JPS61262986A - Producing device for seeming dictionary of cube - Google Patents

Abstract

PURPOSE:To decide quickly the type or the candidate of the cubic form of a matter to be recognized by collating the extracted seeming features of a cube with the contents of a dictionary. CONSTITUTION:The number of visible faces, the types of forms of these visible faces and the connecting relations among these faces which are obtained by observing a cube in a prescribed direction are written on a visible face table 6 in each prescribed direction. While the different appearances of the cube are stored in a dictionary storage part 7 in terms of the connecting relations of polygons. Then the contents of the table 6 are collated 8 with those of the part 7. Based on the result of this collation 8, the combinations among those items concerning the visible faces of the cube described on the table 6 are processed 9. Then the result of this processing 9 is written on the part 7. Thus a triangular prism, for example, is stored in terms of a triangle, a square, a triangle and a square sharing a side, two squares sharing a side and a triangle and two squares sharing a side with each other.

Description

[Detailed Description of the Invention] [Summary] A three-dimensional recognition device such as the Roboso I Vision is equipped with It is used for the identification of
A dictionary of how the body looks that is structured to ensure that there are no omissions.

[Industrial field of use]

The present invention relates to a three-dimensional recognition device, particularly a three-dimensional appearance dictionary creation device used to recognize whether a body is an I'j cube or a pyramid based on the appearance of the three-dimensional object. It's something to think about.

For example, some robots used in 1-assembly T-yards detect a desired part from among the parts being transferred by a conveyor or the like, grip it, and move it to a designated location.

In such a case, it is first necessary for the robot to quickly and reliably identify the type of object, regardless of its orientation.

[Conventional technology]

FIG. 4 is a block diagram of a three-dimensional recognition device used for robot vision and the like.

1 observes the object to be recognized and converts it into two-dimensional image data. For example, an observation section that uses industrial television (ITV).

2 is a feature extraction unit that extracts features of the appearance of the object to be recognized from the two-dimensional image data obtained by the observation unit 1;

3 is a feature dictionary that stores three-dimensional data that describes the shape of each solid for each type of three-dimensional objects to be recognized;

4 is from the three-dimensional data stored in the feature dictionary 3.

A conversion unit that creates several appearance features that are expected to be extracted by the feature extraction unit 2.

Further, in step 5, the three-dimensional shape of the object to be recognized is calculated by comparing the appearance feature of the object to be recognized extracted by the feature extraction unit 2 with some appearance characteristics created by the conversion unit 4. This is an identification part that identifies the type.

For example, each of the 150,000 vertices is set to P as shown in Figure 5 rat.
l, P2...P8, the η-cube is a platform described by the coordinate values and connection relationships (adjacent vertices via the ridge line) of each vertex P1 to P8, as shown in Figure 5 (bl). It can be completely represented by one-body data, and the feature dictionary 3 stores three-dimensional data described in this manner for each type of one-body.

On the other hand, as shown in FIG. Any of the three rectangles that share .

Similarly, the appearance characteristics of a triangular prism are shown in Figure 1 fa.
It is one of five types: t to f[ll.

Therefore, for example, if the appearance feature extracted by the feature extraction unit 2 is a quadrilateral, the solid should be either an eta-cube, a triangular prism, or another square pyramid or square trapezoid.

For this reason, the conversion unit 4 creates various appearance characteristics expected for each solid from the 1″ i-body data of each solid stored in the feature dictionary, and the features of the recognition target extracted by the feature extraction unit 2 are By comparing the appearance characteristics with these, the type of the object to be recognized (h) is limited to a small number of candidates, and if necessary, the type of the three-dimensional object to be recognized is identified using other means. Then, the result is supplied to the robot control unit.

[Problem that the invention seeks to solve]

- In the 1″ one body recognition device having the F configuration, the feature dictionary 3
It takes a long time to create several visual features that are expected to be extracted by the feature extraction unit 2 from the three-dimensional data stored in the three-dimensional data. Therefore, there is a problem in that it takes a long time to recognize the three-dimensional object to be recognized.

SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide a dictionary creation device for how a three-dimensional object appears when included in a three-dimensional recognition device, which can help improve the speed of recognition.

rMeans for solving problems] FIG. 1 is a block diagram of the principle of the present invention.

Reference numeral 6 denotes a visible surface table in which the number of visible surfaces obtained when a solid is observed from a predetermined observation direction, the type of shape of each RJ viewing surface, and the connection relationship of each visible surface are described for each of the predetermined directions.

