JPH08145628A - Method for specifying position of object - Google Patents

Abstract

PURPOSE: To eliminate the effect of fluctuation in the environmental illumination and an obstacle within the view by specifying the position of an object using a combination of most reliable results among a plurality of template pattern matching results. CONSTITUTION: An object for registration is set and a plurality of images of partial feature region, features and the positions of reference point are extracted from the images of the object being picked up by means of a camera 20 through an image pickup means 1 and stored at an image memory section 6 and position memory sections 9, 12, respectively. Subsequently, a pattern matching means 2 determines a plurality of partial region images in the memory section 6 from the images of the object having different position and attitude being picked up through the means 1. A means 3 generates a plurality of positional combinations of template for all positions of a matched template. The combination is then subjected to position/attitude conversion by a means 4 which evaluates the extent of matching of relative positional relationship. The combination at a position corresponding to a best evaluation is regarded as most reliable results, thus specifying the position of the object.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for specifying the position of an object by using image pattern matching, and more particularly, the position of the object for improving resistance to environmental lighting fluctuations and obstacles in the visual field. Regarding the identification method.

For example, in most of the robot systems up to now, the teaching playback system has been used, so that sufficient attachment accuracy is required when fixing the work. For this reason, the cost for jigs has become a problem in production systems that frequently change product use. To solve this problem, a method has been considered in which the jig is shared or simplified to reduce the cost, and the positional deviation of the work that occurs due to this is absorbed using a vision system or the like.

The present invention is not limited to such a production system, but in specifying the position of an object using an image, it is possible to deal with uneven light and dark variations due to shadows and obstacles on an apparent image. This enables accurate positioning of the object.

[0004]

2. Description of the Related Art Conventionally, a single partial area image including the whole object or a part thereof has been used as a template for pattern matching.

[0005]

According to the conventional pattern matching method, in the captured image, changes in the apparent image due to environmental illumination fluctuations or obstacles in the camera field of view within the partial region corresponding to the template. However, there is a problem in that pattern matching is impossible or gives an erroneous result.

The present invention aims to solve this problem.

[0007]

The main feature of the present invention is that the position of an object is specified using a combination of the most reliable results of pattern matching results of a plurality of templates. It differs from the conventional technology in that it uses multiple templates and uses the most reliable combination of results.

[0008]

In the present invention, a plurality of templates are used and pattern matching is performed for each of them. In pattern matching of some, but not all, of these templates, if correctly specified template positions (hereinafter referred to as specified positions) are obtained, N (N is N) of combinations of specified positions. 2 or more natural number)
For all the combinations of the above specified positions, the correspondence between the relative relationship of the positions registered in advance of the template corresponding to the elements of each combination and the relative relationship of these specified positions An assessment of the extent is obtained, and the combination of identified positions that corresponds to the best of these can be the most reliable result.

At the same time, by comparing this best evaluation value with a specific threshold value derived from an error at the time of specifying the template position in the pattern matching, it is possible to evaluate the reliability with higher accuracy for the most reliable result. Can be given.

[0010]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 1 is a diagram showing a configuration example of an object position specifying apparatus using the present invention, FIG. 2 is a process explanatory diagram of an embodiment of the present invention, FIG. 3 is a diagram showing an installation example of an object and a camera,
FIG. 4 is a diagram showing an example of an object image, FIG. 5 is a diagram showing an example of the position of each feature and a reference point position, and FIG. 6 is a diagram showing an example of a feature region image.

In the figure, 20 is a camera for picking up an image of an object, 2
Reference numeral 1 is an object position specifying device that specifies the position of an object using an image. The image capturing means 1 is a processing means for capturing an image from the camera 20 and storing it. The pattern matching means 2 is a processing means for specifying the position of a feature registered in advance on the captured image. The position combination generating unit 3 is a processing unit that creates all combinations of N or more specified positions from the positions of the plurality of features specified by the pattern matching unit 2.

The position conversion / evaluation means 4 pre-registers, for each of these combinations, features corresponding to the specified positions by rotational translational conversion for the specified positions constituting the combination. It is a processing means for aligning a certain position so as to match as much as possible, and storing the evaluation of the degree of matching and the conversion. The reference point position generation means 5 is a processing means for applying a conversion corresponding to the best evaluation among the evaluations made by the position conversion / evaluation means 4 to the position of a reference point registered in advance to create a new position of the reference point. is there.

The feature registration image storage unit 6 is a means for storing information on a partial area image of a feature as shown in FIG. 6 which is registered in advance. The specified feature position storage unit 7 is a unit that stores the position of the specified feature obtained by the pattern matching unit 2. The position combination storage unit 8 is a unit that stores the position combination of the identified features obtained by the position combination generation unit 3. The feature position storage unit 9 is a unit that stores the position of a feature that is registered in advance.

The evaluation storage unit 10 includes a position conversion / evaluation means 4
It is a means for storing the evaluation for the combination of the positions of the identified features obtained in step. The conversion storage unit 11 is a unit that stores information about the conversion applied to the combination of the positions of the identified features. The reference point position storage unit 12 is means for storing the position of the reference point registered in advance. The specified reference point position storage unit 13 is means for storing the position of the specified reference point obtained by the reference point position generating means 5.

