JPS6278684A - Object search method - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed Description of the Invention] [Summary] In a manual search method for unknown objects, we focused on the fact that a two-dimensional projected image of a polyhedral object has only a finite number of views, and provided a dictionary of views regarding the polyhedral object. By comparing the two-dimensional projected image of the input unknown object with this appearance dictionary in advance, the number of references to the three-dimensional coordinates can be reduced.

[Industrial application field]

The present invention relates to an object retrieval method, and in particular, an object retrieval method that can perform a high-speed search for an unknown input object by being equipped with a view dictionary for polyhedral objects.Recent robot technology advances have enabled advanced work. Robots have begun to appear.

On the other hand, there is an increasing number of tasks that are too dangerous for humans to perform, such as inspecting the inside of a nuclear reactor or exploring the ocean floor, and there is an increasing need for robots to perform such high-level tasks on their behalf. There is.

In this case, the robot is required to have accurate and high-speed robot vision for the object being worked on, that is, the robot's eyes.

[Prior art and problems to be solved by the invention] FIG. 4 is a diagram explaining a conventional object search method, in which (a) is a schematic diagram, and (b) is a block diagram. It is.

First, in the observation bag W1, an unknown object 1a is photographed with a television camera 28 or the like to create a two-dimensional projected image 3a, and sent to the line drawing extraction circuit 2.

A line drawing extraction circuit 2 extracts a line drawing from the two-dimensional projected image and stores it in an image memory 3.

The line drawing information stored in the image memory 3 is directly compared with the two-dimensional projection image 6a of the object information 5a read from the three-dimensional dictionary 6 expressed in three-dimensional coordinates by the matching circuit 4.

Therefore, in the conventional method, the direction (viewpoint position) of the input object 1a is predicted in advance by some method, and then input to the projection conversion circuit 5 by another means, as shown in Figure (b). 2. Rotate the object described in the dictionary 6 in the same direction and project it in two dimensions (6a).
] - collate the input two-dimensional projected image 3a, or (v) create a two-dimensional projected image 68 while rotating the object described in the dictionary 6 into all solid angles. It was necessary to check with the dimensional projection image 3a.

Therefore, in the method (2), there is a problem in that the viewpoint position of the unknown object 1a must be known in advance, and in the method (2), there is a problem in that an enormous amount of matching time is required.

In view of the above-mentioned drawbacks of the conventional art, an object of the present invention is to provide a method that reduces the number of times of matching with a 3D dictionary by providing a view dictionary, and allows high-speed searching even when the orientation of an input object is unknown. It is something to do.

[Means for solving problems]

FIG. 1 is a diagram explaining the present invention in detail.

The input object 1a is photographed by a television camera 28 or the like, and becomes a two-dimensional projected image 3a.

Viewing Dictionary 4a describes the perspective from which an object is viewed.
□Describes how it will look (that is, the shape of the polygon and its connection relationship).

The above-mentioned appearance dictionary is disclosed in Japanese Patent Application No. 105355/1983, ``Stereoscopic Appearance Image Creation Apparatus'', which was previously filed by the applicant of the present application, so its details will be omitted, but the second
The following can be summarized using a conceptual diagram of the diagram appearance dictionary.

For example, when considering a square pyramid as a polyhedron, the appearance (shapes of the surfaces and their connection relationships) when viewed from a certain direction is limited to seven types, as shown in this figure.

Therefore, when this appearance is compared with the two-dimensional projected image of the input unknown object, if the unknown human-powered object is a square pyramid, it will match any of these seven types.

The view dictionary 4a stores, corresponding to each polyhedron, a pair of the view and viewpoint information for the view.

When such a view dictionary 4a is compared with the two-dimensional projected image 3a of the input unknown object, it is detected whether it matches the second view in the view dictionary 4a in FIG.

By searching the area corresponding to this view, it is possible to recognize from which direction the object appears.

Therefore, the three-dimensional coordinates of the three-dimensional dictionary 5a are projected in this direction to create a two-dimensional projected image 6a of the three-dimensional dictionary 5a.

In this way, the input two-dimensional projected image 3a and the two-dimensional dictionary 5a
The dictionary is compared with the dimensional projection image 6a, and if they match, the category of this dictionary is the category to be searched.

Therefore, the configuration is such that information associated with this category, for example, "pentagonal pyramid" is output as the search result 7a.

