JP2003232622A - Device, method and program for obtaining three- dimensional information - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device, a method and a program for obtaining more precise three-dimensional information. <P>SOLUTION: In one state of implementation, the method for obtaining three- dimensional information provided here comprises an imaging means for taking an image of an object, a relatively moving means for relatively and continuously moving the object and the imaging means so that the image of the object can be taken from a plurality of visual points by the imaging means, a relative position detecting means for detecting the relative positions of the object and the imaging means from each visual point where the image of the object is taken, and a three-dimensional shape estimating means for estimating the three- dimensional shape of the object by using information on the image of the object taken from the plurality of visual points and the relative positions detected by the relative position detecting means. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a stereoscopic information acquisition device, a stereoscopic information acquisition method and a stereoscopic information acquisition program for acquiring stereoscopic information of a target object.

[0002]

2. Description of the Related Art A three-dimensional information acquisition device as a prior art is
It is disclosed in Japanese Patent Laid-Open No. 10-124704.

In this stereoscopic information acquisition apparatus, a hypothetical existing area is calculated using the boundary between the target object and the background on the image.

The hypothetical existence area is a cone-shaped area having a projection center of the camera as an apex and a boundary between a target object on the image and the background as a cross-sectional shape.

This cone-shaped area (hypothetical existing area) is described by a voxel model (a cubic model having a predetermined size).

The above-described processing is repeated by rotating the target object by a predetermined angle with a rotary table.

Then, the common hypothetical existence area is obtained and the stereoscopic information of the target object is acquired.

Here, the boundary between the target object and the background on the image is an image obtained by photographing only the background without the target object,
It is calculated from the difference between the images taken with the target object.

[0009]

In the prior art as described above, in order to improve the accuracy of stereoscopic information, it is necessary to reduce the above-mentioned predetermined angle, so that the number of times of photographing increases and the photographing time increases. There is a problem.

Further, since the photographing angle is set first, the rotary table is repeatedly moved and stopped, and the target object is subject to acceleration, and there is a possibility that the object may fall or be deformed.

Further, since the stop position needs to be controlled accurately, the moving device and the control device for the rotary table are complicated and expensive.

Further, when the boundary between the target object on the image and the background is obtained, the difference between the image of only the background is used. However, due to variations in exposure, focus, shutter speed, etc. for each photographing of the camera, the boundary can be accurately measured. There is a problem that you may not be asked.

Further, since the common hypothesis existing region is directly obtained using the voxel model, a large number of images are required to confirm whether or not a certain voxel is included in the common hypothesis existing region. There is a problem that the processing time becomes huge.

Further, since it is necessary to store the existence probabilities of all voxels of the voxel model set initially, there is a problem that the memory capacity becomes huge.

The present invention has been made in order to solve the above-mentioned problems, and it is possible to obtain a boundary with high accuracy, obtain more accurate stereoscopic information, and obtain stereoscopic information of a target object. It is an object of the present invention to provide a stereoscopic information acquisition device, a stereoscopic information acquisition method, and a stereoscopic information acquisition program that can significantly reduce the memory capacity of the device, shorten the shooting time, and keep the target object in a stable state.

[0016]

According to the present invention, in order to solve the above problems, (1) a photographing means for photographing an image of a target object, and the photographing means for photographing the image of the target object from a plurality of viewpoints. Relative movement means for relatively and continuously moving the target object and the photographing means so as to enable photographing, and the viewpoint at each viewpoint where images of the target object are photographed from a plurality of viewpoints by the photographing means. Relative position detecting means for detecting the relative position between the target object and the photographing means, images of the target object photographed from a plurality of viewpoints by the photographing means, and information on the relative position detected by the relative position detecting means. And a three-dimensional shape estimation unit that estimates the three-dimensional shape of the target object using the.

(Operation) According to the present invention, since photographing can be performed while the target object is moving, the photographing time is reduced by less acceleration / deceleration during movement, and the target object is less likely to fall or deform due to inertia. In addition, the energy consumption can be reduced.

Further, according to the present invention, in order to solve the above-mentioned problems, (2) a photographing means for photographing an image of a target object, having predetermined optical characteristics, and being arranged behind the target object. The background plate that is the background of the target object at the time of shooting, and the target object and the shooting means are relatively and continuously moved so that images of the target object can be shot from a plurality of viewpoints by the shooting means. Relative movement means for moving the image of the target object from a plurality of viewpoints by the photographing means, relative position detection means for detecting the relative position of the target object and the photographing means at each viewpoint, by the photographing means An image of the target object taken from a plurality of viewpoints and information of the relative position detected by the relative position detecting means are used to take an image by the imaging means. Stereoscopic shape estimating means for recognizing an area occupied by the target object in an image from a plurality of viewpoints and estimating a stereoscopic shape of the target object using the area occupied by the target object. An acquisition device is provided.

(Operation) According to the present invention, since the object can be photographed while moving by the blue back silhouette method and the constant speed movement, the acceleration / deceleration during movement is small,
It is possible to reduce the photographing time, reduce the fall and deformation of the target object due to inertia, reduce the energy consumption, and cut out more accurately.

Further, according to the present invention, in order to solve the above-mentioned problems, (3) a photographing means for photographing an image of the target object, and at least the entire contour portion of the target object within the photographing range of the photographing means or the same. Illuminating means for directly or indirectly illuminating a range including a part of the target object from behind, and the target object and the target object so that images of the target object can be captured from a plurality of viewpoints by the capturing means. Relative moving means for relatively and continuously moving the photographing means, and detecting the relative position of the target object and the photographing means at each viewpoint at which images of the target object are photographed from a plurality of viewpoints by the photographing means Relative position detecting means, images of the target object photographed from a plurality of viewpoints by the photographing means, and relative positions detected by the relative position detecting means. Position information, the area occupied by the target object in each of the images taken from the plurality of viewpoints by the imaging means is recognized, and the three-dimensional shape of the target object is estimated using the area occupied by the target object. The three-dimensional shape estimating unit is provided, and the three-dimensional information acquiring device is provided, which is characterized in that the illumination unit is turned on when an image used for recognizing an area occupied by the target object is captured.

(Operation) According to the present invention, since the back light silhouette method and the constant speed movement allow photographing while moving the target object, there is little acceleration / deceleration during movement.
It is possible to reduce the photographing time, reduce the fall and deformation of the target object due to inertia, reduce the energy consumption, and more accurately cut out regardless of the object color.

Further, according to the present invention, in order to solve the above-mentioned problems, (4) the relative moving means rotates the target object (2) or (3).
The described stereoscopic information acquisition device is provided.

(Operation) The present invention guarantees that at least information around the rotation axis can be obtained by rotation limitation.

