JP7166607B2 - Method for manufacturing omnidirectional camera device, method for designing omnidirectional camera device - Google Patents

Description

The present invention relates to an omnidirectional camera device that acquires 360-degree images and a design method for the omnidirectional camera device.

In recent years, the performance of omnidirectional camera devices is expected in connection with applications to many new information systems such as disaster monitoring, transportation systems, video conferences, and tourism advertisements. An omnidirectional camera device is a camera that acquires an image of a 360-degree surrounding image. This is roughly divided into two types, one that uses the same technology as a stereo camera, and one that uses a refractive and reflective optical system that uses a reflecting mirror. An omnidirectional camera with a refractive and reflective optical system is installed vertically below a convex mirror facing upwards. Light rays entering from the outside are reflected by the rotationally symmetrical convex mirror and focused on the viewpoint of the camera to obtain an image of the surroundings. It is common to An omnidirectional camera with refractive and reflective optical system using a reflecting mirror can observe the surrounding image in real time, so it is suitable as a visual sensor for mobile robots that perform remote control and autonomous action. This is because cost reduction is desired because only one camera is used, and remote control and autonomous movement can be performed efficiently because images of remote locations are acquired in real time.

Many of the reflection mirrors of conventional omnidirectional cameras have a bilobed hyperboloid shape as in Patent Document 1 and Patent Document 2, and acrylic or glass is used for the support structure so as not to be reflected on the reflection mirror. There are many. A curved surface with rotational symmetry in the vertical direction is used for the reflecting mirror, but the two-leaf hyperbolic mirror has a wide field of view in the side direction like a conical mirror, and also has a wide field of view at the feet. In addition, when a two-leaf hyperboloid mirror is used, the focal point is on the extension line of the point reflected on the mirror surface of the two-leaf hyperboloid from the three-dimensional coordinates of the object. Projection conversion can be easily performed.

JP-A-2002-334322 JP 2005-338654 A

The two-leaf hyperbolic mirror, which is used in many omnidirectional camera devices with refractive and reflective optical systems, cannot focus on the whole, has a limited imaging range, and has an extremely large amount of information in the peripheral area, but the central area. It is pointed out that there are few drawbacks. In addition, the bilobed hyperboloid shape is very expensive due to the difficulty of manufacturing, making it difficult to use easily.

An omnidirectional imaging system using a simple hemispherical mirror instead of a bilobed hyperbolic mirror has been proposed as a cost measure, but in this case, there is a part on the periphery of the hemispherical mirror where the projected image cannot be recognized. .

When an omnidirectional camera is used as an input device for a mobile robot or the like, a small one is useful in selecting a place where it can be installed. It is therefore preferred to make it as compact as possible.

SUMMARY OF THE INVENTION It is an object of the present invention to provide an omnidirectional camera device that is less expensive and less distorted than using a two-leaf hyperbolic mirror.

The omnidirectional camera device of the present invention acquires an omnidirectional image by capturing an image reflected by a mirror having a mirror-finished curved surface with a camera module installed facing the mirror. The curved surface is characterized by being less than half a spherical surface.

Since the omnidirectional camera device of the present invention has a spherical mirror, it is easy to process and therefore inexpensive. In addition, since the distortion approaches a two-leaf hyperboloid, which is the opposite of a spherical surface, the distortion is reduced.

1 is a diagram showing the overall configuration of an omnidirectional camera system using a cut hemisphere of the present invention; FIG. FIG. 4 is a diagram for explaining cutting of the outer peripheral portion of the hemispherical mirror; FIG. 4 is a diagram for explaining a bileaf hyperboloidal mirror and a spherical mirror; FIG. 10 is a diagram showing an image and a distortion rate when the distance to an object to be photographed is 50 cm; FIG. 5 is a diagram showing changes in vertical distortion rate when the distance to the object is 50 cm. FIG. 10 is a diagram showing an image and a distortion rate when the distance to an object to be photographed is 100 cm; FIG. 10 is a diagram showing changes in distortion rate in the vertical direction when the distance to the imaging target is 100 cm; FIG. 10 is a diagram showing an image and a distortion rate when the distance to an object to be photographed is 150 cm; FIG. 10 is a diagram showing changes in the distortion rate in the vertical direction when the distance to the imaging target is 150 cm;

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described in detail based on the drawings. It should be noted that the present invention is not limited only to this embodiment.

FIG. 1 is a diagram showing the overall configuration of an omnidirectional camera system. A scene around the camera is transformed into an image reflected on a spherical surface by the spherical mirror 1 . The spherical mirror 1 is connected to a camera module support means 3 having a camera module 10 housed in its bottom via a mirror/camera module connection means 2 which transmits the surrounding image. The camera module supporting means 3 has an opening at the center thereof, through which the camera module 10 captures an image reflected on the surface of the spherical mirror 1. - 特許庁In this case, depending on the type of camera module, it is possible to acquire images such as still images, moving images, infrared images, etc., but an imaging device such as the tip of a fiberscope may be mounted. These devices as a whole are constructed above the installation means 4 and can be easily installed in various places and machines via the installation means 4 . Also, the installation means 4 is provided with a wiring hole 5 for easy connection of a video signal cable from the camera.

