JPH06180216A - Parts position/attitude measuring instrument - Google Patents

Abstract

PURPOSE:To provide a low-cost instrument for measuring the position an attitude of parts, which operate at a sufficiently high processing speed. CONSTITUTION:The title instrument is provided with a plurality of distance sensors 31 an 32 for measuring the distance data at a plurality of measuring points on horizontal surfaces with the same height, a three-dimensional characteristic extraction part 36 for obtaining the three-dimensional characteristics from the distance data which are obtained by the distance sensors 31 and 32, a section model storage part 37 for storing a section shape when the stable attitude of all parts as the shape model of parts is cut on the horizontal surface, and a comparison part 38 for determining the position and attitude of parts by comparing and verifying the three-dimensional characteristics obtained at the above three-dimensional characteristic extraction part 36 and the section shape of parts stored at the above section model storage part 37 on the surface of parts 39.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a component position / orientation measuring device applied to a robot or the like for gripping a component.

[0002]

2. Description of the Related Art As a device for measuring the position and orientation of parts, a device having a structure as shown in FIG. 6 has been conventionally used. As shown in the figure, the measuring device includes distance sensors 11, 3
Dimensional feature extraction unit 12, collation unit 13, and three-dimensional model storage unit 14
Composed of. As the distance sensor 11, for example, as shown in FIG. 7, a binocular stereoscopic method using two TV cameras is used. The images obtained by the two TV cameras are I
Letting L and IR be the centers of the respective lenses be OL and OR, the point P in the three-dimensional space can be calculated as the intersection of straight lines OLPL and ORPR if the images PL and PR are known.

This is generally because many points where the brightness changes are found on the image and a point PR corresponding to the point PL is searched for on another image to obtain three-dimensional coordinates of several thousand points. It is possible. This is commonly referred to as a range image.

Another method is called a range finder as shown in FIG. This range finder is composed of a slit light projector 15, a slit light scanning device 16, a rotating mirror 17 and a television camera 19, and the slit light 18 projected by the slit light projector 15 is driven by the slit light scanning device 16. It is reflected by a controlled rotating mirror 17 and scans over an object 20.

An image is obtained by the television camera 19 while moving the slit light 18 little by little by the rotation of the rotating mirror 17, and the distance image is obtained by measuring the position on the object surface where the slit light 18 strikes by the principle of triangulation. Can be obtained. For example, when 256 times are taken and the three-dimensional coordinates of 256 points in the slit image on each screen are obtained, 256 × 2
A range image of 56 points can be obtained.

From the range image obtained as described above,
The three-dimensional feature extraction unit 12 extracts geometrical features of the part, such as the size of the surface, the normal direction, the angle between the surfaces, and the like.

On the other hand, the three-dimensional shape of the part is stored in advance in the three-dimensional model storage unit 14, and the three-dimensional shape of the stored part and the three-dimensional feature extracted by the three-dimensional feature extraction unit 12 are stored. Are sequentially compared by the matching unit 13 to determine which three-dimensional feature corresponds to which model surface. Then, the matching unit 13
Calculates the position and orientation of the component based on the correspondence result.

For example, in FIG. 9, rotation of coordinate conversion from the extracted normal vectors of the two surfaces S1 and S2 and the corresponding normal vectors of the two surfaces M1 and M2 registered in the three-dimensional model. The component R (vector notation in the figure) is determined, and the translational component t (vector notation in the figure) can be obtained from the vertex coordinate data of the corresponding surface. This coordinate conversion is a coordinate conversion with the sensor coordinate system (0-XYZ) in which the component is actually placed, and corresponds to the position and orientation of the component in the sensor coordinate system.

[0009]

In the above-mentioned conventional apparatus, the processing for the distance sensor 11 to obtain the distance image is extremely complicated and time-consuming, and the cost of the apparatus increases. was there. Similarly, the three-dimensional feature extraction unit 12 also has a problem that the process is complicated and time-consuming, and at the same time, an erroneous feature is often extracted, which prevents the collation unit 13 from performing a correct collation. It was

That is, the cost of the distance sensor 11 is high,
The information of the input distance image is too large and the processing time is too long, and the extraction processing by the three-dimensional feature extraction unit 12 becomes complicated in order to extract the three-dimensional feature with high reliability. Such a component position / orientation measuring device is not practical because it is high in cost and low in processing speed.

The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a highly practical component position / orientation measuring device having a low cost and a sufficient processing speed. is there.

[0012]

SUMMARY OF THE INVENTION That is, the present invention provides a plurality of distance sensors for measuring distance data at a plurality of measuring points on a horizontal plane of the same height on the surface of a component, and a plurality of these distance sensors. A three-dimensional feature extraction unit that obtains three-dimensional features from the distance data obtained by the distance sensor, a cross-sectional model storage unit that stores the cross-sectional shape when the stable postures of all the components are cut on the horizontal plane as the shape model of the component, and A three-dimensional feature obtained by the three-dimensional feature extraction unit and a cross-sectional shape of the component stored in the cross-sectional model storage unit are compared and collated to determine the position and orientation of the component. Not only the structure of the sensor is simplified to reduce the cost, but also the processing required for the three-dimensional feature extraction is simplified because the input data is small, and the processing speed can be greatly improved.

