JPS6049475A - Object detecting method - Google Patents

Abstract

PURPOSE:To detect a 3-dimensional object by irradiating a slit pattern to the object and picking up the object from two directions to obtain a virtual straight line connecting a point on the slit pattern and an optical center and to obtain an intersecting point between the virtual straight line and a slit luminous flux surface as well as the correspondence to a point on an image pickup surface. CONSTITUTION:An object to be detected is irradiated by plural pieces of slit patterns sent from a light source 1 and picked up by the 1st and 2nd cameras 5 and 6 set in different directions. The features of this picked-up image are extracted by a memory. Then an intersecting point between a virtual straight line and a slit plane is obtained by the 1st means. In this case, an equation of the slit plane is previously stored in a CPU. These data are stored in a data memory 38, and a corresponding relation detecting part 40 detects the corresponding relation with the 2nd picture features. Then a 3-dimensional position is detected by a position/form recognizing part 41.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention relates to industrial equipment that handles three-dimensional objects, such as assembly and welding robots.

The present invention relates to an object detection method that recognizes shapes and the like.

Conventional configurations and their problems In recent years, industrial robots have become widespread, and increasingly sophisticated functions are required.

Visual functions using industrial TV cameras are one example of this, and visual feedback is expected to make robots more intelligent.

However, to date, most of the methods that have reached the practical level mainly utilize two-dimensional information, and the recognition of three-dimensional likely objects, which is the most important target for industrial applications, is being actively researched. However, there are few practical examples.

The general conventional method for recognizing the position, shape, etc. of a three-dimensional object is to image the object from different angles using two TV cameras and view it stereoscopically with both eyes. However, there are the following problems.

(1) A huge amount of calculation is required to extract features such as edges and corner points from an input image having gradations.

(2) It is necessary to determine the correspondence between two characteristic line points on both the left and right screens, which is difficult.

Problem (2) is a problem specific to binocular stereoscopic viewing, and cannot be uniquely fitted without prior information on the shape of the target object.

Therefore, as a solution to the problems (1) and (2) above, methods using slit light have been researched in the past.

FIG. 1 shows a principle diagram of a conventional distance measuring method using slit light. In the example shown in the figure, a slit light beam 2 emitted from a slit light source 1 is deflected by a rotating mirror 4,
The target three-dimensional object 3 is irradiated with light, and the reflected light is imaged by the TV camera 5.

The three-dimensional position of the slit light on the object can be determined from the angle information of the slit light beam surface and the position information (angle, distance, etc.) of the camera. In order to obtain distance information about the entire three-dimensional object 3, the rotating mirror 4 scans the slit light sequentially and measures the distance each time.

In the method shown in the figure, since the input image is a slit pattern, a binarized image can be easily obtained, and feature points such as refraction points of the slit pattern can be easily extracted. Furthermore, three-dimensional distance information can be uniquely obtained from the rotation angle of the mirror and the setting angle of the camera, and this is a method that can solve both of the above-mentioned problems (1) and (2).

However, this conventional method still has the problem that image input takes time. In other words, since only one slit pattern can be obtained by one image input, in order to obtain distance information over the entire three-dimensional object, the mirror 4 must be rotated and the slit pattern must be scanned sequentially. Image input must be performed for each step, and from a practical standpoint, the input time is a major problem. If the entire three-dimensional object is irradiated with multiple slit patterns, the number of image inputs is only required once, but since there are multiple slit light beam surfaces, the slit patterns in the input image and , it is difficult to determine the correspondence with the corresponding slit light beam surface, and distance information cannot be determined reliably.

Purpose of the Invention The present invention significantly shortens r- during image input, which was a problem in the conventional methods for determining the position and shape of a three-dimensional object as described above, and provides a practical three-dimensional object. The purpose is to provide a detection method.

