JPS6256814A - Calibration system for three-dimensional position measuring camera - Google Patents

Abstract

PURPOSE:To measure a three-dimensional position with high accuracy irrelevantly to the distortion of a lens system by using a regular pattern and determining the correspondence relation between the position of an optional point in an object space determined by a TV camera and the position on its picked-up image. CONSTITUTION:A calibration plate PLT on which the regular pattern is drawn is placed on a Z0 axis so that it can shift in position, and its pattern edge is detected to project an X0Y0 coordinate system on an image coordinate system. Consequently, the spatial coordinate values of one point on the pattern corresponding to one point A on the image are calculated. Then the pattern is picked up by the TV camera every time the calibration plate PLT is placed at a different position on the Z0 axis, and then the object spatial coordinate values of each calibration plate position corresponding to specific picture elements are calculated. This three-dimensional position measuring method is employed to obtain a sight line equation.

Description

[Detailed Description of the Invention] [Field of Application of the Invention] The present invention relates to a calibration method for a 23-dimensional position measuring camera,
In particular, the present invention relates to a three-dimensional position measurement camera calibration method that enables highly accurate three-dimensional position measurement.

[Background of the invention]

Until now, the relationship between the TV right camera for three-dimensional position measurement and the target space captured by this camera has been expressed by the relationship between two coordinate systems H-XeYeZe, O-X, 0YO20, as shown in Figure 11. There is. In this case, the origin H of the camera coordinate system H-XeYcZe is the front principal point of the lens,
The Ze axis corresponds to the optical axis direction of the camera lens, and the Xe axis and Ye axis correspond to the vertical and horizontal directions on the image, respectively. Now, 1: Now consider a plane that intersects perpendicularly to the Zc axis, and when viewing the target space through that plane from the origin I of the camera coordinate system H-XCYCZC, the image projected onto this plane will be the image captured by the TV right camera. This plane is called the image plane because it corresponds to the image captured by the camera.Of course, the range of the image plane is limited depending on the field of view of the TV right camera, but this situation can be explained in two ways. If the relationship between the coordinate system o-X0YOZO and the camera coordinate system H-XcYcZc is known, then at what position on the image plane will a point in the target space be projected? In addition, since the points in the space that are projected onto one point on the image are all the points on the straight line a shown in Figure 11, the points on this image and the corresponding straight line in the space are known. It is also possible to derive a relationship with

By the way, there are six degrees of freedom (three positions and rotations) between the two coordinate systems, so determining these six variables in some way is the key to determining the three-dimensional position measurement camera. In reality, this has been commonly used as a calibration method within the target space. However, there is slight distortion in the camera lens system that cannot be visually confirmed, and this distorts the image non-linearly. It is not possible to correct the deviation due to this, resulting in problems in high-precision position measurement. Incidentally, examples related to this type of calibration method include, for example, Japanese Patent Laid-Open No. 114887/1987.

[Purpose of the invention]

An object of the present invention is to provide a three-dimensional position measuring camera lens system that enables highly accurate three-dimensional position measurement even when distortion exists in the camera lens system.

[Summary of the invention]

For this purpose, the present invention determines a TV camera line-of-sight equation for each point corresponding to the image, thereby determining the position in the target space captured by the TV camera and the position on the image when that point is captured. This is to determine the correspondence relationship between the two. If the line-of-sight equation for each pixel is determined independently, the correspondence between image coordinates and spatial coordinates is described through these line-of-sight equations. Because it is required by the data, it is absorbed into the overall data. By the way, in order to obtain this line-of-sight equation, it is necessary to measure coordinate points in space through a TV image, but the line-of-sight equation can be obtained by observing a regular pattern through a camera.

[Embodiments of the invention]

The present invention will be explained below with reference to FIGS. 1 to 10.

As mentioned above, if the principal point of the camera lens is the origin and the optical axis of the lens is the Zc axis, the image captured by the camera can be thought of as an image projected onto the image plane perpendicular to the Zc axis. can.

Image coordinates (I, J) are set on this image plane (
(see FIG. 11), a straight line existing in the object space corresponds to a certain point on the image coordinates.

In other words, all points existing on the straight line are projected onto that one point on the image coordinates. Now, if we call such a straight line the line of sight and the linear equation in the target space the line of sight equation, then if we can determine the line of sight equation for each pixel, we can find the correspondence between the image coordinates and the spatial coordinates. That means something.

By the way, in order to obtain the line-of-sight equation, the position in the target space must be measurable through the TV image. Therefore, in order to realize this, a calibration plate PLT on which a regular pattern as shown in FIG. 2 is drawn is used. If this calibration plate PLT is placed in a variable position on the Zo axis as shown in FIG. 3, the target spatial coordinate system will be determined from the pattern plane. In this case, the pattern surface is perpendicular to the Zo axis, and the direction of the vertical and horizontal lattice consisting of checkered edges on the pattern is Xo@
The direction coincides with the Yo axis direction. Further Z
The intersection point with the o-axis is specified on the pattern in advance, but if you specify the value of the position on the Zo-axis where the pattern is first placed, the target seat collar system is uniquely determined by that pattern surface. This is the result. Note that a checkered pattern is considered as the pattern because its edges can be easily extracted through image processing. What is important is that a group of straight lines at regular intervals can be cut out from the pattern in the XQ axis direction and the Yo axis direction.

Next, I will explain how to use this pattern to obtain the spatial coordinate value of a certain point on this pattern from the TV image.By detecting the pattern edge, as shown in Fig. 4, .