[Effect]

That is, according to the present invention, in a search method for an input unknown object, focusing on the fact that a two-dimensional projected image of a polyhedral object has only a finite number of views, a view dictionary regarding the polyhedral object is provided, and [1 input 2D projected image of unknown object,
By checking this view dictionary, the number of references to three-dimensional coordinates is reduced, so
This has the effect of reducing the number of matching matches with the three-dimensional dictionary and speeding up object retrieval.

〔Example〕

Embodiments of the present invention will be described in detail below with reference to the drawings with reference to FIGS. 1 and 2.

FIG. 3 is a block diagram showing an embodiment of the present invention. and verification circuit (7) are functional blocks necessary to implement the present invention.

Note that the same reference numerals indicate the same objects throughout the drawings.

In this figure, 1 is an observation device that projects a three-dimensional object into two dimensions, 2 is a line drawing extraction circuit that extracts a line drawing from the shaded image projected in the two dimensions, and 3 is a memory for storing the extracted line drawing. Reference numeral 8 indicates the above-mentioned perspective dictionary; 7 indicates a matching circuit L6 for comparing the line drawing and the perspective dictionary; 5 indicates the 3D coordinates from the 3D dictionary 6; Projection transformation r1 is projected onto the dimensional plane, and reference numeral 4 is a collation circuit (2) that collates the input two-dimensional projected image with the two-dimensional projected image of the three-dimensional dictionary 6.

Now, the given unknown object 1a is projected onto two-dimensional coordinates by the observation circuit 1. A line drawing extraction circuit 2 extracts line segments corresponding to the ridges of the object from this two-dimensional image, and the line drawings are stored in the image memory 3 as line drawings.

The matching circuit I 7 matches the line drawing read out from the image memory 3 with a large number of line drawings stored in the appearance dictionary 8, and selects polygons with the same polygonal shape and connection relationship as candidates for the appearance. Select as.

The projection conversion circuit 5 reads out a region corresponding to a candidate for the −1-century view from the view dictionary 8, and projects the three-dimensional data stored in the three-dimensional dictionary onto a two-dimensional plane corresponding to this region. do.

The two-dimensional projected image of the input object read from the image memory 3 and the two-dimensional projected image of the three-dimensional dictionary 6 output from the projection conversion circuit 5 are compared in the matching circuit ■4, and if they match, the category of this three-dimensional dictionary is determined. is the category to be searched, and the accompanying information is output.

If -number information is not obtained, this process is repeated until all the verifications are completed for the above-mentioned appearance candidates.

In this way, the present invention makes use of the fact that there are only a finite number of ways in which a two-dimensional projected image of a transhedral object can be seen, and by performing a pre-verification using these ways of seeing, a three-dimensional dictionary storing three-dimensional coordinates can be created. The feature is that the number of references to is reduced.

〔Effect of the invention〕

As described above in detail, the object search method of the present invention is a search method for an input unknown object.
Focusing on the fact that a dimensional projected image has only a finite appearance,
A view dictionary regarding the polyhedral object is provided, and the two-dimensional projected image of the input unknown object is compared with this view dictionary in advance, thereby reducing the number of references to three-dimensional coordinates. Therefore, the number of checks with the 3D dictionary can be avoided without reducing the number of times, and the object retrieval speed can be increased.

[Brief explanation of drawings]

FIG. 1 is a diagram explaining the present invention in detail. Figure 2 is a conceptual diagram of a visual dictionary. FIG. 3 is a block diagram showing one embodiment of the present invention. FIG. 4 is FIG. 2 illustrating a conventional object search method. In the drawing, 1 is an observation device, and 2 is a line drawing extraction circuit. 3 is an image memory, 4 is a matching circuit, 7 is a matching circuit, 8 is a view dictionary. 6 is a three-dimensional dictionary, and 5 is a projection conversion circuit. la is an input object, and 2a is a television camera. 3a is a two-dimensional projected image of the input object. 4a is the contents of the appearance dictionary. 5a is the contents of the three-dimensional dictionary. 6a is a two-dimensional projection 7a of an object stored in a three-dimensional dictionary
is the search result. are shown respectively. (Kano Seiyo's glue #; - 1st suggestion type 0 Maruaki 1ro diagram

Claims (1)

[Claims] A visibility dictionary (8) that stores the visibility of what kind of polygons are visible in what connection relationships when an other-hedral object is viewed from a certain direction, and a 3D object. A three-dimensional dictionary (6) for storing three-dimensional coordinates; and means (7) for comparing the input two-dimensional projected image of the unknown object with the appearance dictionary (8);
7), after first narrowing down the candidate categories of the input object, a detailed comparison with the three-dimensional dictionary (6) is performed to search for an input unknown object.