Structure of the Invention The present invention irradiates an object to be detected with a plurality of slit patterns from a slit light source, images the object to be detected from two different directions by first and second image input devices, and 3. Select any one point on the slit pattern imaged on the imaging surface of the first image input device, and connect the selected point to the optical center of the lens system of the first image input device. Determine the first virtual straight line in the dimensional space,
A first intersection point between the first virtual straight line and any one of a plurality of slit light flux planes irradiated from the slit light source, and a distance between the first intersection point and the second image input device. a second virtual straight line in a three-dimensional space connecting the optical center of the lens system, and a second intersection between the second virtual straight line and the imaging surface of the second image input device. Therefore, from the imaging pattern on the imaging surface of the second image input device at the second intersection, a correspondence relationship between the second intersection and the selected point on the imaging surface of the first image input device is determined. This is a method for detecting a position of a three-dimensional object, characterized in that the three-dimensional position of a point on the surface of the object to be detected corresponding to the selected point is detected by detecting the selected point.

DESCRIPTION OF EMBODIMENTS The principle of the present invention will be explained in detail with reference to the drawings.

FIG. 2 is a diagram showing an example of the configuration of the image input section of the present invention. From a slit light source 1, a plurality of slit light beam groups 2 are
Irradiate the dimensional object 3. As a result, a slit pattern group 7 is formed on the surface of the three-dimensional object 3 and on the floor surface. This is carried out by the first TV camera 5 and the second T.
An image is taken with the V camera 6. FIG. 3a shows an example of an input image (right screen) from the first camera 5, and FIG. 3 shows an example of an input image (left screen) from the second camera 6.

FIG. 4 is a diagram showing the relationship between the camera optical system, the imaging plane, and the input image. Reference numeral 11 represents the lens system of the camera, and point E12 is the optical center (hereinafter referred to as the viewpoint y) of the camera lens system 11.

When a three-dimensional object 3 placed in a three-dimensional space is imaged by a camera, an image as shown in the figure is obtained on the imaging surface 13 of the camera. Ll is the distance between the viewpoint E12 and the imaging surface 13. Normally, the imaging plane 13 is located near the focal plane of the lens system 11. In order to observe the image formed on the imaging surface 13, it is necessary to output the imaged pattern to another image surface 14. The image surface 14 can be any arbitrary object such as a CRT screen or a photograph, but in order to process image data using a computer, it is common to record and store the input image in an image memory. , here, an image memory is considered as the image surface 14.

The imaging pattern on the imaging surface 13 and the input image on the image memory 14 are determined based on the relationship between the amount of parallel movement and the enlargement ratio between the X-Y plane on the imaging surface and the X/Y/plane on the image surface. , transformation matrix T0 (TO is a 3×3 matrix)
, linear coordinate transformation is possible as follows.

(X, Y, 1) = (X'ly', 1) - ToTo
can be determined in advance based on the design specifications of the camera and image memory, or analysis of input image data.

Reference numeral 15 denotes a virtual imaging surface obtained by moving the imaging surface 13 to a point-symmetrical position with respect to the viewpoint E12.

On the other hand, when considering that the origin of the X-Y plane on the imaging surface 13 and the corresponding point on the image plane (the origin if there is no parallel movement) are both on the optical axis of the camera, This is a virtual image surface obtained by moving the image surface 14, which can be seen in a three-dimensional space due to the relationship between the amount and the magnification ratio, to a point-symmetrical position with the viewpoint E12 as the point of symmetry. L2 is
This is the distance between this virtual image plane 16 and the viewpoint E12.

L2 can be determined from the enlargement ratio P between the image plane 16 and the imaging plane 15 as L2=L1×P. Imaging planes 13 and 16.
Since the image planes 14 and 16 are in point-symmetrical positions with respect to the viewpoint E12, the coordinates of the intersection between the line of sight connecting a point on the three-dimensional object 3 and the viewpoint E12 and the imaging plane or image plane. It is equivalent when considering. (However, the coordinates and treatment in the Z-axis direction are different.) Therefore, in order to simplify the explanation, the following explanation will be made assuming 16 or 16 as the imaging plane or image plane.

