JPS6287807A - Method for visually determining entire shape of three-dimensional matter - Google Patents

Abstract

PURPOSE:To visually perceive the entire shape of a matter, by a method wherein a characteristic point is provided to a three-dimensional matter and the position of the characteristic point is calculated from one set of a stereoscopic image pair obtained by picking up the image of said characteristic point from two directions and this process is repeated the necessary number of times. CONSTITUTION:Three-dimensional matter 1 is arranged on a turntable 2 and an image plane 3 is arranged at a position picking up the image of said matter 1. The image of the characteristic point 4 capable of clearly specifying the position of the matter 1 provided to the surface of said matter is picked up to be projected to the characteristic point 5 on the image plane 3. Subsequently, when the turntable 12 is rotated by an angle psi, the point 4 moves to a point 4' and the image pickup position of the point 4' comes to the point 5' on the plane 3. When one set of a stereoscopic image pair is obtained by this method, the position of the point 4 can be determined by predetermined stereoscopic calculation. Next, in order to look other characteristic point not looked at this time, the turntable 2 is rotated. Until all of characteristics are subjected to stereoscopic image picking-up and calculated, the above- mentioned operation is repeated. Therefore, unless the matter 1 has a special shape, the shape of the matter can be determined before three sets of stereoscopic images are picked up.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention provides a method for measuring the shape of an object without touching the object in the process of processing, assembling, inspecting, transporting, etc. an object having a three-dimensional shape. or can be identified,
It can contribute to automation and efficiency of the above process.

[Prior Art] Conventionally, as a method for reproducing the entire shape of a three-dimensional object, a laser spot is swung horizontally by a scanning mirror and projected onto the object, and the object is rotated once on a rotating table. A method of capturing an image of the contour shape of the horizontal cross section, repeating this from the top to the bottom of the object to obtain the entire horizontal cross section contour, and reproducing the object image based on it.1) A cylindrical lens is used to slit the laser in the vertical direction. Create a light and project it onto the object on the rotary table. While the rotation is stopped, the vertical contour is imaged. This is rotated little by little in the horizontal plane until the object has rotated once. Method 2 of obtaining the vertical contour of and reproducing the object image based on this
), and the inventor's method is to place a plane mirror slightly behind the object so that the back side of the object can be seen, and simultaneously capture the direct image and mirror image of the object on one screen and analyze it. There is a method 3) that calculates the position of the feature point of an object. In the first two methods, the equipment is a laser light source, a device for scanning it or making it into a slit light,
A rotary table that rotates the object once or a device that moves the imaging system up and down is required, and it is also necessary to image a large number (perhaps dozens or more) of contour shapes in the horizontal or vertical direction. The method using a plane mirror has the great feature of only capturing one image, but it requires a plane mirror, and in the case of objects with complex shapes, there may be feature points that cannot be determined because they cannot be seen with a single plane mirror. sell.

[Problems to be solved by the invention] As mentioned above, in order to be able to use it in the field of processing, assembling, inspecting, and transporting objects, it is possible to contact objects without using special equipment, operations, or actions as much as possible. In addition, it is considered that the number of times of imaging can be reduced.

[Means for solving the problem] When humans try to ascertain the overall shape of a three-dimensional object as efficiently as possible, after viewing it from a certain direction, the next time the parts or feature points that were not visible at the beginning can be seen as clearly as possible. The person will move in the direction and look again.

And they will try to get to know the whole thing in as few moves as possible. The present invention is based on this idea.

[Operation] As for the operation performed by this method, first, a non-contact visual operation called imaging is performed on the target object. As a result, each feature point of the object, which is originally three-dimensional, is converted to a position (two-dimensional coordinate value) within a two-dimensional plane called the camera's image plane. Using a set of two stereo images taken from slightly different directions, it is possible to determine the three-dimensional position of a feature point within the image by solving simultaneous equations mathematically based on the principles of trigonometry. At this time, of course the back side of the object cannot be seen, so in order to know the overall shape, the relative position of the object and camera must be kept as minimally related to the previous image as possible, so that as many new feature points as possible can be seen. change position. Thus, the method only requires changing the relative position of the object and camera several times and imaging in the usual manner at each position.

[Example] In this method, it is only necessary to be able to change the relative positions of the object and the imaging system, and it is only necessary to know the relative relationship regardless of which of them moves. It is assumed that the imaging system is placed on top of the rotating table and can be rotated around the center of the rotating table as necessary, and the position and orientation of the imaging system are fixed.