The yo coordinate system will be projected onto the image coordinate system, and as a result, the target space coordinate value of one point on the pattern corresponding to one point A on the image can be found as shown in Figure 5. . That is, first, four straight lines surrounding point A in the vicinity of point A on the image are determined, and then these intersection points P, -P are determined.
The image coordinate values of 4 are obtained. After this, in order to determine the relationship between the intersections P1 to P4 and point A, the internal division ratio Q as shown in FIG.
, m are found. In this case, since the target space coordinate values of points P and -P are known from the pattern, this internal division ratio is
By applying this to the target space coordinate values of I to P4, the target space coordinate value of one point A can be obtained.

Now, in order to obtain the line-of-sight equation using such a three-dimensional position measurement method, each time the calibration plate is placed at a different position on the Zo thin axis, the pattern is captured with a camera, and the pattern is captured at each calibration plate position corresponding to a specific pixel. It is necessary to find the target space coordinate values of . This is because if two points in space are determined, the line-of-sight equation can be determined from those two points. The data for determining the line-of-sight equation are at least two object space coordinates (Xo+YO+Zo) for each pixel coordinate (I,,1).
) +(Xo' + Yo' + Zo'), but it is not easy to obtain data for all image coordinates (I, J). However, if such data is obtained for image coordinate points with appropriate sampling widths s and s' as shown in Figure 1, then based on this data, any point on the image can be The line-of-sight equation is obtained as shown in FIG.

That is, first, a sample point S, -34 in the vicinity of an arbitrary point A (Im, Jm) on the image is found, and then internal division is performed to express the relationship between point A and point S, -s on the image. Ratio Q.

m is calculated. Two target space coordinate points for point A are calculated by applying this internal division ratio to the target space coordinate data of points S and -S4. The line of sight equation corresponding to point A can be found from these two points.

The accuracy of this calibration method depends on how the pattern edges are cut out on the image, but since the image has nonlinear distortion, the cut edge lines should not be treated as straight lines on the image but as curved lines by high-order approximation. Bye. As a result, the accuracy of capturing the pattern onto the image is guaranteed, and distortion due to nonlinearity is no longer a problem.

To give an example of a 3D position measurement method using 3D vision using such a line-of-sight equation as data for a stereo vision type and a range finder using slit light, Figure 7 shows the appearance of a stereo vision type visual sensor. This is what is shown. The three-dimensional shape of the target C is known by capturing the target C with two TV cameras A and B and measuring the three-dimensional position of each point on the target C.

As shown in Fig. 8, in this system, measurement points ARpAL are found in the left and right images, and the spatial position of measurement point A is measured as the intersection of the line-of-sight equation nRtQL for those points on the left and right images. be.

FIG. 9 shows the external appearance of a range finder sensor using slit light. This consists of one TV camera A and a slit light projector B. The slit projector B emits a planar slit light C, but when the slit light C is projected onto an object, a light cutting line E appears on the surface of the object.
will occur. The position on the TV screen when this light cutting line E is captured by the TV camera is that the slit light C plane is T.
Since it is not parallel to the optical axis Zc of the V camera, it moves according to the spatial position where the light cutting line E occurs according to the ZC coordinate, but by using this relationship and knowing the position of the light cutting line on the TV screen - , the spatial position of the optical cutting line can be measured. If the target spatial coordinate values of several points on the slit light plane are measured in advance using a regular pattern during calibration, the plane equation of the slit light plane can be calculated.
As shown in Figure 0, one point A on the light section line E on the plane of the object is captured as an image by calculating the equation of the line of sight Q with respect to the image position A' and combining that line of sight with the slit light plane P.
The spatial coordinate value is determined as the intersection point with .

〔Effect of the invention〕

As explained above, according to the present invention, nonlinear distortion of the image of the camera lens can be absorbed into the data describing the correspondence between the image coordinates and the target space coordinates during calibration, so that the camera lens system is free from distortion. Even in the case where there is an effect, three-dimensional position measurement can be performed with high accuracy.

[Brief explanation of the drawing]

Figures 1 and 6 are diagrams for explaining the method for determining the line-of-sight equation for arbitrary image coordinates, Figure 2 is a diagram showing a calibration plate according to the present invention, and Figure 3 is a diagram at the time of calibration. Diagram to explain its use in. Fig. 4 2 Fig. 5 is a diagram for explaining the method of obtaining the target space coordinate values using the regular pattern on the calibration board; Fig. 7;
FIG. 8 is a diagram for explaining the application of the present invention to stereo vision, FIGS. 9 and 10 are diagrams for explaining the application to a range finder, and FIG. 11 is a diagram for explaining the three-dimensional position. FIG. 2 is a diagram for explaining a general relationship between a camera coordinate system of a measurement TV camera and a target space coordinate system. H-XeYcZc = camera coordinate system, O-X, YoZ. ...Target space coordinate system, ■)), 1'...Calibration plate. Representative Patent Attorney Tadashi Akimoto Figure 1 ■ Figure 3 a Figure 4 Figure 5 ■ Figure 6 ■ Figure 7 Figure 8 LMR Figure 9 Ugi Figure 10 Figure 11

Claims (1)

[Claims] T as a 3D position measurement camera for capturing images of objects
A three-dimensional position measurement camera calibration method in a visual recognition device comprising a V camera, a monitor for displaying images captured by the camera, and an image processing device for analyzing the images on the monitor, the method comprising: Using a regular pattern that is placed in a variable position in space, the three-dimensional position on the pattern is read directly from a monitor, and the TV is used for each point on the image.
The feature is that by determining the line of sight of the camera, the correspondence relationship between the position of an arbitrary point in the target space captured by the TV camera and the position on the image when the point is imaged is determined. 3D position measurement camera calibration method.