FIG. 5 is a diagram for explaining the configuration and principle of the present invention. In the figure, 0 is the coordinate origin in absolute space;
It is set on the floor surface on which the three-dimensional object 3 is placed. El,
F2 is the first . This is the optical center (view point) of the lens system of the second TV cameras 5 and 6. 21 and 22 are respectively the 1st. This is the imaging surface of the second TV camera. However, as explained in FIG. 4, the imaging plane considered here is a plane obtained by moving the actual imaging plane to a point-symmetrical position with the viewpoint as a point of symmetry. 11 and 12 are distances between the viewpoints of the first and second TV cameras and the imaging planes, respectively. Instead of the imaging planes 21, 22, one may consider an image plane 16 in FIG. 4.

However, in that case, for 11 and 12, the distance between each image plane and the viewpoint must be considered.

01 is the origin of a hypothetical x1-Y1-21 coordinate system (the z1-axis direction is the direction of 0l-El) on the imaging surface 21. o2 is X2-F2- assumed on the imaging surface 22
This is the origin of the Z22 coordinate system (the Z2 axis direction is the direction of 02-F2). In the example shown in the figure, the optical axis of the first T-camera 6 passes through the origin 0, and α with respect to the x-Z plane of the absolute coordinate system.
The second camera 6 is installed at an angle of θ with respect to the IFYZ plane, and the second camera 6 is installed with its optical axis passing through the origin 0 and at an angle of α2 with respect to the x-Z plane of the absolute coordinate system. It is said that

Origin 0 of the absolute coordinate system and each electric point E1? Assuming that the distances from E2 are dl and d2, the above camera system setting conditions (α1.α2.θr11+12+d1 rd2 are hereinafter referred to as camera parameters) define the two cameras 6.6 in absolute space. When installing the device inside the device, it is possible to determine in advance by analyzing the actually measured input image and the reference input image. Therefore, from now on, it will be treated as known data.

If the above camera parameters are known, x on the imaging surface 21
From the transformation matrix T1 from the l-yl-z1 coordinate system to the absolute coordinate system, and from the absolute coordinate system, x2- on the imaging plane 22
The transformation matrix T2 to the T2-Z2 coordinate system can also be uniquely determined. Generally T1. T2 can be expressed as a 4×4 matrix. T1 in the configuration example of FIG. 5 is illustrated below.

(X+3't"t1') = CX1+Y1+Z1+')
-T1T1. T2 shall also be determined in advance.

FIG. 6 also shows three slit/luminous flux planes 2-1.2-2.degree. 2-3 irradiated from the slit light source 1. The position of each slit light flux plane in the three-dimensional space can be uniquely determined from the design specifications of the slit light source 1, the set position and irradiation angle of the slit light source 1 in the three-dimensional space (hereinafter referred to as light source parameters). Also, the position of the slit light source 10 in three-dimensional space,
It is also possible to determine the position of each slit beam plane by measuring the slit pattern image formed on the floor surface when the object 3 is removed. By the method described above, the expression S1. S2. Regarding S3, we will also consider it as known data in advance. Now, after making the above preparations, we will proceed to the first step. When inputting the slit pattern image formed on the surface of the three-dimensional object 3 by the second TV right camera 6, a slit pattern as shown in FIG. 5 is obtained on the imaging plane 21.22. .

Assume now that a point P on the imaging surface 21 in FIG. 6 is selected as a feature point, considering the refraction point of the slit pattern.

In the example shown in the same figure, in order to avoid complicating the drawing, there are three slit light flux surfaces and the target object has a simple shape, so Pi is the slit formed by the second slit light flux surface 2-2. - Humans can easily tell that this is a point on the pattern, but in order to confirm with a computer that it is formed by the slit light beam surface 2-2,
Kanashi data processing is required. More generally, due to position detection accuracy, in practice it is normal to irradiate a large number of slit beams, and if the shape of the three-dimensional object becomes complex, the formation of The slit butter/ becomes like a broken line, so
It becomes impossible to uniquely determine which slit light beam surface forms the selected feature point.