Further, according to the present invention, in order to solve the above-mentioned problems, (5) the relative moving means moves the target object in a direction parallel to a rotation axis (4). A stereoscopic information acquisition device is provided.

(Operation) According to the present invention, by the spiral movement, it is possible to recognize a part which cannot be recognized by rotation.

Further, according to the present invention, in order to solve the above-mentioned problems, (6) the relative moving means rotates the target object at a constant angular velocity, and the relative position detecting means uses a reference angular position. A reference angular position detecting means for detecting the time difference, a time difference calculating means for calculating a time difference from the time when the reference angular position detecting means detects the reference angular position to the time when the photographing is performed by the photographing means, the constant angular velocity and the time difference. An angle difference specifying unit that specifies an angle difference rotated from the reference angle by the time when the image is taken by the image taking unit using the result obtained by the calculating unit (4). The three-dimensional information acquisition device is provided.

(Operation) The present invention, by rotating at a constant angular velocity,
Since no special angle detecting means is required, the device configuration becomes simple.

Further, according to the present invention, in order to solve the above-mentioned problems, (7) the image obtained at the time of photographing for estimating the three-dimensional shape of the target object by the three-dimensional shape estimation means is a contour of the target object. The stereoscopic information acquisition device according to (3) is characterized in that the region occupied by the target object is a relatively dark silhouette image compared to a background portion in the vicinity.

(Operation) In (3), the present invention provides an image for estimating a three-dimensional shape.

Further, according to the present invention, in order to solve the above-mentioned problems, (8) the three-dimensional shape estimating means has a luminance difference between a contour area of the target object and a background area around the target object in the silhouette image. The stereoscopic information acquisition device according to (7) is characterized in that the contour of the target object is extracted by utilizing, and the area occupied by the target object is estimated using the extracted contour.

(Operation) In the present invention, in (7), the shape cutting method is specified.

Further, according to the present invention, in order to solve the above-mentioned problems, (9) the illuminating means is arranged at the back of the target object with respect to the photographing means and a light source which emits light in at least a visible light region. And a light scattering means for scattering the light from the light source, and the stereoscopic information acquisition apparatus according to (3).

(Operation) In the present invention, the back illumination scattering is used in (3).

Further, according to the present invention, in order to solve the above-mentioned problems, (10) the three-dimensional shape estimating means is operated by the photographing means at least once when the illuminating means does not illuminate the target object. The stereoscopic information acquisition apparatus according to (3) is characterized in that the texture of the surface of the target object is estimated using the captured texture image of the target object.

(Operation) In the present invention, the texture is estimated in (3).

Further, according to the present invention, in order to solve the above-mentioned problems, (11) the three-dimensional shape estimating means sets a set of closed regions closely arranged in a three-dimensional space with respect to the target object. Based on the closed region setting means for setting and the image obtained by photographing the target object from a plurality of viewpoints by the photographing means, the closed region is calculated by calculating the probability that each closed region exists outside the target object. A closed region outside determination means for determining whether the closed region exists outside the target object, and the closed region outside determination means that the probability that the closed region exists outside the target object exceeds a predetermined threshold value. The stereoscopic information acquisition apparatus according to (2) or (3) is characterized in that, when the means determines, the closed area is excluded from the determination processing targets of the closed area outside determination means thereafter.

(Operation) In the present invention, in (2) or (3), the processing speed is remarkably increased by not processing the unnecessary portion such as the engraving by the external judgment.

Further, according to the present invention, in order to solve the above-mentioned problems, (12) the three-dimensional shape estimating means selects the first image from among the images photographed by the photographing means from a plurality of viewpoints.
While recognizing the area occupied by the target object based on the image captured with the line of sight from the viewpoint, the recognition estimation process for estimating the three-dimensional shape of the target object is performed, and then, from the first viewpoint, A recognition estimation process for recognizing an area occupied by the target object based on an image captured with a line of sight from a distant second viewpoint and estimating a three-dimensional shape of the target object is performed. Among the remaining viewpoints corresponding to images not used, among the viewpoints sandwiched by the maximum angle difference formed by the line of sight corresponding to each of the two images used in the recognition estimation process, The stereoscopic information acquisition apparatus according to (4) is characterized in that the recognition estimation processing is performed using an image from the viewpoint closest to the angle at which the angle difference is interpolated, and the processing is sequentially repeated.

(Operation) In the present invention, in (4), by considering the boundary image processing order, it is possible to more efficiently determine a portion where processing is unnecessary.

Further, according to the present invention, in order to solve the above problems, (13) the solid shape estimating means has a predetermined probability that a closed region belonging to the set of closed regions exists outside the target object. It becomes a probability range, and the boundary closed region determination means for determining whether or not the closed region exists near the boundary between the inside and the outside of the target object, and the closed region existing near the boundary by the boundary closed region determination means. Boundary closed area dividing means for dividing the closed area determined to exist into smaller divided closed areas, and the boundary closed area determination for the divided closed area divided by the boundary closed area dividing means. It is again determined by the means whether or not it is a closed region existing near the boundary, and based on this determination result, the boundary closed region dividing unit re-divides the divided closed region, and the boundary closed region determination unit makes a determination. The stereoscopic information acquisition apparatus according to (2) or (3) is characterized in that the processing of dividing by the boundary closed area dividing means is repeated until the closed area has a predetermined size.

(Operation) In the present invention, in (3), the boundary voxels are subdivided so that the accuracy is remarkably improved.

According to the present invention, in order to solve the above problems, (14) the target object in the image is occupied by using images of the target object photographed from a plurality of viewpoints and information on the viewpoint positions. A three-dimensional information acquisition method for recognizing a region, estimating the three-dimensional shape of the target object based on the region occupied by the target object, and acquiring the three-dimensional information of the target object, which is three-dimensional with respect to the target object. Based on a closed region setting step of setting a set of closed regions closely arranged in space and an image obtained by photographing the target object from a plurality of viewpoints, the probability that each closed region exists outside the target object is calculated. By determining, the closed region external determination step of determining whether the closed region exists outside the target object, and the probability that the closed region exists outside the target object exceeds a predetermined threshold value, and Closed area external format And a step of removing the closed region from the determination processing target of the closed region outside determination process thereafter, when the determination is made in the regular process.

(Operation) According to the present invention, the processing speed is remarkably increased by not processing the unnecessary portion such as the carving by the external judgment.

