JPH055609A - Cubic recognition method of image - Google Patents

Abstract

PURPOSE:To enable accurate detection of the position of an object concerning a cubic recognition method of images by which the object is photographed with a plurality of cameras to determine features thereof. CONSTITUTION:A plurality of cameras 21 and 22 are set so that an object 20 can be projected on screens of the cameras 21 and 22. An operator is selected to extract a contour line closer to a vertical component of the object with respect to a second straight line which is generated by projecting a plane containing a straight line connecting projection centers of the two cameras representing the plurality of the cameras 21 and 22. The contour line of the object is extracted and stored by the operator to be displayed on a screen. A reference line is drawn parallel with the second straight line on the screens to determine a reference point as intersection between the reference line and the contour line. Thus, the position of a part corresponding to the reference point of the object is determined from the positions of the reference points on the screens and a relative position between the cameras.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image stereoscopic recognition method for photographing an object by using a plurality of cameras and capturing its characteristics.
In particular, the present invention relates to an image stereoscopic recognition method for extracting the outline of an object, displaying it on the screen, and capturing its features.

Such an image stereoscopic recognition method is used for a robot or the like operating in a factory or various facilities. For example, in a factory, in order to handle a work or inspect a product, it is necessary to know information such as a distance between the robot and the work. For that purpose, it is necessary to accurately detect the distance between the camera and the work.

[0003]

2. Description of the Related Art A conventional image stereoscopic recognition method is shown in FIG.
As shown in FIG. 7, when the object 63 is photographed by the two cameras 61 and 62, a straight line 71 connecting the projection centers 61a and 62a of the two cameras is created, and one plane 64 including the straight line is formed.
To form. The plane 64 is a surface generally called an epipolar surface. FIG. 7 is a diagram showing the video screens 61b and 62b of the cameras 61 and 62, respectively. The lines 65 and 66 on each screen are reference lines for obtaining information on the object 63. Here, the projections of the plane 64 of FIG. 6 on the video screens 61b and 62b become the reference lines 65 and 66.

In order to detect the characteristics of the object 63, first, an intersection point where the reference lines 65 and 66 intersect with the contour line of the object 63 on each of the video screens 61b and 62b is obtained, and this intersection point is used as a reference point. In the figure, points 67 and 68 correspond to this reference point, but normally either one is selected. Here, the point 67 is selected.

When the reference point 67 is determined, the positions of the reference point 67 on the video screens 61b and 62b and the cameras 61 and 62 are determined.
The position of the reference point 67 can be specified from the distance between them. Thereby, the distance between the robot and the work can be obtained by triangulation.

[0006]

By the way, in order to obtain a more accurate position, in addition to the reference lines 65, 66, the reference line 69,
Suppose 70 is drawn as shown (or the first drawn reference line happens to overlap this position). In such a case, since the reference lines 69 and 70 overlap the contour line of the object 63, the reference points extracted on the screens 61b and 62b are often at different positions. In particular, when the shading of the contour line is different on each screen due to the adjustment of light or the like, an accurate reference point cannot be extracted.

The present invention has been made in view of the above circumstances, and provides an image stereoscopic recognition method capable of accurately determining the position of an object by extracting an accurate reference point even on an arbitrary reference line. The purpose is to provide.

[0008]

FIG. 1 is a principle diagram of an image stereoscopic recognition method of the present invention for achieving the above object. A plurality of cameras are installed so that objects can be imaged on the screens of the plurality of cameras (step S1), and a plane including a straight line connecting the projection centers of two cameras represented by the plurality of cameras is projected on the screen. The operator for extracting the contour line close to the vertical component of the object with respect to the second straight line that can be created is selected (step S2), and the contour line of the object is extracted and stored by the operator and displayed on the screen. (Step S3), a reference line parallel to the second straight line is drawn on each screen (step S4), a reference point which is an intersection of the reference line and the contour line is obtained (step S5), and on each screen. The position of the portion corresponding to the reference point of the object is obtained from the position of the reference point and the relative position between the cameras (step S6).

[0009]

In extracting the contour of the object, the contour line of the component substantially perpendicular to the reference line is extracted, so that the reference point which is the intersection of the reference line and the contour line can be accurately obtained.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 2 is a diagram showing the configuration of an image recognition apparatus for implementing the image stereoscopic recognition method of the present invention. In this embodiment, two cameras 21 and 22 are used to capture the object 20. The cameras 21 and 22 are installed so that the straight line 26 connecting the respective projection centers 21a and 22a is included in the epipolar surface 23. The positions and shooting angles of the cameras 21 and 22 are controlled by actuators. Alternatively, it may be fixed at a fixed position. In addition, here, for simplification of description, it is assumed that the optical axes 21c and 22c are parallel to each other. Furthermore, it is assumed that the horizontal components of the screens of the cameras 21 and 22 are set to be parallel to the epipolar plane.

The images taken by the cameras 21 and 22 are sent to the control device 24. The image sent to the controller 24 is
It is displayed on the display screen of the display device 25 and stored in the memory in the control device 24. The control device 24 processes the sent image to obtain information on the object 20, and controls the robot and the like.