Therefore, in general, it must be considered that it is unknown which slit light flux surface Pi is formed by. Therefore, in the present invention, from the image data obtained by binocular stereoscopic viewing using the two TV left cameras, it is determined which slit light beam surface P is formed by which slit light beam surface.

In absolute space, consider a first virtual straight line 23 connecting the feature point Pi on the imaging surface 21 and the viewpoint E1 of the first camera, and aim for the intersection of the extension of this straight line 23 and the slit light flux plane. Ru. At this time, even if we consider an arbitrary image plane corresponding to the image plane 16 in FIG. Since it always passes through a point corresponding to Po, the same result can be obtained as the intersection of the straight line and the slit light flux plane.

If the intersections of the straight line 23 and the three slit light flux surfaces 2-1 to 2-3 are Pil and P 2tPi3 in FIG. 6, then "i"
One of Pi3 is a characteristic point on the surface of the three-dimensional object that is formed when the slit light flux plane intersects with the three-dimensional object 3, and is a true intersection. On the other hand, the remaining 2 of Pi1 to Pi3
One is a false intersection point, which is not a point on the slit pattern formed on the surface of the three-dimensional object 30.

Considering a second virtual straight line 24 connecting Sokote, P, 1 to Pi3 and the viewpoint E2 of the second TV camera 6 in absolute space,
Intersection point P between this straight line and the imaging surface 22 of the second TV camera
'11+P'>2+P' Add 13.

Here, in the example of FIG. 6, since ij:P12 is a true intersection, point P'1□ on the imaging plane 22 (left screen) is on the imaging plane 2
1 (right screen) corresponds to the point Pi on the right screen. Therefore, on the imaging surface 22, point P',2 is located on a slit pattern shape similar to the feature point P0.

However, since Pil and Pi3 are false intersections, points P′,1 and “i3” are on the imaging plane 22.
is generally not a point on the slit pattern.

As mentioned above, P,
is the refraction point of the slit pattern formed on the surface of the three-dimensional object 30 by the slit beam surface 2-2. The position can be uniquely determined from the absolute coordinate values of . Once the corresponding slit light flux surface is known, on the imaging surface 21, since a series of slit patterns including P are all formed by the same slit light flux surface, Pi can be determined on the imaging surface 21. For any point on the included slit pattern, its position in absolute space can be determined by finding the intersection between the first virtual straight line and the slit light flux plane 2-2.

Depending on the setting angle of the camera and the slit light source and the shape of the three-dimensional object, in rare cases, it may be an extension of the first virtual straight line connecting P□ and El, or an extension of the second virtual straight line connecting Plj and E2. It is also possible that there is another true intersection point. Even if there are two or more true intersections, it is generally possible to confirm the correspondence between them, but in rare cases there are combinations of intersection positions for which the correspondence cannot be uniquely determined.

In such a case, the correspondence is determined by a combination of characteristic points such as both end points of the straight line or several points on the straight line by a method such as linear approximation of the slit pattern image in the input image.

Furthermore, feature points that exist on the right screen may be hidden and cannot be seen on the left screen. In this case, the second
The correspondence relationship cannot be confirmed for any of the intersection points P'i1 to "i3. Therefore, it is determined that the feature point Pi is hidden on the left screen, and processing moves on to the next feature point.

An embodiment in which a three-dimensional object position/shape recognition device is configured by employing the object detection method of the present invention having the configuration described above will be described below. FIG. 6 is a block diagram of the same embodiment. In this embodiment, the input image data of both the left and right screens are processed by the CPU, and the position detection method according to the present invention is used to determine the correspondence between the left and right screens and the 3 Determine the absolute position on the surface of a dimensional object. Based on these data, the position and shape of the target three-dimensional object are recognized.

In FIG. 6, 1 is a slit light source, 6, 6 is a first . This is the second TV camera. 31.32 are the 1st.
The first one stores input images a and b from the second TV camera.
．． This is a second image memory.

33 is the CpU, and the following components are all connected to this CpU.
Operated by pU. 34 is a signal bus of the CRU. 42 is a camera parameter, 43 is a light source parameter, which is given as data in advance (35, 36 are the first and second feature extraction units, input images a', blc
On the other hand, feature data c and d such as feature points in the input image are extracted by conventional methods such as binarization, subdivision, and linear approximation.

37 is a means for finding the intersection of the first virtual straight line 23 and the slit light beam surface, and 37 is a means for finding the intersection point of the first virtual straight line 23 and the slit light beam surface,
The three-dimensional coordinates of the intersection point Pij of the first virtual straight line connecting the first TV left camera viewpoint E1 and the slit light flux surface Sj are determined. S, is the formula of the first slit beam plane stored in advance in the CpU 33, pi, s,
Instructions are given from the CPU. tl is a transformation matrix that performs coordinate transformation from the first image memory to the imaging plane of the first TV left camera, and T1 is xl-y, - taken on the imaging plane of the first TV left camera. This is a conversion matrix from the z1 coordinate system to the absolute coordinate system, and both are predetermined from camera parameters and the like and stored in the CpU 33. The three-dimensional coordinates of Pij determined by means 37 are stored in data memory 3.
82 becomes the input data of the means 39.

A data memory 38 stores the three-dimensional coordinates of Pii determined by the means 37 as data. When it is confirmed through subsequent processing that Plj is an intersection point (true intersection point) with the corresponding slit light beam surface, it is used as valid data representing the absolute position of the corresponding point on the three-dimensional object surface. 39 is means for determining the intersection between the second virtual straight line 24 and the imaging plane of the second TV right camera, and the intersection P1j with the slit plane determined by means 37 and the viewpoint of the second TV right camera. Find the plane coordinates on the imaging plane of the intersection point P',] between the second virtual straight line 24 connecting E2 and the imaging plane of the second TV right camera. T2 is a transformation matrix from the absolute coordinate system to the X2-T2-Z2 coordinate system taken on the imaging plane of the second TV right camera.

4o is a correspondence detection unit which converts the coordinate value of P′ obtained by the means 39 into a coordinate value 1 on the second image memory, and extracts the feature data obtained by the second feature extraction unit 36. Check whether the corresponding feature point exists in d. t2
is from above the imaging surface of the second TV right camera. This is a transformation matrix for coordinate transformation onto the second image memory.

41 is a position and shape recognition unit, and a correspondence detection unit 40
Based on the data of the correspondence between the feature points on both the left and right screens detected by the camera and the slit light beam surface, and the absolute position data of the feature points on the surface of the three-dimensional object stored in the data memory 38, etc. First, the position, shape, etc. of three-dimensional objects are recognized.

To further explain the details of each of the above components, 1. Regarding the second feature extraction unit 35, 36 and the position and shape recognition unit 41, many research publications have been made in the past, such as “Doi et al., “Recognition of 3D objects by laser beam cutting method”, measurement Proceedings of the Society of Automatic Control Engineers, v
ol 9. NO, 1゜PP18-21, (197
3) J, R Oshima et al., 'Special hardware for three-dimensional object recognition', Electronics Technology Research Institute, vol. 3
7. NO, 6, pP493-501.

(19ryo 3)” and other documents.

Regarding the series of operations and specific methods of the components in the means 37 to the corresponding position detection section 40''!, see FIGS. 7A and 7B.
This is shown in the flow chart below. The numbers on the right side of the flowchart are the numbers of the corresponding components in FIG.

As an initial condition, i=1. Start from the state where j=1. First, an arbitrary feature point P is selected from the feature data C on the first image memory 31.

Coordinate P of the Xl-Y1 coordinate system where Pi is taken on the imaging surface 21 of the first TV left camera using the coordinate transformation matrix t1
Convert to i-Pi×t1. The above operation is not necessary if an image plane corresponding to the first image memory is considered from the beginning instead of the imaging plane.