The operation of the apparatus shown in FIG. 1 is as follows. First, in order to register an object in advance, the object 22 is placed at a standard position on the surface plate 23 as shown in FIG.
An image including an object as shown in FIG. Then, the partial image of a plurality of features as shown in FIG. 6 on this image and the positions of these features and the positions of the reference points as shown in FIG. ，
It is stored in the characteristic position storage unit 9 and the reference point position storage unit 12.

Based on these, the position of the newly placed object is specified by the processing shown in FIG. Therefore, first, the image capturing means 1 captures and stores an image including an object whose position and orientation are different from those at the time of registration from the camera 20. In the captured image, the pattern matching means 2 finds the positions of a plurality of partial area images as shown in FIG.

Next, the position combination generating means 3 generates N (N ≧ 2) or more template position combinations for all the template positions for which pattern matching has succeeded.

For example, when making two or more combinations,
The combination of four templates A (hereinafter referred to as A), template B (hereinafter referred to as B), template C (hereinafter referred to as C), and template D (hereinafter referred to as D) when the pattern matching is successful is ( A, B, C,
D) and (A, B, C), (A, B, D), (A, C,
D), (B, C, D) and (A, B), (A, C),
(A, D), (B, C), (B, D), (C, D) position '.

The position conversion / evaluation means 4 registers the positions of the templates stored in the position storage unit 7 of the specified features constituting each combination thereof and the position storage unit 9 of the features in advance with respect to these templates. The position and orientation conversion by rotational translation is performed on a given position, a conversion that minimizes the sum of squares of the distances of the respective new positions obtained is found, and the conversion information is converted and stored. The sum of squares of this distance is stored in the evaluation storage unit 10 at the same time.

Next, the reference point position generating means 5 obtains the conversion having the smallest sum of squares of the distances from the evaluation storage unit 10 and the conversion storage unit 11, and registers it in the reference point position storage unit 12. The obtained reference point is converted and stored in the reference point position storage unit 13 identified as the position of the new reference point. At the same time, the success of the position identification is examined by, for example, whether the sum of squares of the minimum distance is within four times the square of the position identification error obtained by the pattern matching means 2.

The image capturing means 1 described above
Can be easily realized by, for example, an image capturing board mounted on a personal computer. Further, the pattern matching means 2 can be easily realized by a normalized correlation pattern matching or the like.

Rotational translational change in position conversion / evaluation means 4
In the conversion process, the position registered in advance is (x_i, Y_i),
The position specified by turn matching is (x '_i,
y ’_i), E = Σ_k{[(X_k・ Cos θ−y_k・ Sin θ + x_s) −
x '_k]²+ [(X_k・ Sin θ + y_k・ Cos θ + y_s)
-Y '_k]²}, Simultaneous equations (∂E / ∂θ) = 0, (∂E / ∂x
_s) = 0, (∂E / ∂y _s) = 0 using the Newton method
It can be easily realized by solving the problem. In addition,
In the above equation, θ is rotated, (x_s, Y_s) Indicates translation
You.

[0023]

As described above, according to the present invention,
Even if changes in the apparent image due to environmental lighting fluctuations or obstacles in the camera's field of view occur in some characteristic parts, the position of the target object will be highly accurate if the characteristic parts that do not change remain. It is possible to specify with
There is an advantage that it is not necessary to rigorously prepare the environment when specifying the position of the object using the image as in the conventional case.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration example of an object position specifying device using the present invention.

FIG. 2 is an explanatory diagram of a process according to an embodiment of the present invention.

FIG. 3 is a diagram showing an installation example of an object and a camera.

FIG. 4 is a diagram showing an example of an object image.

FIG. 5 is a diagram showing an example of a position of each feature and a reference point position.

FIG. 6 is a diagram showing an example of a characteristic region image.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Image capturing means 2 Pattern matching means 3 Position combination generating means 4 Position conversion / evaluation means 5 Reference point position generating means 6 Feature registration image storage unit 7 Specified feature position storage unit 8 Position combination storage unit 9 Features Position storage unit 10 evaluation storage unit 11 conversion storage unit 12 reference point position storage unit 13 identified reference point position storage unit 20 camera 21 object position identification device

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification number Reference number within the agency FI Technical indication location G06T 7/60 (72) Inventor Eiji Mitsuya 1-6, Uchisaiwaicho, Chiyoda-ku, Tokyo Nihon Telegraph Phone Co., Ltd.

Claims (1)

[Claims] 1. A first step of capturing and storing an image including an object, and a plurality of characteristics of the object registered in advance, each position on the image captured by the first step is described. A second step of specifying, a third step of making all combinations of N (N ≧ 2) or more specified positions from the plurality of specified positions, and the N or more specified positions. For each and every combination of positions, the rotational and translational transformations of the identified positions that make up the combination ensure that these identified positions and the corresponding pre-registered positions of the features correspond best. In accordance with the above, the fourth process of storing the evaluation of the degree of coincidence and this conversion, and the conversion corresponding to the best evaluation among these evaluations is applied to the position of the reference point registered in advance, and a new reference point Fifth making position An object position specifying method, comprising: