JPS61114381A - Stereoscopic viewing device - Google Patents

Abstract

PURPOSE:To recognize an object at a high speed by extracting the feature points of each of binary images obtained by >=3 units of 2-dimensional image pickup means, securing the correspondence between the feature points and calculating the 3-dimensional coordinates of the corresponding point on the subject. CONSTITUTION:Feature points PA, PB and PC of a point P on an object appear on images HA, HB and HC obtained through TV cameras 1A, 1B and 1C respectively. An image epipolar line l2 of a straight line connecting the focal point of the camera 1A and said feature points is set on the 2nd image HB. The feature point PB of the image HB is set on the line l2, and the feature point PC of the 3rd image HC is set on an intersecting point of epipolar lines l3 and m3. Thus it is understood that the feature points PA-PC are opposite to each other as images at the point P. Therefore the 3-dimensional coordinates of the point P are obtained in the form of an intersecting point among straight lines FAPA, FBPB and FCPC.

Description

[Detailed Description of the Invention] <Technical Field of the Invention> This invention relates to the recognition of the shape of a three-dimensional object, or the recognition of the shape of a three-dimensional object.
This invention relates to a stereoscopic viewing device used to measure the position and orientation of a three-dimensional object, and in particular, the present invention relates to a stereoscopic viewing device used to measure the position and orientation of a three-dimensional object.
A new device for measuring dimensional coordinates at high speed is provided.

<Summary of the Invention> This invention obtains gray scale images of a three-dimensional object viewed from each direction using three or more two-dimensional imaging means, binarizes each gray scale image to generate a binary image, and then It is configured to extract feature points from each binary image and make correspondences between the feature points to calculate the three-dimensional coordinates of the corresponding points on the object. Efforts are being made to speed up the process.

<Background of the Invention> Conventional stereoscopic viewing devices use a two-dimensional imaging device to obtain a grayscale image of a three-dimensional object, perform edge enhancement processing on this grayscale image, and then draw straight lines to the series of points that make up the edges. By applying the data to a line drawing, and analyzing this line drawing, shape recognition of a three-dimensional object is performed. However, in the case of this method, the calculation cost for edge enhancement processing is extremely high, and it is also easily affected by noise on the image, making it difficult to always perform accurate shape recognition.

In addition to the above method, slit light is irradiated onto a three-dimensional object,
A system in which the slit light image is observed using a two-dimensional imaging device while scanning the slit light has also been implemented, but in this method, since it involves scanning the slit light, it takes a long time to observe, and The disadvantage is that the recognition process cannot be accelerated.

<Purpose of the Invention> The present invention is intended to eliminate the drawbacks of the above-mentioned conventional methods, and its purpose is to provide a stereoscopic viewing device that can accurately and rapidly measure the three-dimensional coordinates of a point on a three-dimensional object. do.

<Configuration and Effects of the Invention> In order to achieve the above object, the present invention includes three or more two-dimensional imaging means for imaging an object from at least three directions;
means for receiving video signals from each imaging means and generating grayscale images of an object viewed from each direction; means for generating binary images by binarizing the data of the grayscale images; A stereoscopic viewing apparatus is configured by means for extracting feature points and means for associating the feature points in each binary image and calculating the three-dimensional coordinates of the corresponding points on the object.

According to the stereoscopic viewing device of the present invention, complicated or time-consuming processing such as edge coercion processing or slit light scanning is not required, so it is possible to realize faster processing and lower calculation costs, and also to reduce noise. This invention has remarkable effects that achieve the purpose of the invention, such as being less susceptible to influence and improving the accuracy of object recognition.

<Description of Embodiments> FIG. 1 shows a stereoscopic viewing apparatus of the present invention using three television cameras IA, IB, and Ic. Each of the above TV cameras IA
, IB, and IC are for capturing images of a three-dimensional object from three different directions, and use, for example, a two-dimensional surface imaging device. Each TV camera IA,
An image input section 2 having a synchronization signal generation circuit and a video signal amplification circuit is connected to the output side of IB and Ic, and this image input section 2 is connected to each television camera IA, IB, and Ic.
It receives video signals from c and generates grayscale images of the object viewed from each direction. The grayscale image G shown in FIG. 7 from one direction is shown as an example.

Each grayscale image is input to a binarization unit 3 consisting of, for example, a floating binarization circuit, and in this binarization unit 3, the data of each grayscale image is binarized to generate a binary image. .

Each of these binary images is then input to the feature point extraction unit 4,
For each image, points corresponding to the corners of the object on the image are connected to a contour tracing unit 41 that obtains feature sequence data, and a line drawing processing unit 42 that converts the binary image into a line drawing based on the contour point series data. The feature points are extracted from the intersections of each of the 100 lines that make up the line drawing, and the data is output.