Further, according to the present invention, in order to solve the above-mentioned problems, (15) the target object in the image is occupied by using images of the target object taken from a plurality of viewpoints and information on the viewpoint positions. A stereoscopic information acquisition method of recognizing a region, estimating the stereoscopic shape of the target object based on the region occupied by the target object, and acquiring stereoscopic information of the target object, the image being captured from the plurality of viewpoints. A first recognition estimation processing step of recognizing a region occupied by the imaging target based on an image captured with a line of sight from a first viewpoint and performing a recognition estimation processing of estimating a three-dimensional shape; A second method of recognizing a region occupied by the target object based on an image captured with a line of sight from a second viewpoint farthest from the first viewpoint and performing a recognition estimation process of estimating a three-dimensional shape.
Of the remaining viewpoints corresponding to the images not used in the recognition estimation processing, and the angle difference formed by the line of sight corresponding to each of the two images used in the recognition estimation processing is maximum. Among the viewpoints sandwiched between the two, the third recognition estimation processing step for performing the recognition estimation processing using an image from the viewpoint closest to the angle for interpolating the angle difference. A stereoscopic information acquisition method is provided, which is characterized by sequentially repeating the third recognition estimation processing step.

(Operation) The present invention makes it possible to more efficiently determine a portion that does not require processing by considering the boundary image processing order.

Further, according to the present invention, in order to solve the above-mentioned problems, (16) the target object in the image is occupied by using images of the target object photographed from a plurality of viewpoints and information on the viewpoint positions. A three-dimensional information acquisition method for recognizing a region, estimating the three-dimensional shape of the target object based on the region occupied by the target object, and acquiring the three-dimensional information of the target object, which is three-dimensional with respect to the target object. A closed region setting step of setting a set of closed regions closely arranged in space, and a probability that a closed region belonging to the set of closed regions exists outside the target object becomes a predetermined probability range, and the target A boundary closed region determination step of determining whether or not it is a closed region existing near the boundary between the inside and the outside of the object, and a closed region determined to be a closed region existing near the boundary by the boundary closed region determination step. The area A boundary closed region dividing step of further dividing the closed closed region into smaller divided closed regions, and performing a process of executing the boundary closed region determination step and the boundary closed region dividing step on the divided closed region, A stereoscopic information acquisition method is provided, which is characterized in that is repeated until the size becomes a predetermined size.

(Operation) In the present invention, the accuracy is remarkably improved by subdividing the boundary voxels.

Further, according to the present invention, in order to solve the above-mentioned problems, (17) by using an image obtained by photographing a target object from a plurality of viewpoints by a computer and information on the viewpoint position, the target object in the image Recognizing the area occupied by, the three-dimensional information acquisition program for estimating the three-dimensional shape of the target object based on the area occupied by the target object, to acquire the three-dimensional information of the target object, the computer, Based on images of the target object captured from a plurality of viewpoints, a step for setting a closed region for setting a set of closed regions closely arranged in the three-dimensional space for the target object, A step for determining whether or not the closed region exists outside the target object by determining a probability that the region exists outside the target object; If the probability that a closed region exists outside the target object exceeds a predetermined threshold value, if it is determined in the step for the closed region outside determination, the closed region is
There is provided a stereoscopic information acquisition program, characterized in that it has a program for executing the following steps of the step for determining the outside of the closed region from the determination processing target.

(Operation) According to the present invention, the program is processed by the external judgment so that unnecessary parts are not processed so as to cut the engraving, thereby significantly increasing the processing speed.

According to the present invention, in order to solve the above problems, (18) the target object in the image is obtained by using images of the target object photographed by a computer from a plurality of viewpoints and information on the viewpoint positions. Recognizing the area occupied by, the three-dimensional information acquisition program for estimating the three-dimensional shape of the target object based on the area occupied by the target object, to acquire the three-dimensional information of the target object, the computer, The first of the images taken from the plurality of viewpoints
A step for a first recognition estimation process for recognizing an area occupied by the imaging target based on an image captured with a line of sight from the viewpoint, and performing a recognition estimation process for estimating a three-dimensional shape of the target object; A second step of recognizing a region occupied by the target object based on an image captured with a line of sight from a second viewpoint farthest from the first viewpoint, and performing a recognition estimation process of estimating a three-dimensional shape of the target object. For the recognition estimation process, and the angle of the line of sight corresponding to each of the two images used in the recognition estimation process, which is the remaining viewpoint corresponding to the image not used in the recognition estimation process. For a third recognition estimation process that performs the recognition estimation process using an image from the viewpoint closest to the angle for interpolating the angular difference among the viewpoints sandwiched by the ones with the largest difference Has a step, a program for execution, the three-dimensional information acquisition program characterized by sequentially repeating the steps for the third recognition estimation process is provided.

(Operation) The present invention provides a program that can more efficiently determine a portion that does not require processing by considering the boundary image processing order.

According to the present invention, in order to solve the above-mentioned problems, (19) the target object in the image is obtained by using images of the target object photographed by a computer from a plurality of viewpoints and information on the viewpoint positions. Recognizing the area occupied by, the three-dimensional information acquisition program for estimating the three-dimensional shape of the target object based on the area occupied by the target object, to acquire the three-dimensional information of the target object, the computer, A step for setting a closed region for setting a set of closed regions closely arranged in the three-dimensional space for the target object, and a closed region belonging to the set of closed regions exists outside the target object. The probability of being within a predetermined probability range, a step for boundary closed area determination for determining whether or not the area is a closed area existing near the boundary between the inside and outside of the target object, and the closed area determination A step of dividing the closed region, which is determined as a closed region existing near the boundary by the determining step, into smaller divided closed regions, and a program for executing the step of dividing the boundary closed region. Then, the processing for executing the step for determining the boundary closed area and the step for dividing the boundary closed area is repeated for the divided closed area until the closed area reaches a predetermined size. A stereoscopic information acquisition program is provided.

(Operation) The present invention has a program in which the accuracy is remarkably improved by subdividing the boundary voxels.

[0054]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of a stereoscopic information acquiring apparatus according to the present invention will be described below with reference to the drawings.

In the description of the embodiment of the three-dimensional information acquisition apparatus according to the present invention, mainly the movement of the target object will be explained, but the same can be realized by moving the photographing means.

(First Embodiment) FIG. 1 is a block diagram showing the configuration of a first embodiment of a stereoscopic information acquisition apparatus according to the present invention.

In the present embodiment, as shown in FIG. 1, a camera CU as a photographing means, a flash FU for illuminating an object, a rotary table (photographing table rotating device) RU as a relative position moving means, and A sensor RU for recognizing that the rotary table RU has rotated once, a blue back BB as a background plate, a target object BE, and a photographing time recording device TM.
And a relative position calculating device PC as relative position specifying means
And a shape estimation device SR that performs shape estimation.

FIG. 2 is a flow chart showing the flow of rough processing in the present embodiment.