[0013]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 shows the configuration thereof, and reference numerals 31 and 32 are distance sensors placed at a height h with respect to the work surface 30, respectively. These distance sensors 31 and 32 are horizontally moving devices 33 and 34.
Work surface 30 while moving at equal intervals with each
Distance to the placed component 39 d _j ¹ , D _j ² (J =
1, 2, ...) Is measured, and the obtained measurement data is sent to the three-dimensional feature extraction unit 36.

The three-dimensional feature extraction unit 36 outputs a signal for instructing movement to the horizontal movement control unit 35 which controls the drive of the horizontal movement devices 33 and 34, while inputting the measured data d _j ¹ , D _j ² To the measurement point direction vector vk (k =
1, 2, ...) is obtained.

Further, the three-dimensional feature extraction unit 36 finds all stable postures for the part 39. The stable posture of a part is found by first finding the smallest convex polyhedron that includes all vertices of the part called the convex hull, and then sequentially checking whether one face of the convex hull can be stably placed on the horizontal plane. To be Whether or not it can be placed stably can be determined by whether or not the vertical line passing through the center of gravity of the part passes through the plane placed on the horizontal plane.

All stable postures with respect to the part 39 are obtained, that is, the model coordinate system (0
If the coordinate transformation (Ri, ti) from −XYZ) to the stable posture Si (i = 1, 2, ...) Is obtained, the three-dimensional feature extraction unit 36 determines the distance sensor 1 for each stable posture.
The cross-sectional shape Ci obtained by cutting the stable posture is obtained on the horizontal plane of the height h where 1 and 12 are placed, and the obtained cross-sectional shape Ci is stored in the cross-section model storage unit 37.

The collation unit 38 checks which cross-sectional shape Ci stored in the cross-section model storage unit 37 the direction vector vk extracted by the three-dimensional feature extraction unit 36 is measured on, and based on the result. The position and orientation of the part 39 is calculated.

In the above-mentioned structure, the distance sensor 3
1 and 32, for example, measurement data d _j ¹ as shown in FIG. , D
_j ² Then, the three-dimensional feature extraction unit 36 obtains the direction vectors v1, v2, v3 by linear approximation of the measurement points.

Here, the measurement point d ₁ ¹ , D ₂ ¹ , D
₃ ¹ Is able to determine the direction vector v1 to be linearly approximated as a group, the measurement points d ₄ ¹
Is not on the straight line, the measurement point d ₅ ¹ And another direction vector v2 is calculated. D ₁ ² obtained from the distance sensor 32 ~d ₅ ² Are all on the same straight line, the direction vector v3 can be obtained as one group.

Further, the three-dimensional feature extraction unit 36 obtains a cross-sectional shape C1 as shown in FIG. 3 which is taken along a horizontal plane of a certain height h with respect to one stable posture S1 of the component 39. In this case, the data representing the sectional shape C1 stored in the sectional model storage unit 37 is the vector e1 indicating the length and direction of each side,
e2 and e3.

The matching unit 38 uses the direction vector v shown in FIG.
It is checked to which cross-sectional shape Ci stored in the cross-section model storage unit 37, 1, v2 and v3 are measured. This is the side e _j ⁱ of the sectional shape Ci. Since (j = 1, 2, ...) Is stored in the cross-section model storage unit 37, the following condition

[0022]

[Equation 1] Cross section Ci and side e _j1 ⁱ satisfying , E _j2 ⁱ , E _j3 ⁱ Search for.

For example, in FIG. 2, three-dimensional features v1, v2,
When v3 corresponds to e1, e2, and e3 of FIG. 3, respectively, the above equations (1) and (2) are satisfied, so that the coordinate transformation (θ, u) from the coordinate system of FIG. 3 to the coordinate system of FIG. Desired. Therefore, the position and orientation of the component 39 are coordinate transformation (R1, t1) and coordinate transformation (θ,
Determined by synthesis with u).

Using the above method, 3
It is possible that the edge corresponding to the dimensional feature vk cannot be determined, that is, multiple corresponding candidates may be found. Such multiple interpretations can occur frequently, especially depending on the shape of the part. In this case, the two distance sensors 31, 32
Can be installed at different heights or another distance sensor can be added to increase the number of three-dimensional features vk, thereby eliminating the above multiple interpretation.

Here, when the second and third distance sensors are installed with different heights, the cross-sectional shape cut along the horizontal plane of that height is stored in advance in the cross-section model storage unit 37 according to the height. You need to register. . Further, the distance sensors 31 and 32 do not necessarily have to be installed opposite to each other as shown in FIG. 1, and the measurement directions may be changed and installed.

