JPH07136959A - Robot positioning method - Google Patents

Abstract

PURPOSE:To correct the position and attitude of a robot after the movement of the robot based on the attitude estimation of an object so as to enable the high accuracy positioning of the robot. CONSTITUTION:An input image feature extracting part 20 receives the two-dimensional image of an object photographed by a fixed camera 10 and extracts a feature segment and a feature point. An image-object coordinating part 40 coordinates the extracted feature segment and feature point respectively with the features of segment description and point description of an object model previously prepared in an object model description storage part 30, and an object attitude estimating part 50 estimates the attitude of the object from this coordination. A robot primary induction part 60 induces a robot according to the estimated attitude, and a robot secondary induction part 80 corrects the position of the robot according to the position information from a range sensor 70 on a robot hand obtained at the time of inducing the robot and position information from the camera 10.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a robot positioning method for robot work, and more specifically, to a position and orientation of a three-dimensional object input as a two-dimensional image from a fixed camera in a three-dimensional space. The method relates to a method of estimating a position, guiding a robot, and correcting the position by a range sensor attached to the robot.

[0002]

2. Description of the Related Art Conventionally, in robot work, many methods have been adopted in which the work of a previously described operation procedure is faithfully performed by numerical control for an object whose position / orientation in a robot coordinate system is given in advance. Has been. In addition, when the approximate relationship between the robot coordinate system and the object coordinate system is known, a method of observing the object position using a range sensor attached to the robot and matching the robot coordinate system and the object coordinate system is used. ing.

On the other hand, as described in JP-A-1-232484 and JP-A-2-51008, there is a method of estimating the position / orientation of an object using a fixed camera and guiding a robot.

[0004]

Of the above-mentioned conventional techniques,
The numerical control method has a problem that it cannot be applied to robot work in various environments. Further, although the coordinate matching by the range sensor is effective for the correction of the minute positional deviation, it is still far from the realization of the work in various environments. On the other hand, only by estimating the position and orientation of an object using a fixed camera, it is possible to obtain only a certain degree of orientation estimation accuracy, and robot work that can be used as it is is limited to special applications.

An object of the present invention is to eliminate the above-mentioned problems of the prior art, and to eliminate the problems of the known shape object corresponding to the camera input image.
It is an object of the present invention to provide a method for positioning a robot at high speed and with high accuracy by estimating a posture in a dimensional space and correcting the position and posture by a range sensor.

[0006]

In order to achieve the above object, a robot positioning method according to the present invention picks up an image of a target object by an image pickup means fixed in a work environment and extracts features from the picked up two-dimensional image. The step, the step of associating the extracted feature night with the object model of the target object prepared in advance, the step of obtaining a pose candidate from the association, and estimating the pose of the target object, the estimation The step of guiding the robot to the vicinity of the target position according to the posture, and after the guiding, the target object is observed by the range sensor attached to the robot and the image pickup means fixed to the working environment, and the object position is obtained, and the object position And a step of correcting the robot position.

[0007]

In the present invention, two of the target objects imaged by the imaging means are
By associating the characteristic line segments and characteristic points of the three-dimensional image with the features of the object model line segment description and object model point description of the target object that have been prepared in advance, the object posture is estimated, and the robot is guided. After that, by comparing the object position observed by the range sensor with the object obtained from the two-dimensional image of the image pickup means to correct the robot position, it is possible to realize the positioning with the desired accuracy.

[0008]

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram of an embodiment of the present invention. 10 is a camera, 20 is an input image feature extraction unit, 30 is an object model description storage unit, 40 is an image / object associating unit,
Reference numeral 50 is an object posture estimation unit, 60 is a robot primary guidance unit, 7
Reference numeral 0 represents a laser range sensor, and 80 represents a robot secondary guiding unit.

The camera 10 is fixedly installed in the work environment of a robot (not shown). The input image feature extraction unit 20 inputs a two-dimensional image of a three-dimensional object (hereinafter referred to as a target object) which is a work target captured by the camera 10 whose focal length and lens center are known, and three or more feature lines are input. Minutes and at least two or more feature points are extracted. FIG. 2 shows this, in which 100 is a target object, arrows are characteristic line segments,
Circles indicate feature points. The geometric shape of the target object 100 is known.

In the object model description storage unit 30, for three or more effective line segments in the target object 100 in advance, a three-dimensional direction display obtained in an arbitrary coordinate system defined in the object and the line segments thereof are displayed. An object model line segment description that describes a feature (for example, an optical feature) and a line segment and a three-dimensional coordinate value of one point at a collinear position, and a coordinate system defined in the object of at least two or more points The object model point description in which the three-dimensional coordinate values obtained in (3) and the feature (for example, optical feature) of the point are described is stored.