The general processing flow in this embodiment is as follows: step S0 for performing calibration, step S1 for photographing, and step S2 for creating a boundary image.
And the step S3 of performing shape estimation.

However, the step S0 for performing the calibration may be performed once unless the focus position and zoom position of the camera CU and the positional relationship between the camera CU and the turntable RU are changed.

(Step S0) First, this step S0
At, calibration is performed.

The calibration according to the present embodiment refers to the internal parameters of the camera CU, the camera CU, and the rotation table RU in order to know to which point in the captured image the point in the three-dimensional space is projected. Is a process for obtaining the positional relationship of.

First, the internal parameters of the camera CU will be described.

Here, the internal parameters of the camera CU are the vertical enlargement ratio α _u of the captured image, the horizontal enlargement ratio α _v, and the optical center (from the principal point position as shown in FIG. 3). Feet of a perpendicular line lowered to the image plane) u ₀ , v ₀ .

Here, the enlargement ratio is the ratio of the width of the pixel on the vertical and horizontal images to the distance between the optical principal point of the camera CU and the image plane.

FIG. 3 is a diagram showing the relationship between the camera coordinate system and the image coordinate system.

For example, the origin Oi at the upper left on the image plane IP,
Image coordinate system O with u-axis set horizontally and v-axis set vertically
i, origin Oc at the principal point position of the camera CU, X parallel to the u axis
A camera coordinate system Oc is set in which the Y axis is set parallel to the axes and the v axis and the Z axis is set in the image plane direction.

When a point can be expressed as W = (x, y, z) ^ T in the camera coordinate system Oc and I = (u, v) ^ T in the image coordinate system Oi, these two coordinate systems Oc and Oi The relationship is I ′ = (u ′, v ′, w ′) ^ T = U × W ′ I = (u ′ / w ′, v ′ / w ′) ^ T W ′ = (x, y, z, 1) ^ T.

Here, ^ T represents the transpose of the vector, and U
Represents a transformation matrix from the camera coordinate system Oc to the image coordinate system Oi.

U is the perspective ratio α _{u in the} u and v directions.
And α _v and the optical centers u ₀ and v ₀ ,

[Equation 1] Can be expressed as

As a method for obtaining these parameters, there is a document "3D CG made from photographs, Takeshi Xu, Modern Science Company, 2001".

Next, the positional relationship between the camera CU and the turntable RU will be described.

This is the above-mentioned camera coordinate system Oc and the rotary table coordinate system Or set on the rotary table RU.
Is equivalent to the relationship.

FIG. 4 is a diagram showing the relationship between the camera coordinate system Oc and each coordinate system of the rotary table coordinate system Or set on the rotary table RU.

Here, in the rotary table coordinate system Or, the rotary axis is set in the Z direction and the rotary plate is set in the XY plane.

When W = (x, y, z) ^ T in the camera coordinate system Oc and Z = (p, q, r) ^ T in the rotary table coordinate system Or, the relationship between these two coordinate systems Oc, Or W = M × Z 'W' = (x, y, z, 1) ^ T Z '= (p, q, r, 1) ^ T.

Here, M is a rotation matrix Rcr representing the positional relationship of the rotation table coordinate system Or viewed from the camera coordinate system Oc.
And the translation vector Tcr

[Equation 2] Can be expressed as

As a method for obtaining the rotation matrix Rcr and the translation vector Tcr, for example, a pattern flat plate PB in which patterns PC as shown in FIG. 5 are arranged at intervals d is placed upright on the rotary plate on the rotary table RU as shown in FIG. There is a method of photographing with the camera CU from a plurality of different angles that are rotated by 10 degrees.

For example, if images are taken from five different angles, the taken images IPB1, IPB2, ...
All patterns PC11, PC1 in the image of IPB5
2 ... PC15, PC21, PC22, ..., PC25, P
C31, PC32, ..., PC35, PC41, PC4
2, ... P, C45, PC51, PC52, ..., PC55
Of the center of gravity of each of the coordinates PC111, P in each image coordinate system Oi
C112 ... PC115, PC211, PC212, ...
PC215, PC311, PC312, ..., PC31
5, PC411, PC412, ..., PC415, PC5
11, PC512, ..., PC515 are calculated.

Considering a pattern flat plate coordinate system Op in which the vertical direction of the pattern flat plate is the Z direction, the horizontal direction is the Y direction, the normal direction of the flat plate is the X direction, and the lower left origin of the flat plate is set, the pattern flat plate of each pattern is considered. Coordinates PC11 and PC1 in the coordinate system Op
2, ..., PC15, PC21, PC22, ..., PC2
5, PC31, PC32, ..., PC35, PC41, P
C42, ..., PC45, PC51, PC52, ..., PC
55 is PCnm = (0, mxd, nxd) ^ T.

Here, when the conversion matrix from the pattern flat plate coordinate system Op to the rotary table coordinate system Or is G and the vector after conversion is PCRpnm, PCRpnm = (PCRpnmx, PCRpnmy, P
CRpnmz) ^ T PCRpnm '= (PCRpnmx, PCRpnmy,
PCRpnmz, 1) ^ T = Rp × G × PCnm ′ PCnm ′ = (0, m × d, n × d, 1) ^ T

[Equation 3] Becomes

Here, the subscript p represents one of a plurality of images.

Rp is a known matrix representing the rotation of the rotation table RU.

That is, by solving the simultaneous equation Ppnm ′ = U × M × Rp × G × PCnm ′, the relationship between the camera coordinate system Oc and the rotary table coordinate system Or, that is, the matrix M can be obtained.

Using this M, a point F on a certain three-dimensional space
Let Fr be the coordinates seen from the rotary table coordinate system Or and Fi be the coordinates seen from the image coordinate system Oi. Fi ′ = (u ′, v ′, w ′) ^ T = U × M × Fr ′ Fi = (U '/ w', v '/ w') ^ T Fr '= (Fr ^ T, 1) ^ T, and it becomes clear which point of the image an arbitrary point in the three-dimensional space is projected on. .

Here, instead of the pattern flat plate, a characteristic pattern is arranged on a solid surface of a cylinder, a square prism or a special shape having a three-dimensional shape, and the three-dimensional coordinates of the center of gravity of the pattern can be accurately known. Anything will do.

(Step S1) Next, in step S1, the target object BE is placed on the turntable RU and the target object B is rotated by the camera CU while rotating the target object BE.
E is photographed together with the background BB, and the object images A1, A2,
…, Get An.

For example, the target object B is photographed 36 times at substantially equal intervals (that is, photographed at 60/36 = 1.67 second intervals) while making one revolution while rotating the target object B as shown in FIG. , A plurality of object images A01, A02, ..., A
Get 36.