Regarding the number of measurement points, in FIG. 2, 5 points were measured for each of the distance sensors 31 and 32, and a total of 10 points were measured. However, by measuring more finely, the accuracy of the direction vector extracted is improved and a complicated shape is obtained. Features can be accurately extracted even from parts. Therefore, how many measurement points are necessary may be determined according to the complexity of the shape of the target component.

Similarly, with respect to the number of distance sensors, the more sensors are installed, the more three-dimensional characteristics can be obtained, so that the accuracy and reliability are improved, but the cost and the processing are increased accordingly. This leads to a decrease in speed. Therefore, the number of distance sensors may be determined by the trade-off between performance and cost.

For the distance sensors 31 and 32, a laser beam is normally projected by a projector 41 as shown in FIG.
A so-called displacement sensor is used, which can receive the light reflected by 42 through a lens 43 by a position detector 44 and measure the distance d to the object surface 42 according to the principle of triangulation according to the light receiving position. Such distance sensors 31 and 32 are moved at equal intervals by driving the horizontal movement devices 33 and 34 by the horizontal movement control unit 35.

As an equivalent to the distance sensor having such a horizontal movement mechanism, a fixed slit light type range finder as shown in FIG. 5 can be considered. In this fixed slit light type range finder, the slit light projector 45 projects the slit light 46 onto the object 48, and the image thereof is picked up by the television camera 47. From the position of the slit image on the image, 3
Based on the principle of angular measurement, it is possible to obtain the three-dimensional coordinates of the portion on the surface of the object on which the slit light 46 is shining.

When this range finder is used, the horizontal moving devices 33 and 34 are unnecessary, and the slit light projector is arranged so that the slit light surface is parallel to the horizontal plane and has a specific height h.
45 should be installed. According to this fixed slit light type range finder, as compared with the range finder shown in FIG. 8, a slit light scanning mechanism and the like can be eliminated, so that the structure is simple and the cost of the apparatus can be greatly reduced. .

[0031]

As described above, according to the present invention, a plurality of distance sensors for measuring distance data at a plurality of measuring points on a horizontal plane of the same height on the surface of a component, and a plurality of these distance sensors. 3D features from the distance data obtained by the distance sensor
A three-dimensional feature extraction unit, a cross-sectional model storage unit that stores a cross-sectional shape when a stable posture of all the components is cut on a horizontal plane as a shape model of the component, and the three-dimensional feature extraction unit described above.
Since the dimensional feature and the cross-sectional shape of the component stored in the cross-section model storage unit are compared and collated to determine the position and orientation of the component, the structure of the distance sensor is simplified and the cost is reduced. In addition, the processing required for three-dimensional feature extraction is simplified due to the small amount of input data,
It is possible to provide a highly practical component position / orientation measuring device capable of significantly improving the processing speed.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration according to an embodiment of the present invention.

FIG. 2 is a view showing an example of measurement data according to the same embodiment.

FIG. 3 is a diagram illustrating a cross-sectional shape corresponding to the measurement data of FIG.

FIG. 4 is a diagram illustrating the principle configuration of the distance sensor of FIG.

FIG. 5 is a diagram showing another configuration example of the distance sensor of FIG.

FIG. 6 is a block diagram showing a configuration of a conventional component position / orientation measuring device.

FIG. 7 is a diagram illustrating the principle configuration of the distance sensor of FIG.

FIG. 8 is a diagram illustrating a principle configuration of the distance sensor in FIG.

9 is a diagram illustrating a method of determining the position and orientation of a component by the collating unit in FIG.

[Explanation of symbols]

11, 31, 32 ... Distance sensor, 12, 36 ... Three-dimensional feature extraction unit,
13, 38 ... Collating unit, 14 ... Three-dimensional model storage unit, 15, 45 ... Slit light projector, 16 ... Slit light scanning device, 17 ... Rotating mirror, 18, 46 ... Slit light, 19, 47 ... TV camera, 2
0, 48 ... Object, 33, 34 ... Horizontal movement device, 35 ... Horizontal movement control section, 37 ... Section model storage section, 39 ... Parts, 41 ... Projector, 42 ... Object plane, 43 ... Lens, 44 ... Position Detector.

Claims (1)

[Claims] 1. A plurality of distance sensors for measuring distance data at a plurality of measurement points on a horizontal plane of the same height on the surface of a component, and three-dimensional data from the distance data obtained by the plurality of distance sensors. A three-dimensional feature extraction unit that obtains a feature, a cross-sectional model storage unit that stores a cross-sectional shape when the stable postures of all the components are cut on a horizontal plane as a shape model of the component, and the three-dimensional feature obtained by the three-dimensional feature extraction unit And a collating unit for comparing and collating the sectional shapes of the components stored in the sectional model storage unit to determine the position and orientation of the component.