The spherical mirror 1 is generally formed by cutting the outer peripheral portion of a hemispherical mirror made of mirror-polished stainless steel or the like. FIG. 2 is a diagram for explaining the cutting of this hemispherical mirror. The mirror-finished hemispherical mirror is denoted by 7 in FIG. 2, the camera module installed facing it is denoted by 10 in FIG. At this time, the tangent line of the spherical mirror 7 reaching the camera module 10 is represented by 9 in FIG. 2, and the shaded portion 8 in FIG. Thus, the curved surface of the spherical mirror 1 is less than half the spherical surface. Also, the vicinity of the contact point is an outer peripheral portion where optical distortion is large and an unclear image is projected. The spherical mirror 1 removes a predetermined portion of the outer peripheral portion where a blurred image is projected. The size of this portion may be predetermined geometrically based on the angle of view and focal length unique to the camera module 10 and the radius of the spherical mirror 1 .

As described above, many existing omnidirectional camera devices use a two-leaf hyperboloid as the shape of the mirror. And the layout of the camera module when using a two-leaf hyperboloid mirror is also established. Here, in an omnidirectional camera device in which the arrangement of camera modules is determined on the premise of using a two-leaf hyperbolic mirror, the optimum shape of the spherical mirror will be described when changing the mirror to a spherical surface. FIG. 3 is an xz plan view for explaining a two-leaf hyperboloidal mirror and a spherical mirror. The two-leaf hyperboloid is represented by the dotted line 11 in FIG. 3 and is obtained by rotating the hyperbola represented by the following equation (1) along the z-axis.
(x/a) ² −(z/b) ² =−1 (1)
However, a is half of the imaginary axis (conjugate axis) and b is half of the real axis (principal axis). Also, c satisfies the relationship of c=(a ² +b ² ) ^1/2 . In this case, the lens of camera module 10 is centered at OC and _OM is the focal point of the _bileaf hyperbolic mirror. A solid line 13 in FIG. 3 is an asymptote of the bilobed hyperboloid.

A solid line 12 in FIG. 3 represents the spherical surface of the spherical mirror. Similarly, the curved surface of the spherical mirror can be obtained by rotating part of the following equation (2) along the z-axis.
x ² +(zh) ² =r ² (2)
If r' is the radius of the outer circumference of the spherical mirror (the radius of the circle on the cut surface after cutting the hemisphere), the value of r'/r is Takes values from 0.8 to 0.93. The center of this spherical surface is on the principal axis of the camera, and the distance from the origin O is h.

The following equation (3) is obtained by transforming equation (1) with respect to z, where z _hyp is the z-coordinate value of the bilobed hyperboloid at each x-coordinate.
Z _hyp =b(1+(x/a) ² ) ^1/2 (3)
On the other hand, the following equation (4) is obtained by transforming equation (2) with respect to z, where z _sph is the z-coordinate value of the spherical surface on the x-coordinate.
Z _sph =−(r ² −x ² ) ^1/2 +h (4)

この場合の積分範囲はｘ軸の原点から球面ミラーの外周円の半径ｒ’の値が適用される。このＥＲＲＯＲＳＩＺＥを最小にするように球面の半径ｒと球面の中心と原点Ｏの距離ｈを決定すればよい。 Letting ERRORSIZE be the integral of the difference between the positions of the two-leaf hyperboloid and the spherical surface on the xz plane, it is obtained by the following equation (5) using z _hyp and Z _sph .

In this case, the range of integration applies from the origin of the x-axis to the radius r' of the outer circumference of the spherical mirror. The radius r of the spherical surface and the distance h between the center of the spherical surface and the origin O should be determined so as to minimize this ERRORSIZE.

Since the two-leaf hyperboloid parameters a and b have already been determined, the distance _c between the origin O and the lens center OC of the camera is determined. The sum of the distances h and _c minus the radius r of the spherical surface is the distance between the lens center OC of the camera module on the principal axis and the surface of the spherical mirror.

The distance between the spherical mirror 1 on the principal axis and the camera module 10 is adjusted by the length of the mirror/camera module connection means 2 .

The spherical mirror 1 designed in this manner has a wider imaging range than a two-leaf hyperboloid mirror, is smaller than a hemispherical mirror, and has less image distortion. In addition, since the omnidirectional camera device of the present invention has a spherical mirror, it is easy to process, which reduces the cost.

The imaging range and distortion rate among the above effects will be described in detail with reference to FIGS. FIG. 4 is a diagram showing an image and a distortion factor when the distance to the object is 50 cm, and FIG. 5 is a diagram showing changes in the distortion factor in the vertical direction when the distance to the object is 50 cm. FIG. 6 is a diagram showing an image and a distortion factor when the distance to the object is 100 cm, and FIG. 7 is a diagram showing changes in the distortion factor in the vertical direction when the distance to the object is 100 cm. FIG. 8 is a diagram showing an image and a distortion factor when the distance to the object is 150 cm, and FIG. 9 is a diagram showing changes in the distortion factor in the vertical direction when the distance to the object is 150 cm.