At this time, if necessary, the flash FU for illuminating the subject may be synchronized and turned on.

In this case, the shutter speed is 1/500
, The target object B is 360
÷ 60 ÷ 500 = Because it rotates only 0.012 degrees, the problem of image blurring is unlikely to occur.

Here, the rotation speed may be any rotation speed at which the object B is stably rotated and the image is not blurred.

Here, the object images A01, A02, ..., A
Times T1, T2, ..., T36 (s) when 36 is photographed are recorded by the photographing time recording device TM.

As a method of saving the times T1, T2, ..., T36, for example, the images may be added or they may be paired with the shooting order and saved as a table in a separate file.

Here, a sensor RS is provided on the table RU so that the rotation of the rotary table RU can be recognized.

As a result, it is possible to measure the time required for the rotary table RU to make one rotation, and by calculating the average angular velocity of one rotation, the three-dimensional reconstruction of the angular velocity fluctuation of the rotary table RU can be performed. The impact can be reduced.

As the background, for example, a blue background BB as shown in FIG. 1 is used.

However, any background can be used as long as it can be recognized as the background.

For example, red, yellow, green or the like may be used instead of blue, or a pattern such as a lattice pattern may be provided.

(Step S2) In step S2, as shown in FIG.
Of the object images A01, A02,
A boundary image is created from A36.

That is, each object image A01, A02, ...
The background BB is recognized from A36, and a plurality of boundary images B01, B02, ..., B36 are created as shown in FIG.

These are binary images in which the area where the object exists is "1" and the place where the object does not exist is "0" in principle.

However, an area in which it is not possible to distinguish between an object and a background may be saved as a number other than “0” and “1” in an analog manner.

Object images A01, A02, ..., A36 are 3
Since there are 6 sheets, the number of boundary images also becomes 36.

(Step S3) Next, in step S3, shape estimation is performed in the shape estimation apparatus SR shown in FIG.

FIG. 9 is a flow chart showing the flow of processing in step S3.

First, the voxel setting processing step S301
In, the voxel BOX is set in the rotary table coordinate system Or as shown in FIG.

In this case, the setting range of the voxel BOX is
Set it to the area where the target object can be completely covered.

The size of one voxel BOX and the number of voxels BOX are set according to desired accuracy.

For example, the setting range is (-1, -1, 0),
Assuming a cube with (1,1,2) as a diagonal, the size of one voxel BOX is 0.001, and the size of voxel B is
The number of OXs is 2000 × 2000 × 2000 = 8 × 10 ^
It is 9.

Here, ^ 9 represents the 9th power.

In this setting range, a sphere having a diameter of 2 placed at the center of the turntable RU can be measured.

Although a cube is used as the voxel or the setting range of the voxel in the present embodiment, for example, a rectangular parallelepiped, a triangular prism, a hexagonal prism or the like may be used, or a fan-shaped voxel BOX in a cylindrical coordinate system. May be used.

In particular, when the object is flat, the amount of calculation and the amount of memory can be greatly reduced by setting the voxel setting range to a rectangular parallelepiped.

Next, the processing image selection processing step S303.
One image is selected from the images that have not been processed in step S304 and the shooting angle A from the reference angle is calculated in the shooting angle calculation processing step S304.
A1, Aa2, ..., Aa36 are obtained.

This is the relative position calculating device P shown in FIG.
Performed in C.

The reference angle may be based on, for example, the angle at which the image is first captured, or may be based on other image capturing angles.

In this embodiment, the rotary table RU is set to 1 rp.
Since it is rotated at m, it rotates 6 degrees per second.

Therefore, Aan = (Tn−T1) × 6 Becomes

Further, in the judgment target voxel selection processing step S306, an unjudged voxel is selected from the vertices of each voxel which is not judged to be outside.

At the voxel vertex coordinate projection processing step S308, the selected voxel is subjected to boundary image B01, B02, ... , B36
Voxel external determination processing step S30
If all the vertices are not included in the object area at 9, it is determined to be outside.

Here, in the external determination, the centroid of the voxel may be projected onto the boundary images B01, B02, ..., B36, and the case where the projected point is not included in the object area may be determined to be external. .

The determination result is stored in the determination result storage unit (not shown).

FIG. 11 is a diagram showing an example of this determination.

That is, FIG. 11A shows the case where it is judged to be inside, and FIG. 11B shows the case where it is judged to be outside.

FIG. 12 is a diagram showing how voxels are judged to be outside by the boundary image B01.

Here, the shaded areas are the voxels judged to be outside.

FIGS. 13 (a) to 13 (d) are diagrams showing how a target object is cut out by this method in a two-dimensional example with respect to a simple shape.

Here, the voxels in which the shaded portions are cut out are the voxels which are not determined because the mesh pattern portions have previously been determined to be outside.

Further, a voxel whose number of times of being judged to be outside by the voxel outside judging section is a threshold value or more is registered as an external voxel.

This determination may be performed using, for example, the value of external probability = the number of determinations / the number of images used in the determination.

Finally, when the processing is completed for all the images, the voxels that are not the outside are set as the three-dimensional shape of the target object.

Further, the photographing by the photographing section may be carried out through a plurality of rotations.

In this case, the following effects can be obtained by photographing from different viewpoints at each rotation.

For example, in the first rotation, the image is taken at around 0, 180 degrees (this angle may not be exactly if the shooting angle itself is known), and in the second rotation, it is around 90,270 degrees, If the resolution in the rotation direction is gradually made finer, such as around 45, 135, 225, and 315 degrees at the third rotation, a gradually finer shape can be seen for each rotation.

As a result, the area to be shape-estimated by the shape estimator can be known at an early stage of the processing, and the processing can be speeded up.

Further, depending on the photographing unit, 0, 10, ... 34
Even if the images were taken in sequence around 0 and 350 degrees, 0 and 1
80, 90, 270, 40, 130, 220, 310 ...
If shape estimation is performed in the order of degrees, the same effect as described above can be obtained.

Further, as shown in FIG. 14, a photographing table lifting device UDU for moving the target object BE up and down in the rotation axis direction at a constant speed is further provided, and the rotation table RU is moved at a constant speed in accordance with the rotation of the rotation table RU. You may comprise so that it may shoot while lifting.

In this case, since the relative position in the vertical direction also changes, it is possible to estimate the part of the target object that could not be estimated by simple rotational movement.

The photographing position changes not only in angle but also in movement in height, but since it is moving at a constant speed, the height can be obtained by multiplying the photographing time at a constant speed as in the calculation of the angle described above. it can.

(Second Embodiment) The device configuration of the stereoscopic information acquisition apparatus according to the second embodiment is the same as that of the first embodiment.