The coordinates of Pi in the Xl-Yl-Z1 coordinate system are Pi(xp
i ry ,, o), then convert this to the absolute coordinate system using the transformation matrix T1 of Pi to the absolute coordinate Pi=Pi×T
Convert to 1. On the other hand, the viewpoint E1 of the first TV left camera
The coordinates of the X1-Y1-21 coordinate system are El (0,021
1) Yes, similarly convert this into absolute coordinates E1=E1×T1. Next, in the absolute coordinate system, an equation e of the first virtual straight line connecting the above-mentioned P and El is determined. Next, the plane equation Sj of the first slit light beam plane, which is stored in advance as known data, and the above straight line 20 are made to coincide, and the absolute coordinates of the intersection point Pij are determined.

The practical method for performing the above series of operations is to determine the coordinate transformation and the solution of the simultaneous equations in advance in the form of literal expressions, and then use the known data of camera parameters and light source parameters, and
The three-dimensional coordinate values of the intersection points P are calculated by giving the coordinate values of i on the first image memory and the coefficients of the equation representing the slit light beam surface as data.

l ] In the above explanation, both Pi and El are converted to absolute coordinates and then the equation of the straight line connecting the two points is calculated, but p and E
It is also possible to express l in the xl-yl-z1 coordinate system, find the equation of the straight line connecting the two points in the same coordinate system, and then convert the equation of the straight line to the absolute coordinate system, and the result is You will get the same thing as .

After storing the absolute coordinates of Pi determined above in the data memory 38, Pij is converted from the absolute coordinate system to the X2-Y2-Z2 coordinate system taken on the imaging surface 22 of the second TV camera 6. By the transformation matrix T2, the coordinate “i” in the same coordinate system
j ""i j Convert to XT2. Since the coordinates of the viewpoint E2 of the second TV camera in the x2-Y2-Z2 coordinate system are E2 (0, 0, 12), the equation f of the second virtual straight line connecting Pi and E2 can be expressed as x2- In the Y2-Z2 coordinate system.

Next, by combining the equation f of the straight line and the equation Z2-o of the imaging surface of the second TV camera 6,
-22 Get the coordinate data in the coordinate system.

At this time, the Z2 axis coordinate value is always 0.

The above series of operations is also solved in advance in the form of character expressions, and by substituting the camera parameters (known data) and the absolute coordinates of P (unknown data), the intersection point p/'t, X
2. A practical method is to calculate the Y2 coordinate value.

Next, the intersection point p/, j is transformed into the coordinate value P'1j=P'1jXt2 on the second image memory using the transformation matrix t2. If the image plane taken on the second image memory is considered from the beginning in place of the imaging plane 22, the above conversion is not necessary. The correspondence detection unit 40 detects the P determined above.
By comparing and examining the coordinate data of 'ij and the contents of the feature data d obtained by the second feature extraction section 36, feature points Pi and Check whether feature points of the corresponding pattern shape exist.

Regarding the specific method of the correspondence detection unit, research reports have been made for a long time (such as the above-mentioned documents), and it is not new. However, by adopting the method of the present invention, the positions of candidate points on the image memory that should be examined for correspondence can be determined meaningfully, thereby reducing data processing time and making unique decisions. This gives rise to the advantage of being able to

As a result of checking the correspondence at the positions of p/, .l] If no correspondence is detected, J-1+1 is performed and the routine moves on to check the relationship between Pi and the next slit light flux surface. If the correspondence cannot be confirmed even after checking up to the last slit beam surface, the feature point Pi is registered as being hidden on the left screen, and i =
As i +1, the routine moves on to examine the next feature point. p
If the correspondence is confirmed at the position /i, the correspondence between Pi and the corresponding slit beam surface Sj is registered, and then i=1-1-1 is set and the process moves to the next feature point. If there is a possibility that there are two or more true intersections on the extension of the first virtual straight line 23, leave i as is and set j=j+1, and also calculate the intersections between Pi and the remaining slit light flux planes. All you have to do is continue to consider the correspondence relationship. Once the correspondence relationships and three-dimensional positions have been detected for all necessary feature points, control is transferred to the position and shape recognition unit 41,
Based on the above data, the position and shape of the target object are recognized.

