JP3365443B2 - 3D position measurement method - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a non-contact measurement method for measuring a three-dimensional position of an object using a camera. 2. Description of the Related Art When measuring a distance, a three-dimensional position, and a shape of an object using an image from a camera, first, an edge is obtained by differentiating two input images taken from two different viewpoints. Extract and binarize to obtain a line image. Then, a point on the edge of one line image is selected, a corresponding point on another line image is detected, and three-dimensional coordinates are calculated based on the principle of triangulation. [0003] However, there is a problem that it is difficult to detect a correct corresponding point because there are a plurality of points that can be handled when obtaining a corresponding point. In the edge extraction processing, an edge corresponding to the other edge may not be extracted due to noise or the like, and a corresponding point may not be found in many cases. Therefore, an object of the present invention is to provide a three-dimensional position measurement method capable of easily finding a corresponding point between two images and stably measuring a three-dimensional position of an object feature point. According to the present invention, there is provided a three-dimensional position measuring method using an image processing apparatus having a template matching function between a camera and a multi-valued image. A feature point of an object to be measured is extracted from one input image, an input image in the vicinity of the feature point is stored as a template image, and a position where the horizontal axis or the vertical axis of the camera imaging plane at the first point is parallel. The image input from the camera at the second point in the first position is defined as a second input image, and the template matching between the template image and the second input image is performed by extracting the lens center of the camera at the first point and the extracted feature points. Is performed only in the vicinity of a straight line area captured on the second input image, in which a straight line in the three-dimensional space connecting the above is obtained to find the position of the corresponding feature point on the second input image The three-dimensional position of the feature point is measured from the position of the feature point on the first input image and the position on the second input image according to the principle of triangulation. The present invention is based on the finding that the three-dimensional position of a feature point in the first input image is on a straight line connecting the lens center and the feature point on the actual object. . That is, this straight line is projected as a straight line on the second input image, and the equation of the projected straight line can be easily calculated from the position of the feature point on the first input image and the mutual position of both cameras. Can be. Therefore, the three-dimensional position of the feature point can be easily and accurately obtained by the above means. Hereinafter, the present invention will be described in detail with reference to examples. As shown in FIG. 1, the present embodiment includes two cameras having optical axes installed in parallel and an image processing apparatus having a template matching function. In the image processing apparatus, first, the images from the two cameras are stored in the first input image memory and the second input image memory by the image input unit. Next, the feature points of the target object are extracted from the first input image by the feature point extraction unit. In this processing, the vertices of the line image are selected by a conventional edge extraction technique. Further, the feature points may be detected based on a template image of the feature points given in advance. Using the positions of the feature points, the first input image in the vicinity of the feature points is cut into a size given in advance as a template image by the template cutout unit and stored in the template memory. Next, template matching between the template image and the second input image is performed by the template matching unit, and the position of a feature that substantially matches the feature of the template image is detected from the second input image. Finally, the three-dimensional position calculation unit calculates the three-dimensional position of the feature point from the feature position of the second input image corresponding to the feature position of the first input image according to the principle of triangulation. FIG. 2 shows an example in which the above processing is applied to block measurement. One vertex A of the block is extracted as a feature point from the first input image, and the vicinity thereof is cut out as a template. This template image is subjected to template matching with the second input image, and a feature point of a corresponding vertex is obtained. Although the features of the second input image corresponding to the template image have different shapes due to parallax, they are approximately equal and can be sufficiently detected by template matching. Further, in the template matching, since the multi-valued image information near the feature point is used, it is possible to stably detect the corresponding point. [0008] The template matching of this embodiment is the second
Although it is possible by checking the matching with the template in the entire area of the input image, the area that can correspond to the feature points of the first input image is a straight line area on the second input image. It is useless. As shown in FIG.
It is clear that the three-dimensional position of the feature point A in the first input image is on a straight line connecting the lens center and the actual feature point A on the object. This straight line is projected as a straight line on the second input image. The equation of the projected straight line can be easily calculated from the position of the feature point on the first input image and the mutual position of both cameras. If only the vicinity of this linear region is used as the search region for template matching, the search time required for matching is greatly reduced. As a second embodiment, one camera is attached to the tip of a robot, and the camera is moved.
FIG. 4 shows an example in which dimensional coordinates are measured. Except that the camera image before movement is stored in the first input image memory and the camera image after movement is stored in the second input image memory,
This is exactly the same as the three-dimensional coordinate measurement method described above. In this example, it takes time to move the camera, but there is an advantage in that the number of cameras can be reduced to one to reduce the camera portion. As described above, according to the present invention, the detection of a corresponding point, which has conventionally been difficult to detect stably, is performed by using a multivalued input template image near a characteristic point and the other input. By performing template matching with an image, there is an effect that the matching can be easily and accurately performed.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a configuration diagram showing a specific embodiment of the present invention; FIG. 2 is an explanatory diagram of corresponding point detection by template matching; FIG. 3 is an explanatory diagram showing a search region of template matching; 4 Configuration diagram of an embodiment having a camera moving mechanism [Description of References] 1 First camera 2 Second camera 3 Image processing device 4 Image input unit 5 First input image memory 6 Second input image memory 7 Template cutout unit 8 Template memory 9 Feature point extraction unit 10 Template matching unit 11 3D position calculation unit 12 First input image 13 Second input image 14 Object 15 Feature point A 16 Feature point neighborhood template 17 Corresponding point of feature point A Robot 19 Robot controller

Continuation of front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G01B 11/00 G06T 1/00 315 G06T 1/00 400

Claims (1)

(57) [Claim 1] In a three-dimensional position measurement method using an image processing apparatus having a template matching function between a camera and a multi-valued image, a first position input from a camera at a first point is provided. A feature point of the object to be measured is extracted from the input image, and the input image in the vicinity of the feature point is stored as a template image, and is located at a position where the horizontal or vertical axis of the camera imaging surface at the first point is parallel. An image input from the camera at the second point is used as a second input image, and template matching between the template image and the second input image is performed by connecting the lens center of the camera at the first point to the extracted feature points. The position of the corresponding feature point on the second input image is obtained by performing only in the vicinity of a straight line region where a straight line in a three-dimensional space is captured on the second input image, A three-dimensional position measuring method, comprising: measuring a three-dimensional position of the feature point from a position on the first input image and a position on the second input image according to the principle of triangulation.