FIG. 15 is a flow chart showing the flow of processing of the stereoscopic information acquisition apparatus according to the second embodiment.

The processing flow of the three-dimensional information acquisition apparatus according to this embodiment is the same as that of the first embodiment in steps S0 to S3.

In the texture mapping processing step S4, the color information of voxels on each surface is determined using the object images A01, A02, ..., A36.

For example, first, the coordinates of the center of a voxel on the surface are set to object images A01 and A0 in which the voxel is visible.
2, ..., A36, and the color information of the projected image is averaged to obtain the color information of the voxel.

As a result, not only the three-dimensional shape of the object but also the color information of each part can be acquired.

(Third Embodiment) FIG. 16 shows the third embodiment.
It is a block diagram showing a configuration of a stereoscopic information acquisition device according to an embodiment of.

In this embodiment, the photographing step S2 and the boundary image creating step S3 of the first embodiment are different.

First, a light source that makes the entire back surface bright is provided to the camera CU as a target object BE as an object.
Be installed behind.

For example, as shown in FIG. 16, a diffuser plate FB is installed, and a flash BF is provided from behind to synchronize with shooting.
Make U glow.

Further, the target object BE is set on the transparent base CB.

Then, the rotary table RU is rotated in the same manner as in the first embodiment, so that, for example, 36 silhouette images S01, S02, ..., S36 are photographed (see FIG. 17).

By photographing in this way, it is possible to obtain the same image as that in the case of so-called backlighting in which the background area has high brightness and the area where the object exists is very dark.

Next, in the boundary image creating processing step S3, a dark area is extracted from the silhouette image.

For example, a pixel having a brightness value equal to or higher than a certain threshold is used as a background, and the remaining area is used as an object.

The details of the photographing method and the clipping method are based on Japanese Patent Application No. 2001-245593.

(Fourth Embodiment) FIG. 18 shows the fourth embodiment.
It is a block diagram showing a configuration of a stereoscopic information acquisition device according to an embodiment of.

The present embodiment is different from the photographing step S1 of the second embodiment shown in FIG. 15, the boundary image creation processing step S2, and the texture mapping processing step S4.

First, a light source that makes the entire back surface bright is installed behind the object with respect to the camera CU.

For example, as shown in FIG. 18, the diffusion plate FB
Install a flash BF from behind to match the shooting
Make U glow.

By photographing in this way, it is possible to obtain the same image as that in the case of so-called backlighting in which the background area has high brightness and the area where the object exists is very dark.

In addition, a flash F for illuminating the front surface
Equipped with U and alternate with the flash BFU behind, or
Flash switching device FC for flashing every few times
Equipped with U.

The rotary table RU is provided with a transparent base CB, on which the target object BE as an object is mounted.
, The texture images T01 and T0 captured by illuminating a plurality of FUs while rotating the turntable RU.
, ..., T36 and silhouette images S01, S02, ..., S36 photographed by shining a plurality of BFUs are photographed alternately, for example (see FIG. 19).

Next, in the boundary image creation processing step S3, a dark area is extracted from the silhouette image.

For example, a pixel having a brightness value equal to or higher than a certain threshold is used as a background, and the remaining area is used as an object.

Details of the photographing method and the clipping method are in accordance with Japanese Patent Application No. 2001-245593.

In the texture mapping processing step, the texture image is used to give color information to the voxels as in the second embodiment, so that not only the three-dimensional shape of the object but also the color information of each part is acquired. can do.

(Fifth Embodiment) In the fifth embodiment, the shape estimation processing step S3 of the first to fourth embodiments is different.

20 (a) to 20 (d) are two-dimensional views showing how the voxels of this embodiment are changed.

In this embodiment, first, one voxel is set to a size larger than the desired accuracy.

Further, as in the first to fourth embodiments, the voxel external determination is performed, but voxel division is performed on a voxel in which the determination of eight vertices is mixed externally and internally.

In FIG. 20, one square is divided into four because it is a two-dimensional description, but in three dimensions, it is divided into eight cubes.

Then, the external judgment is similarly performed on the divided cubes and the division is repeated recursively.

The same processing is performed using the boundary image that has ended when the voxel size has reached the desired accuracy and has not yet been processed.

As described above, by performing the three-dimensional reconstruction while recursively dividing, the number of voxels to be processed is drastically reduced because the voxels are divided by the voxels larger than the desired accuracy except the boundary between the inside and the outside of the object. To do.

If it is determined that all voxels inside the voxel that are once divided are outside, then those voxels are integrated and regarded as one voxel. By doing so, it is possible to further reduce the number of voxels.

Here, in addition to the eight vertices, the voxel external determination may be performed using, for example, the corner centroids of the six surfaces or the midpoints of the sides.

This makes it possible to perform three-dimensional reconstruction of an object having a more complicated shape.

In addition to the claims 1 to 19 shown in the scope of the claims, the present invention shown in the above-described embodiments includes the inventions shown as the following supplementary notes 1 to 3. include.

(Supplementary Note 1) A photographing means for photographing an image of the target object, and the target object and the photographing means at a predetermined relative speed so that the photographing means can photograph the image of the target object from a plurality of viewpoints. Relative moving means for moving, reference position detecting means for detecting a reference relative position of the target object and the photographing means, time when the reference position detecting means detects the reference relative position, and the photographing means for measuring the target object. A relative position specifying unit that specifies a relative position between the target object and the photographing unit based on a difference between the time when the image is photographed and the predetermined relative speed, and from a plurality of viewpoints photographed by the photographing unit. Using the image of the target object and the information of the relative position specified by the relative position specifying means,
And a stereoscopic shape estimating means for estimating a stereoscopic shape of the target object.

(Operation) According to the present invention, since photographing can be performed while the target object is moving, since the acceleration / deceleration during movement is small, the photographing time is reduced and the target object is less likely to fall or deform due to inertia. In addition, the energy consumption can be reduced.

(Supplementary Note 2) A range including the entire boundary between the photographing means for photographing the image of the target object and at least the background portion of the target object within the photographing range of the photographing means is directly or directly behind the target object. Illumination means for indirectly illuminating, and relative movement means for moving the target object and the photographing means at a predetermined relative speed so that images of the target object can be photographed from a plurality of viewpoints by the photographing means. A reference position detecting unit that detects a reference relative position between the target object and the image capturing unit, a time when the reference position detecting unit detects the reference relative position, and a time when the image capturing unit captures an image of the target object. From a plurality of viewpoints photographed by the photographing means, the relative position specifying means for specifying the relative position of the target object and the photographing means from the difference of The image of the target object, and the three-dimensional shape estimation means for estimating the three-dimensional shape of the target object using the information of the relative position specified by the relative position specifying means, the illumination means, A stereoscopic information acquisition device, which is turned on when performing imaging for estimating the stereoscopic shape of a target object.