FIG. 3 shows a display screen 21b of the display device 25,
It is a figure which shows the state of 22b. The display screen 21b displays the outline of the image captured by the camera 21, while the display screen 22b displays the outline captured by the camera 22 and subjected to the image processing described below. As can be seen from the figure, the contour of the object 20 shows only the component perpendicular to the reference line of the screen.

FIG. 4 is a diagram showing an example of an operator used to extract the contour line. In the present embodiment, image processing is performed using the operator of FIG. 4A to emphasize and extract the outline of the component perpendicular to the horizontal plane of the screen. However, when the reference line is inclined, that is, when the relative position of the camera is licked, it is not necessarily limited to this.
For example, as shown in FIG. 4B, an operator who emphasizes a contour line inclined by 30 ° from a vertical line may be used. Further, in FIG. 4, an example of a 3 × 3 operator is shown, but 5 ×
An operator having a large number of matrices such as 5 or 7 × 7 can extract more accurate contour lines.

Returning to FIG. 3, on the screens 21b and 22b, reference lines 31 and 32 are drawn in order to obtain reference points for obtaining information such as the position of the object 20. This may be a so-called epipolar line obtained by projecting the epipolar surface 23 on the screens 21b and 22b, or may be a straight line parallel to the epipolar line. The reference point is the object 20 on each screen.
A point at which the contour line of and the reference line intersect is selected. However,
When there are a plurality of points, in this embodiment, the leftmost point is used as the reference point. Therefore, in FIG. 3, the points 33 and 34 are determined as the reference points, respectively. These reference points 33 and 34 both correspond to the point P in FIG.

Thus, on the screens 21b and 22b,
Since only the vertical contour line is displayed, the reference line 31,
It does not overlap with 32. Therefore, an accurate reference point can be obtained no matter where the reference line is drawn.

When the reference points 33 and 34 are obtained, the control device 24 performs a calculation for specifying the position of the point P. Thereby, the position of the object 20 is specified. FIG. 5 is a diagram showing a method for specifying the position of the object 20. The line connecting the projection center 21a of the camera 21 and the reference point 33 is 41, and the camera 22
The line connecting the projection center 22a of the and the reference point 34 is 42,
It is considered that the point P of the object 20 is at the intersection of the line 41 and the line 42.

When the angle between the line 41 and the optical axis 21a is α and the angle between the line 42 and the optical axis 22a is β, these angles are from the center points 21d and 22d of the screens 21b and 22b. The distance between the projection centers 21a and 22a and each reference point 3
It can be easily obtained from the positions of 3, 34 on the screen. Furthermore, the lengths of the lines 41 and 42 can be calculated from the angles α and β and the distance L between the cameras which is known in advance by the principle of triangulation. As a result, the distance between the camera and the object can be accurately obtained.

In the above embodiment, in order to simplify the explanation, the optical axes 21c and 22c of the camera are made parallel to each other,
Furthermore, although the horizontal components of the screens of the cameras 21 and 22 are set to be parallel to the epipolar surface 23, this is not always the case, and the objects 20 on the screens 21b and 22b are not necessarily in compliance with this.
If the cameras 21 and 22 are installed so that the image can be reliably captured, the distance between the camera and the object can be obtained as described above.
That is, the optical axes 21c and 22c are deviated, and the screen 21
Even if the horizontal components of b and 22b are not parallel to the epipolar surface 23, accurate reference points 33 and 34 can be obtained by extracting the contour lines that are perpendicular to the obliquely drawn reference lines 31 and 32. .. Therefore, as in the above embodiment, the distance between the camera and the object can be accurately obtained.

[0019]

As described above, according to the present invention, when the contour of the object is extracted, the contour line of the component almost vertical to the reference line is extracted, and therefore, it is the intersection of the reference line and the contour line. The reference point can be accurately obtained. Therefore, the position of the object can be accurately obtained.

[Brief description of drawings]

FIG. 1 is a principle diagram of an image stereoscopic processing method of the present invention.

FIG. 2 is a diagram showing a configuration of an image recognition device.

FIG. 3 is a diagram showing a state of a display screen.

FIG. 4 is a diagram illustrating an example of an operator.

FIG. 5 is a diagram showing a method for identifying the position of an object.

FIG. 6 is a block diagram of a conventional image processing apparatus.

FIG. 7 is a diagram showing a screen state of a conventional image stereoscopic recognition method.

[Explanation of symbols]

 20 Object 21, 22 Camera 24 Control Device 25 Display Device

Claims (1)

Claim: What is claimed is: 1. An image stereoscopic recognition method in which an object is photographed by using a plurality of cameras and the features thereof are captured, and the plurality of cameras are installed so that the object appears on the screens of the plurality of cameras. Then, (step S1), a vertical line of the object is close to a second straight line formed by projecting a plane including a first straight line connecting the projection centers of two cameras of the plurality of cameras on the screen. An operator for extracting the contour line is selected (step S2), the contour line of the object is extracted and stored by the operator and displayed on the screen (step S3), and the second line is displayed on each screen. A reference line parallel to a straight line is drawn (step S4), a reference point which is an intersection of the reference line and the contour line is obtained (step S5), and the position of the reference point on each screen and the camera are separated. Relative position and Determining the position of the portion to further corresponding to the reference point of the object (step S6) image solid recognition method, characterized in that. 2. The stereoscopic image recognition method according to claim 1, wherein the reference line is an epipolar line.