Note that on the first image memory, a series of slit patterns including feature points Pi whose correspondence has been confirmed are all formed by the same slit light beam surface, so Since the corresponding slit light flux plane for any point is known in advance, the three-dimensional position can be directly determined using only the means 37.

The setting conditions (position, angle, etc.) for the two cameras and the slit light source shown in the example in Fig. It is also possible to place However, by properly arranging the two cameras and the slit light source,
It is desirable to carefully select the setting conditions for the two cameras and the slit light source, since this provides advantages such as improved position detection accuracy and simplified coordinate value calculation. In addition, in the present invention, two cameras are used for the crab.

Of course, a method of sequentially inputting images at 75 different angles using a single camera using a mirror or the like can also be adopted. It is also possible to employ a mesh pattern (grid light) as the slit light. If necessary, a part of the present invention can be configured by software.

In the configuration of the embodiment shown in FIG. 6 and the flow charts shown in FIGS. 7A and 7B, each processing step is divided into blocks, but in the actual configuration, 37 to 39 are shown as a series. Xl, Y1 coordinates of the selected feature point Pi,
By inputting the number j of the slit beam surface, directly p
/i, on the second image memory coordinates (x21 Y
2) Since the calculation processing time can be shortened by configuring the feature points so as to increase A slit formed on the curved surface of a three-dimensional object.
Even if it is a point on a pattern, its 3D position information,
Effects of the Invention As described above, the object detection method of the present invention has a simple configuration and can be applied to a wide range of objects. , it is possible to solve the problem of matching the left and right image areas when viewing stereoscopically with both eyes, which was difficult with conventional methods, and also to directly determine the 3D positions of feature points on 3D objects. . By solving the problem of correspondence, n slit images can be
What used to require multiple image inputs and preprocessing for feature extraction can now be reduced to just one time, resulting in a significant reduction in image input and preprocessing time. Furthermore, since the subsequent data processing requires only simple calculations, the present invention can be expected to have a great industrial effect in the sense that it can realize a practical three-dimensional object recognition device.

[Brief explanation of the drawing]

Fig. 1 is a principle diagram showing a conventional distance measuring method using slit light, Fig. 2 is a conceptual diagram showing an image input section in an embodiment of the object detection device of the present invention, and Figs. FIG. 4 is a conceptual layout diagram showing the relationship between the camera optical system, the imaging surface, and the input image, and FIG. 5 is a diagram showing the configuration of the same embodiment. 6 is a block diagram showing the configuration of the embodiment, and FIGS. 7A and B are flowcharts explaining the operation of the embodiment.
It is a chart. 1...Slit light source, 2...Slit light flux surface group, 5.6...1st. second image input device, 21.22...first. Imaging surface of second image input device, 23°24...1st. Second virtual straight line, 37... Means for determining the intersection of the first virtual straight line and the slit plane, 38... Data meeting memory, 39... Number 2 virtual straight line and the 2nd TV
Means for determining the intersection with the imaging plane of the camera, 4o...
Correspondence detection section, 41...Position and shape recognition section. Name of agent Patent attorney Toshio Nakao and 1 other person 1st
Figure 2 Figure 3 Figure Name 4 Figure 7 Figure 7 (8)

Claims (1)

[Claims] irradiating the object to be detected with a plurality of slit patterns from a slit light source, imaging the object to be detected from two different directions by first and second image input devices,
Selecting any one point on the slit pattern imaged on the imaging surface of the first image input device, and connecting the selected point and the optical center of the lens system of the first image input device. Aim at a first intersection point between the first virtual straight line and any one of a plurality of slit light flux planes irradiated from the slit light source. , a second virtual straight line in a three-dimensional space connecting the first intersection point and the optical center of the lens system of the second image input device; From the imaging pattern on the imaging surface of the second image input device at the second intersection, the second intersection with the imaging surface of the second image input device is determined. An elephant-dimensional object device characterized in that the three-dimensional position of a point on the surface of the object to be detected corresponding to the selected point is detected by determining the correspondence with the selected point on the imaging surface of an image input device. How to detect a good person.