(Operation) According to the present invention, since the object can be photographed while moving by the light cutting and the cutting, the photographing time is reduced due to less acceleration / deceleration during movement.
The target object may be less likely to fall and deform due to inertia, and the energy consumption may be less.

(Supplementary Note 3) The relative position moving means rotates the target object at a constant angular velocity, and the relative position detecting means detects the reference angular position and the reference angular position detecting means. Using the time difference calculation means for calculating the time difference from the time when the position detection means detects the reference angular position to the time of photographing by the photographing means, and the constant angular velocity and the result obtained by the time difference calculation means, the reference From an angle difference specifying means for specifying an angle difference rotated by a time taken by the imaging means from an angle, and a time from the reference angular position detecting means once detecting the reference angular position to detecting it again, The three-dimensional shape information acquisition apparatus according to claim 4, further comprising: a constant angular velocity calculation unit that calculates an average angular velocity and sets the constant angular velocity.

(Operation) In (4), the present invention can reduce the influence of fluctuation of the angular velocity due to, for example, placing the target object on the turntable as the constant angular velocity measurement.

[0185]

As described above, according to the present invention, it is possible to obtain a boundary with high accuracy and acquire stereoscopic information with higher accuracy, and to increase the memory capacity for acquiring stereoscopic information of a target object. It is possible to provide a three-dimensional information acquisition device, a three-dimensional information acquisition method, and a three-dimensional information acquisition program that can reduce the image capturing time, reduce the shooting time, and keep the target object in a stable state.

[Brief description of drawings]

FIG. 1 shows a first embodiment of a stereoscopic information acquisition device according to the present invention.
3 is a block diagram showing the configuration of the embodiment of FIG.

FIG. 2 is a first diagram of a stereoscopic information acquiring apparatus according to the present invention.
4 is a flowchart showing a rough processing flow in the embodiment.

FIG. 3 is a diagram showing a relationship between a camera coordinate system and an image coordinate system.

FIG. 4 is a camera coordinate system Oc and a rotary table R.
It is a figure which shows the relationship of each coordinate system of the rotary table coordinate system Or set on U.

FIG. 5 shows a rotation matrix Rcr and a translation vector Tc.
It is a figure which shows the pattern flat plate PB used as an example which calculates | requires r.

FIG. 6 shows a rotation matrix Rcr and a translation vector Tc.
FIG. 6 is a diagram shown for explaining a method of taking an image with a camera CU from a plurality of different angles in which the pattern flat plate PB of FIG.

FIG. 7 shows object images A01, A02, ..., A36.
FIG.

FIG. 8 shows boundary images B01, B02, ..., B36.
FIG.

FIG. 9 is a flowchart showing a flow of processing of step S3 of FIG.

FIG. 10 is a diagram showing a set voxel BOX.

FIG. 11 is a diagram illustrating an example of external determination.

FIG. 12 is a diagram showing how a voxel is determined to be outside by the boundary image B01.

FIG. 13 is a diagram showing how a target object is cut out in a two-dimensional example with respect to a simple shape.

FIG. 14 is a block diagram showing a configuration of a modified example of the first exemplary embodiment.

FIG. 15 is a flowchart showing a processing flow of the stereoscopic information acquisition device according to the second embodiment of the present invention.

FIG. 16 is a block diagram showing a configuration of a stereoscopic information acquisition device according to a third embodiment of the present invention.

FIG. 16 shows silhouette images S01, S02,
... It is a figure which shows S36.

FIG. 18 is a block diagram showing a configuration of a stereoscopic information acquisition device according to a fourth embodiment of the present invention.

FIG. 19 is a diagram showing a captured image according to a fourth embodiment.

FIG. 20 is a two-dimensional view showing how voxels are changed in the fifth embodiment of the present invention.

[Explanation of symbols]

CU ... Camera as photographing means, FU ... Flash for illuminating subject, RU ... Rotating table (imaging table rotating device) as relative position moving means, RU ... Sensor for recognizing that the rotating table RU has made one rotation, BB ... Blue background as background plate, BE ... Target object, TM ... Imaging time recording device, PC ... Relative position calculation device as relative position specifying means, SR ... Shape estimation device for shape estimation, PB ... Pattern flat plate, FB ... Diffusion Board, BFU ... Flash, CB ... Transparent base.

Front page continuation (51) Int.Cl. ⁷ Identification code FI theme code (reference) H04N 5/222 H04N 13/02 5L096 13/02 G01B 11/24 KA F term (reference) 2F065 AA04 AA20 AA53 DD06 DD07 FF02 FF05 FF65 GG08 JJ03 JJ19 JJ26 LL49 MM04 PP13 QQ31 5B050 BA01 DA02 DA04 EA06 EA09 EA28 EA30 5B057 AA20 BA02 BA19 CA01 CA08 CA13 CB20 CC01 DA08 DA11 DA20 DB03 FA02 AB01 FA02 AC02 AB02 FA02 AC02 AB05 AB02 AC02 AB15 AB02 AC15 AB02 AB02 AB02 AB15 AB02 AC02 AB15 AB02 AC15 AB02 AB02 AB15 AB02 AB15 AB06 JA11

Claims (19)