Figure 8 il+ +21 +31 shows the line drawing images H, HB, and Ho in three directions, respectively.For example, in each image, points PA, PB, and Po overlap the same object point on the object. Shows feature points.

In this way, each feature point extraction unit 4 outputs feature point data for each image, and the control unit 6 outputs each television camera IA,
Camera parameters such as the positions and orientations of 1B and 1c, as well as the imaging distance of the lenses, are respectively taken into the stereoscopic vision section 5. For example, in the stereoscopic vision section 5 shown in FIG. 3, the corresponding point determination section 51 performs correspondence between the feature points, and the triangulation unit 52 calculates the three-dimensional coordinates of the corresponded points on the object using a known triangulation method.

FIG. 9 is a principle diagram illustrating the operation of the correspondence between feature points in the corresponding point determination unit 51, and shows how each television camera IA, I
B, Image HA on IC,)iB.

Ho (hereinafter, third image HA, second image HB, third image H
The feature point P for the object point P on the object
A, PB, and Po are scratched. Also, the third image H
6, there is an image of a straight line FAPA connecting the focal point FA of the first camera IA and the feature point PA (this straight line image is called an epipolar line)! ! Similarly, on the third image H6, the epipolar lines 13 and m3 of straight line FAPA and straight line FBPB are set.

Figure 10 ill +21 +31 is each of the above images HA, H
B, indicates Hc. According to the figure, the feature point PB in the third image H6 is located on the epipolar line 12, and the feature point PB in the third image H6 is located on the epipolar line 12.
Feature point P in image H6. is the epipolar line ls
Located on the intersection of r m3. From this, the feature point PA
, PB, and Po are points that correspond to each other as images of the object point P, and therefore the three-dimensional coordinates of the object point P are straight lines FAPA, FBPB, and FoP.

It can be found as the intersection of Furthermore, Figure 10 +21
+31 also indicates the feature points Rn and Rc of other object points located on the extension of the straight line FAP in FIG.
In this case, the feature point R6 is not located on the intersection of the epipolar line/3m3.

FIGS. 4 and 5 show an example of the feature point extracting section 4. FIG. The embodiment shown in FIG. 4 includes a run-length encoding unit 43 that encodes the starting points of white pixels and black pixels and their continuous amount for each row constituting a binary m-image, and a run-length encoding unit 43 that analyzes the run-length code of each row. It also includes a run-length code analysis section 44 that extracts feature points.

Further, in the embodiment shown in FIG. 5, a binary image is scanned with a mask of a predetermined size, and a partial pattern within this mask is compared with a reference partial pattern read out from a memory by a partial pattern selection section 46 in a matching section 45. Feature points are extracted by sequential comparison.

The operations of the image input section 2, binarization section 3, feature point extraction section 4, and stereoscopic viewing section 5 are controlled in series based on control information issued by a control section 6, and are particularly shown in FIG. In the embodiment, video signals from three television cameras IA, IB, and IC are sequentially switched, and corresponding processing is executed in each unit of a single configuration.

FIG. 6 shows three image input units 2A, 2B, 2C, binarization units 3A, 3B, 3C, feature point extraction units 4A, 4 corresponding to each television camera LA, IB, IC, respectively.
This shows another embodiment in which circuits B and 4C are provided. Although this embodiment has a complicated circuit configuration, the processing proceeds in parallel, so it is possible to speed up the processing.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a stereoscopic viewing device according to the present invention, FIG. 2 is a block diagram showing an example of the configuration of a feature point extracting section, FIG. 3 is a block diagram showing an example of the configuration of a stereoscopic viewing section, FIG. FIG. 5 is a block diagram showing another embodiment of the feature point extraction section, FIG. 6 is a block diagram showing another example of the device according to the present invention,
Fig. 7 is a diagram showing an example of a grayscale image, Fig. 8 fil T2
1431 is a diagram showing a line-drawn image, and FIGS. 9 and 10 are diagrams for explaining the principle of the bone correspondence method between feature points. IA, IB, IC...Television camera 2...Image input unit 3...Binarization unit 4...Feature point extraction unit 5...Stereoscopic vision unit Patent applicant Tateishi Electric Co., Ltd.→l Figure -〉A+q T2, hqz卜0ζ入-1 に Ｇｕｎｔ？ -y+7 Kuchitake! 3 countries (2nd cultivation q country dull 10 figures

Claims (1)

[Claims] three or more two-dimensional imaging means for capturing images of an object from at least three directions; means for receiving video signals from each imaging means to generate gray scale images of the object viewed from each direction; A means for binarizing to generate a binary image, a means for extracting feature points for each binary image, and a means for associating the feature points in each binary image and calculating the three-dimensional coordinates of the corresponding point on the object. A stereoscopic viewing device comprising means.