Reference numeral 7 denotes a dictionary storage unit that stores the appearance of three-dimensional objects based on connection relationships of polygons.

8 compares the description contents of the visible surface table 6 for each direction with the contents of the dictionary storage section 7, and calculates the number of oJ viewing surfaces described in the pJ viewing front table 6, the type of each visible surface, and the number of oJ viewing surfaces described in the pJ viewing front table 6. A collation unit that checks whether a combination with a connection relationship is stored in the dictionary storage unit 7.

9 shows the visible surface table 6 as a result of the matching in the matching section 8.
A combination of the number of visible surfaces, the type of each visible surface, and the connection relationship of each visible surface described in
If there is no such combination, it is a write processing section that performs processing for writing this combination into the dictionary storage section 7.

[Effect]

In other words, 1. For example, for a triangular prism whose -face and bottom face are parallel, as shown in Figure 2 rat, each face is f1 to f5, and each face f1 to f5 is considered to be all independent, and this is observed. At this time, there are 5 directions in which only one side can be seen.

There are nine directions in which only two sides can be seen, and six ways in which three sides can be viewed, for a total of 20 views. ) is written in the visible surface table 6.

By the way, in the visible surface table 6, each surface fl-f5 is considered to be independent, so there are 20 different views, but for example, the characteristics of the view when observed from one direction (one triangle) The characteristics of the appearance when observed from directions 2, 3, and 4 (1 triangle) are the same, and the characteristics of appearance when observed from directions 2, 3, and 4 (11 squares) are the same.
11) There are some similar items.

So by integrating these.

Only appearance features that may be extracted from two-dimensional image data obtained through observation are stored in the dictionary storage section 7 as a three-dimensional appearance dictionary.

As a result, the appearance of various solids is as follows: 1. For example, for a triangular prism, there are 1 triangle, 1 quadrilateral, 1 triangle and quadrilateral that share sides, and 2 squares that share one side and 1 that share one side with each other
These triangles and two quadrangles (see FIG. 7) are stored in the dictionary storage unit 7.

〔Example〕

FIG. 3 is a configuration diagram of the embodiment.

Reference numeral 10 denotes an appearance dictionary that stores the appearance of various three-dimensional objects constructed as described above.

Reference numeral 11 denotes an identification unit that limits the candidate types of the three-dimensional object to a small number of types by comparing the appearance characteristics extracted by the feature extraction unit 2 with the stored contents of the appearance dictionary 10. It is.

By comparing the appearance characteristics of the object to be recognized extracted by the feature extraction unit 2 with the stored contents of the appearance dictionary 10, the type of the object to be recognized is determined, or some small number of appearance characteristics are determined. limited to candidates.

Furthermore, if necessary, the type of the object to be recognized is recognized by another recognition section (not shown), and the recognition result is supplied to the robot control section.

〔Effect of the invention〕

As explained above, according to the present invention, the three-dimensional shape type or candidate of the object to be recognized can be determined in a short time, and the response speed can be improved when applied to robots, tovisions, etc. I can do it.

[Brief explanation of drawings]

FIG. 1 is a block diagram of the principle of the present invention. Fig. 2 (al ft11 is an explanatory diagram of the action. Fig. 3 is a block diagram of the embodiment. Fig. 4 is a block diagram of the stereoscopic recognition device. 7(a) to 7 are explanatory diagrams of the conventional example. In the figure, l is an observation unit, 2 is a feature extraction unit, 6 is a visible surface table, 7 is a dictionary storage unit, 8 is a collation unit, and 9 is a Write processing unit. Alpha-th Ak-th invasion Himi Shizuka 2ji-go, Henguchi (α) (#) (c) Osaihyo A-row's Kaya 7 (ol) (e) Meimeihaen

Claims (1)

[Scope of Claims] A visible surface in which the number of visible surfaces, the type of shape of each visible surface, and the connection relationship of each visible surface obtained when a solid is observed from a predetermined observation direction are described for each of the predetermined directions. A table (6), a dictionary storage unit (7) that stores the appearance of solid objects based on connection relationships of polygons, and a dictionary storage unit (7) that stores descriptions of the visible surface table (6) for each observation direction. A collation unit (8) that compares the content, and the number of visible planes, the type of each visible plane, and the connection relationship between each visible plane, which are described in the visible plane table (6) according to the matching result of the collation unit (8). A writing processing unit (9) for writing a combination of the three-dimensional appearance dictionary into a dictionary storage unit (7).