The image / object associating unit 40 stores the feature line segments and feature points of the camera input image (two-dimensional image) extracted by the input image feature extracting unit 20 and the object model description storage unit 30. The object model line segment description and the feature description in the object model point description are associated with each other, and three line segments and two points are selected from the associated items,
A correspondence table of three-dimensional coordinates and two-dimensional image coordinates is generated for line segments and points.

The object pose estimation unit 50 estimates all possible pose candidates in the three-dimensional space of the line segment set and the designated point pair selected by the image / object association unit 40. Next, the image of the line segment set and the point set included in the object model description storage unit 30 is estimated from each pose candidate,
By comparing this with a two-dimensional image, an appropriate posture is selected from the posture candidates. That is, the posture of the target object is estimated.

Since the processing up to this point is described in detail in, for example, Japanese Patent Laid-Open No. 2-51008, detailed description will be omitted.

The robot primary guiding unit 60 guides the robot in the vicinity of a predetermined position (target position) of the target object according to the posture selected (estimated) by the object posture estimating unit 50. At this time, measures such as obstacle avoidance may be taken as necessary.

The robot secondary guiding unit 80, in the position / posture guided by the robot primary guiding unit 60, follows the position / posture of the robot according to the object position observed by the laser range sensor 70 provided on the robot hand of the robot. To correct. Specifically, as shown in FIG.
The position of the characteristic point of the target object 100 (circled in FIG. 3) observed by the laser range sensor 70 attached to the tip of the robot hand 200, and the object model of the laser slit image simultaneously captured by the fixed camera 10. The position and orientation estimation of the object is corrected by the position of the feature point above.

FIG. 4 shows an example of a processing procedure in the robot secondary guiding unit 80. A laser range sensor (also referred to as a laser range finder) 70 attached to the tip of the robot hand 200 converts a planar laser beam into a target object 1
00 (410), and the laser image is read by the laser range sensor 70 to detect the characteristic points (42
0), the three-dimensional position of the feature point viewed from the robot hand 200 is calculated (430). Specifically, as shown in FIG. 3, the target object 10 is detected by the laser range sensor 70.
A laser image (slit image) of a portion including a ridgeline of 0 is observed, an intersection portion of the ridgeline is extracted from the laser image, and its position is used as a feature point (circle mark) to calculate its three-dimensional coordinates.
At the same time, a laser image (slit image) is captured together with the object image by the fixed camera 10 (440), and the position of the intersection (feature point) of the laser image and the ridgeline on the object (on the model) is calculated (450).

The above operation is repeated at least three times,
The object position observed by the laser range sensor 70 and the object position obtained from the image of the fixed camera 10 are matched to accurately guide the robot (460).

[0019]

As described above, according to the present invention,
The posture of the target object is estimated by associating the features of the two-dimensional image of the target object captured by the image capturing means with the object model of the target object prepared in advance, and the robot is moved according to the estimated posture thus obtained. After that, by integrating the object position observed by the range sensor and the object position obtained from the camera information, the accurate position of the object can be estimated at high speed, thus realizing the industrial visual sensor or the visual system of the robot. Widely applicable to. Further, according to the present invention, the positioning of the object can be realized even when the accurate shape model is not given in advance.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of the present invention.

FIG. 2 is a diagram showing a relationship between a camera and a target object.

FIG. 3 is a diagram showing a relationship between a camera, a laser range sensor, and a target object after the robot is moved by estimating the posture of the target object.

FIG. 4 is a diagram showing an example of a processing procedure of a robot secondary guiding unit.

[Explanation of symbols]

 10 camera 20 input image feature extraction unit 30 object model description storage unit 40 image / object association unit 50 object orientation estimation unit 60 robot primary guidance unit 70 laser range sensor 80 robot secondary guidance unit 100 target object

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI technical display location G05D 3/12 T 9179-3H

Claims (1)

[Claims] 1. A robot positioning method based on posture estimation of a shape-known object, the method comprising the steps of capturing an image of a target object with an image capturing means fixed in a work environment and extracting features from the captured two-dimensional image. A step of associating the extracted feature with an object model of the target object prepared in advance; a step of obtaining a pose candidate from the association and estimating the pose of the target object; a robot according to the estimated pose. Step of guiding to the vicinity of the target position, after the guidance, the target object is observed by the range sensor attached to the robot and the imaging means fixed to the working environment to obtain the respective object positions, and the object positions are compared. A robot positioning method comprising: a step of correcting a robot position.