) as above. BC obtained by applying multiplication and the condition A>O are given to obtain IAl, IBl, and IcI.

These conventional processing methods require a huge amount of information for calculation, so they are not suitable for use in situations where high speed is strongly required, such as the robot vision system mentioned above. There wasn't.

[Purpose of the invention]

This invention was made in view of the above-mentioned conventional problems, and its purpose is to provide a visual system that can quickly determine the three-dimensional inclination of an object plane through simple information processing. .

[Summary of the invention]

Therefore, in this invention, as shown in FIG.
Process the two-dimensional image using the optical cutting method obtained from IQ from 1,
A visual system for determining the three-dimensional inclination of a certain object plane included in a scene, comprising a converting means 53 for converting height information in the scene of the object plane into gradation information of a two-dimensional image, and a conversion means 53 for converting height information in the scene of the object plane into gradation information of a two-dimensional image; Approximate projection means 55 approximates the distribution of information by an ellipse that is inertially equivalent to the information distribution, and projects the shading information onto the long sleeve and short axis of this ellipse, and from this projection value, calculates the three-dimensional A visual system comprising: a tilt calculation means 57 for calculating a tilt angle;

[Example: - Fig. 2 shows the hardware configuration of a visual system according to an embodiment of the present invention, Fig. 3 shows the main part of its software configuration, and Fig. 4 is an explanatory diagram of the processing process.

As shown in FIG. 2, a slit light 2 is emitted from a projector 1 using a laser or the like, and the slit light 2 is reflected by a -r=J moving mirror 3 and irradiated onto a target object 4. Target object 4
On the surface of
occurs.

As information about the target object 4, a portion including the optical cutting line 2a is imaged with a television camera 5 installed. The obtained image signal is
The image signal is written into the image memory 7 from the image signal acquisition section 6 under the υ control of the calculation section 9 . The movable mirror 3 is driven by the suspension control section 11 under the control of the calculation section 9, and is adapted to change the irradiation position of the slit light 2 on the target object 4. In addition to the image memory 7 for storing captured images, the image memory includes an image memory 8 for storing a screen that is currently being processed. The calculation results and the contents of the image memory can be monitored on the display unit 10 at any time.

The slit light image 2a observed by the camera 5 has a step corresponding to the height of the target object 4. observed (X, Y)
The size of the difference in point is converted to the actual height of the surface of the target object based on the principle of three-dimensional distance measurement using the optical cutting method, and an image with rough shading information and 8-bit dimensions is created. In the case of the memory, the information is converted into (0 to 255 information), and the image memory 7 is converted to (x.

Y) position (step 202 in FIG. 3).The angle of the movable mirror 2 is changed little by little, and the
2 has been applied to all pixels in the image memory 7, in step 205, the observation plane of the target object 4, that is, the object plane (in this example, since the slit light is irradiated from directly above, The coordinates (XG, Ya) of the center of gravity of the target object 4 (the upper surface of which is observed) are calculated. Therefore, the image stored in the image memory 7 is converted into black and white binarized using a certain threshold T and stored in the image memory 8 (the part where the object is present is represented by white and the background part is represented by black, or vice versa). Based on the contents of the image memory 8, the arithmetic unit calculates the barycenter coordinates (Xa, YG) using the following formula.

Here, fB(X, Y) is (X) of image memory B8.

Y) is the content of point f B (X, Y) = Oor
It is 1.

Next, in step 206, the yaw angle θ of the object plane is calculated.

That is, . . . rA (X, Y) is the content of the image memory 7.Furthermore, in step 207, the major axis a and the minor axis of the inertial equivalent ellipse of the content of the image memory 7 are calculated. This inertial equivalent ellipse is
It is an ellipse that shares a principal axis of inertia with a two-dimensional object and whose moment of inertia about its axial ring is equal to that of the object.

・・・・・・■ Next, in step 208, the information in the image memory 7 is set (
The image is rotated by -θ° around XG, Ya) and stored in the image memory 8. That is, image conversion is performed so that the five axes of the inertial equivalent ellipse obtained previously are horizontal. As a method of rotational movement, the image memory 7 (X
, Y) point information on the image memory 8 ((X-XG)
cos θ+(Y-Ya) sin θ, -
X −XG ) Sin θ + (Y−YG)Ca
Just transfer it to SO3 point. At this time, the same value (0) as the frontal area is stored in the area (X) where no data is transferred, such as the shaded area in FIG. 4(e).

Next, in step 209, the image memory 8 is subjected to window processing. That is, (XG −
a/2. Ya −b /2), (XG +a
/2. Ya −b /2), (XG −a
/2. Ya”b /2>, (XG +a /2.
A window specified by four points (YG +b/2) is multiplied (see FIG. 4(f)), and all data outside the window is set to O.

In step 210, the one-dimensional information within the window is projected onto the long axis of the one-dimensional equivalence ellipse, and the average value for each X-ray is determined. That is, (XG-α, Yc-b
/2), (Xa-α, Ya-b /2+1),
...(XG-α, Yg upper b/2) The relaxation of the shading information of each point divided by the number of points whose shading information is other than O is the average of the object at X=Xa-α Regarding the height, the data in the shaded area in FIG. 4(f) is O, so it is excluded from the calculation data for the average height).

The average height distribution thus obtained is as shown in FIG. 4 ((1,). This distribution state is stored in the calculation unit 9,
In step 211, the pitch angle ω of the object plane is calculated by least squares approximation.

Subsequently, in step 212, the shading information within the window is projected by q4 on the short axis of the inertia equivalent ellipse, and the average value for each Y value is determined. That is, (YG-β,
Xa −a /2), (YG-β.

XG -a /2 half 1), -・-・・-(Yv-β,
Y = Ya
-β is the average n of the objects t.

The average height distribution thus obtained is as shown in FIG. 4(h). This distribution state is stored in the calculation unit 9, and in step 213, the roll angle 8 of the object is calculated by least squares approximation.

〔Effect of the invention〕

As explained in detail above, the visual system according to the present invention converts the height information in the scene of the target object plane into grayscale information of a two-dimensional image, and inertially transforms the distribution of the obtained grayscale information into Approximate with an equivalent ellipse, project the above shading information onto the long axis and short axis of this ellipse, and calculate this I>
Since the three-dimensional inclination of the object plane is calculated from the 1-shadow value, the amount of information that becomes the counterstain is compressed by the projection,
The amount of information 8 is significantly reduced compared to the conventional method, making calculations faster, and the measurement accuracy of this calculation is the same as that of conventional processing methods.

[Brief explanation of the drawing]

Fig. 1 is a complaint correspondence diagram, Fig. 2 is a configuration diagram showing the hardware of the 81 sense system according to an embodiment of the present invention, Fig. 3 is a flow chart showing the software of the above system, and Fig. 4 is the same system as above. FIG. 1... Floodlight 2... Slit light 2a.
...Light cutting line 3...Movable mirror 11...Target object 51...Visual sensor 53...Conversion means 55...Approximate projection means 57...Inclination calculation means

Claims (1)

[Claims] (1) A visual system that processes a two-dimensional image obtained from a visual sensor using a light sectioning method to determine the three-dimensional inclination of a certain object plane included in a scene, which calculates the height of the object plane in the scene. A conversion means for converting information into gradation information of a two-dimensional image, approximating the distribution of the obtained gradation information with an ellipse that is inertially equivalent to the distribution, and projecting the gradation information onto the long axis and short axis of this ellipse. A visual system comprising approximate projection means for calculating the three-dimensional slope of the object plane from the projection value.