Now, the object 1, the rotary table 2, and the image plane 3 of the imaging system
are arranged as shown in Figure 1. OtXtYtZt is the rotation table coordinate system (Zt axis is the rotation axis), OcXcYc
Zc is the camera coordinate system (Oc is the center of the lens), OiPQ is the image coordinate system, f is the focal length, and φ is the depression angle. Now perform the first imaging and select one feature point 4 (position Xt) of the object.
is projected onto the feature point 5 (position P) on the image plane.

Then, when rotating the rotary table by ψ as shown in Figure 2 to obtain the second image of the first pair of stereo images, feature point 4 (Xt) was moved to 4'(Xt'). shall be. Here, ψ cannot be set to a very large angle in order to maintain the correspondence between the two images or the similarity of features, and if it is too small, the accuracy when determining the position of feature points by performing stereo calculations will be reduced. to degrade. In experiments, around 30 degrees was appropriate for a polyhedron with 13 vertices (feature points) (though it also depends on the shape of the object). At this time, the feature point 5'(P') on the image plane corresponding to the feature point 4' becomes as shown in the figure. When a pair of stereo images is obtained as described above, one feature point 4 is obtained by the following stereo calculation.

Now, ■t■ [O, O, 1/f, O] Also, as a geometric transformation matrix, when Ttc is rotated from the rotary table coordinate system to the camera coordinate system and TR is rotated by ■ around the Zt axis, As a matrix, I2×4■[I2×2■O2×2], O2×4■[O2
×2■O2×2], and A1-A2■[A:b], then the simultaneous equations AXt=-b are obtained. By solving this, the position Xt of the feature point 4 can be found as Xt=-(AtA)-1Atb.

Next, in order to see other feature points that were not visible at this time,
Change the relative position of the object and camera. In the case of this embodiment, this corresponds to how much angle the rotary table should be turned.

Figure 3 is a two-dimensional drawing of the relationship between various quantities in order to consider this matter. The feature point 4 that was visible when the first set of stereo images was taken may be hidden when the second set of stereo images was taken, so the critical state is that the feature point 4 moves to 4' in FIG. This happens when 6 comes to 6'. From a geometrical relationship, the rotation angle at that time is π-2ap. Here, ap is the angle at which the feature point 4 is viewed as shown in Figure 3, and since this can be known, the rotation angle π
-2ap can also be known. Since this is a critical angle, the actual rotation angle is chosen to be a value that does not exceed this.

In the above manner, all the feature points of the object are stereo-imaged and this process is repeated until they are calculated. However, unless the object has a very special shape (such as having a complicated concave part), up to three sets of stereo-images can be used for the purpose. is often reached.

[Effects of the Invention] As described above, the present invention provides a means for determining the overall shape of an object in a non-contact manner without applying any special scanning or projection light to the object.

[Brief explanation of drawings]

FIG. 1 shows the relationship between the main elements of the embodiment according to the invention,
Figure 2 is a diagram to show the arrangement, Figure 2 is a diagram to show the relationship of various quantities during stereo calculation, and Figure 3 is a diagram to explain determining the critical rotation angle between stereo images. be. 1...Object 2...Rotary table 3...Image plane 4...Feature points 5...Feature points on the image plane 6...Plane [Reference materials]

Claims (2)

[Claims] (1) Having a three-dimensional, or three-dimensional, shape such as a polyhedron, whose position can be clearly identified, such as the vertex of the polyhedron or some additional marking on the surface. A method for quantitatively measuring and determining the overall shape of an object, as expressed by the coordinate values of characteristic points (hereinafter referred to as feature points), and for determining the position relative to the object. Both images are taken from a pair of stereo images taken from directions that differ by the minimum required angle (approximately 30 degrees) using a variable imaging system (such as a television camera). The positions of the feature points are calculated based on the two-dimensional coordinate values on the image plane, and then a pair of stereo images is captured again from a direction that newly includes the feature points that were invisible at the time of imaging. By repeating this process the minimum number of times necessary, all the coordinate values of the entire feature points of the object can be calculated and the shape of the object can be expressed. A method for visually determining the overall shape of a three-dimensional object. (2) In the visual determination method of the overall shape of a three-dimensional object, in order to capture as few stereo image pairs as possible, focus on one main surface on the object and make sure that the surface has exactly one shape. A method for determining the imaging direction of the next pair of stereo images from the angle of the direction that appears as a line (i.e., the surface is degenerated into a line).