The upper photographs of FIGS. 4, 6, and 8 are images taken with an omnidirectional camera device. (1) is an image when using a two-leaf hyperboloid mirror, (2) is an image when using a spherical mirror that minimizes ERRORSIZE, and (3) is an image when using a hemispherical spherical mirror. Blocks in which black and gray are alternately overlapped are arranged on the upper side of each image. This block is the object to be photographed, all of which are images of the same block, and each block has a width of about 16 cm and a thickness (height) of about 5 cm. When a bileaf hyperbolic mirror is used, the layers of each block appear thick. Conversely, when a spherical mirror is used, many blocks are reflected. In other words, the spherical mirror can image an object placed in the horizontal direction of the omnidirectional camera device in a wider vertical direction (the imaging range is wider) than the two-leaf hyperbolic mirror.

4, 6 and 8, the ratio of the thickness of the block that appears thickest among the blocks that could be imaged to 100% and the ratio of the thickness to the thickness of the other blocks was evaluated as the distortion rate (%). Below is the table. 5, 7, and 9 show changes in distortion rate for each number of stages. (1) is the change in distortion rate when using a two-leaf hyperboloid mirror, (2) is the change in distortion rate when using a spherical mirror that minimizes ERRORSIZE, and (3) is a hemispherical spherical mirror. It is the change in the distortion rate when In the case of a two-leaf hyperboloid mirror, the upper block is reflected thicker, so the third or fourth stage is 100%. In the case of a spherical mirror, since the lower block is reflected thicker, the first or second stage is 100%. In other words, the two-leaf hyperboloid and the spherical surface are distorted in opposite ways. ERRORSIZE is a parameter for obtaining a spherical surface close to a bilobed hyperboloid. Therefore, with the omnidirectional camera device of the present invention, the distortion is reduced by approaching a two-leaf hyperboloid, which is the opposite of a spherical surface. 5, 7, and 9 also show that the distortion factor is smaller in the spherical mirror when ERRORSIZE is minimized than in the hemispherical spherical mirror. The above effect can be obtained by making the sphere smaller than the hemisphere within a range close to the bilobed hyperboloid.

INDUSTRIAL APPLICABILITY The present invention is suitable for omnidirectional imaging, and is particularly applicable to image monitoring devices, vision systems for mobile robots, and the like.

1 Spherical mirror 2 Mirror-camera module connection means 3 Camera module support means 4 Installation means 5 Wiring hole 6 Main axis of camera module 7 Mirror-polished hemispherical mirror 8 Area that cannot be acquired by camera module 9 Hemisphere leading to camera module Mirror tangent line 10 Camera module 11 Bilobe hyperboloid 12 Spherical surface 13 Bilobe hyperboloid asymptote

Claims (3)

A method for manufacturing an omnidirectional camera device for acquiring an omnidirectional image by photographing an image reflected by a mirror having a mirror-finished curved surface with a camera module installed facing the mirror, comprising:
A spherical surface is used instead of a predetermined bilobed hyperboloid as the curved surface,
a and b are the parameters that determine the shape of the hyperbola (x/a) ² -(z/b) ² =-1 of the bilobed hyperboloid which is predetermined, r' is the radius of the outer circle of the mirror of the bilobed hyperboloid, Z _{Let hyp} be a point on the hyperbola, Z _sph be a point on the mirror, r be the radius of the sphere, h be the distance between the center of the sphere and the origin,
Z _hyp =b(1+(x/a) ² ) ^1/2
Z _sph =−(r ² −x ² ) ^1/2 +h

and when
Designed by a design method that determines the shape of the mirror by setting r and h so that the ERRORSIZE is smaller than when r = r' and h = b + r
A method of manufacturing an omnidirectional camera device, characterized by: A method for manufacturing an omnidirectional camera device according to claim 1 ,
A method of manufacturing an omnidirectional camera device, wherein the mirror has a shape in which r and h are determined so as to minimize the ERRORSIZE. A method for designing an omnidirectional camera device for acquiring an omnidirectional image by photographing an image reflected by a mirror having a mirror-finished curved surface with a camera module installed facing the mirror, comprising:
A spherical surface is used instead of a predetermined bilobed hyperboloid as the curved surface,
where a and b are predetermined parameters for determining the shape of the hyperbola (x/a) ² −(z/b) ² =−1 of the two-leaf hyperboloid, r′ is the radius of the outer circle of the mirror of the two-leaf hyperboloid, Let Z _hyp be the point on the hyperbola, Z _sph be the point on the mirror, r be the radius of the sphere, h be the distance between the center of the sphere and the origin,
Z _hyp =b(1+(x/a) ² ) ^1/2
Z _sph =−(r ² −x ² ) ^1/2 +h

and when
A method of designing an omnidirectional camera device, wherein the shape of the mirror is determined by determining r and h so that the ERRORSIZE is minimized.

Images