[Claims] 1. A photographing means for photographing an image of a target object, and the target object and the photographing means are relative and continuous so that the photographing means can photograph the image of the target object from a plurality of viewpoints. A relative movement means for moving the target object from a plurality of viewpoints by the photographing means, and a relative position detection means for detecting a relative position between the target object and the photographing means at each viewpoint, Using the image of the target object photographed from a plurality of viewpoints by the photographing means, and the information of the relative position detected by the relative position detection means, a three-dimensional shape estimation means for estimating the three-dimensional shape of the target object, A stereoscopic information acquisition device comprising: 2. A photographing means for photographing an image of a target object, a background plate having predetermined optical characteristics, which is arranged behind the target object and serves as a background of the target object at the time of photographing, and the photographing means. A relative movement unit that relatively and continuously moves the target object and the photographing unit so that images of the target object can be photographed from a plurality of viewpoints by means of the photographing unit; Relative position detecting means for detecting the relative position of the target object and the photographing means at each viewpoint where the image of the object is photographed, images of the target object photographed from a plurality of viewpoints by the photographing means, and the relative Using the relative position information detected by the position detecting means, the target object occupies an image from a plurality of viewpoints photographed by the photographing means. Recognizes the band, three-dimensional information acquisition apparatus characterized by comprising: a three-dimensional shape estimation means for estimating a three-dimensional shape of a target object using an area where the target object occupies. 3. A photographing means for photographing an image of the target object, and a range including at least the entire contour portion of the target object or a part thereof within the photographing range of the photographing means, directly or indirectly from behind the target object. Illumination means for illuminating the target object, and relative movement means for relatively and continuously moving the target object and the image capturing means so that the image of the target object can be captured from a plurality of viewpoints by the image capturing means. A relative position detecting unit that detects a relative position between the target object and the image capturing unit at each viewpoint where the image of the target object is captured by the image capturing unit from a plurality of viewpoints; An image of the target object captured and the relative position information detected by the relative position detection means are used to capture an image. A three-dimensional shape estimating unit that recognizes an area occupied by the target object in each of images from several viewpoints and estimates the three-dimensional shape of the target object using the area occupied by the target object; A stereoscopic information acquisition device, characterized in that the illumination means is turned on when an image used for recognizing an area is photographed. 4. The three-dimensional information acquisition apparatus according to claim 2, wherein the relative movement unit rotates the target object. 5. The stereoscopic information acquiring apparatus according to claim 4, wherein the relative moving unit moves the target object in a direction parallel to a rotation axis. 6. The relative moving means rotates the target object at a constant angular velocity, and the relative position detecting means includes a reference angular position detecting means for detecting a reference angular position, and the reference angular position detecting means. Using the time difference calculation means for calculating the time difference from the time when the reference angle position is detected to the time of photographing by the photographing means, and the constant angular velocity and the result obtained by the time difference calculation means, The three-dimensional information acquisition device according to claim 4, further comprising: an angle difference specifying unit that specifies an angle difference rotated by the time when the image is taken by the image taking unit. 7. An image obtained at the time of photographing for estimating the three-dimensional shape of the target object by the three-dimensional shape estimation means is such that an area occupied by the target object is relatively larger than a background portion near the contour of the target object. The stereoscopic information acquisition device according to claim 3, wherein the stereoscopic image is a very dark silhouette image. 8. The three-dimensional shape estimating means extracts the contour of the target object by utilizing a brightness difference between a contour area of the target object and a background area around the target object in the silhouette image, and the extracted contour is extracted. The stereoscopic information acquisition apparatus according to claim 7, wherein the area occupied by the target object is estimated by using the contour. 9. The illuminating means includes a light source that emits light in at least a visible light region, and the illuminating means is disposed behind the target object with respect to the photographing means,
The light-scattering means for scattering light from the light source, and the stereoscopic information acquisition device according to claim 3. 10. The three-dimensional shape estimation means, in a situation where the illumination means does not illuminate the target object,
The stereoscopic information acquisition apparatus according to claim 3, wherein the texture of the surface of the target object is estimated using a texture image of the target object captured at least once by the image capturing unit. 11. The three-dimensional shape estimating unit sets a closed region setting unit that sets a set of closed regions arranged in close contact with each other in a three-dimensional space with respect to the target object, and the photographing unit from a plurality of viewpoints. A closed region outside for determining whether or not the closed region exists outside the target object by obtaining a probability that each closed region exists outside the target object based on an image obtained by photographing the target object. Determining means, and, if the closed area outside determination means determines that the probability that the closed area exists outside the target object exceeds a predetermined threshold, the closed area is the closed area after that. The stereoscopic information acquisition device according to claim 2 or 3, wherein the stereoscopic information acquisition device is excluded from the determination processing target of the external determination means. 12. The three-dimensional shape estimating means recognizes an area occupied by the target object based on an image taken by a line of sight from a first viewpoint among images taken from a plurality of viewpoints by the photographing means. At the same time, a recognition estimation process for estimating the three-dimensional shape of the target object is performed, and subsequently, a region occupied by the target object based on an image captured with a line of sight from a second viewpoint farthest from the first viewpoint. And a recognition estimation process for estimating the three-dimensional shape of the target object, and thereafter, the remaining viewpoints corresponding to images not used in the recognition estimation process, which are used in the recognition estimation process. Among the viewpoints in which the angle difference made by the line of sight corresponding to each of the two captured images is the largest, the recognition estimation is performed using the image from the viewpoint closest to the angle at which the angle difference is interpolated. The three-dimensional information acquisition apparatus according to claim 4, wherein processing is performed and the processing is sequentially repeated. 13. The three-dimensional shape estimating means has a probability that a closed area belonging to the set of closed areas exists outside the target object within a predetermined probability range, and exists near the boundary between the inside and the outside of the target object. Boundary closed region determination means for determining whether or not the closed region, and the closed region determined to be a closed region existing near the boundary by the boundary closed region determination means is divided into smaller divided closed regions. Boundary closed region dividing means, and for the divided closed region divided by the boundary closed region dividing means, the boundary closed region determining means determines again whether or not the closed region exists near the boundary. On the basis of this determination result, the divided closed region is re-divided by the boundary closed region dividing unit, and the process of the determination by the boundary closed region determining unit and the division by the boundary closed region dividing unit is performed. The stereoscopic information acquisition apparatus according to claim 2, wherein the closed area is repeated until the closed area has a predetermined size. 14. A region occupied by the target object in the image is recognized by using images of the target object photographed from a plurality of viewpoints and information on the viewpoint positions, and the target is determined based on the region occupied by the target object. A three-dimensional information acquisition method for estimating the three-dimensional shape of an object and acquiring the three-dimensional information of the target object, wherein a set of closed regions closely arranged in a three-dimensional space is set for the target object. Whether the closed region exists outside the target object by determining the probability that each closed region exists outside the target object based on the closed region setting step and images obtained by photographing the target object from multiple viewpoints. A closed region external determination step of determining whether or not, the probability that the closed region exists outside the target object exceeds a predetermined threshold value, if determined in the closed region external determination step, the closed region, Subsequent said closure A stereoscopic information acquisition method, comprising: a step that is excluded from the determination processing target of the area outside determination step. 15. An image obtained by capturing images of a target object from a plurality of viewpoints and information on the viewpoint positions are used to recognize a region occupied by the target object in the image, and the target is determined based on the region occupied by the target object. A stereoscopic information acquisition method for acquiring stereoscopic information of the target object by estimating a stereoscopic shape of an object, comprising: an image captured from a first viewpoint among images captured from the plurality of viewpoints. A first recognition estimation processing step of recognizing an area occupied by the imaging target based on the first recognition estimation processing for estimating a three-dimensional shape; and photographing with a line of sight from a second viewpoint farthest from the first viewpoint. A second recognition estimation processing step of recognizing an area occupied by the target object based on the obtained image and performing a recognition estimation processing of estimating a three-dimensional shape; and an image not used in the recognition estimation processing. Of the remaining viewpoints, the angle between which the angle difference is interpolated among the viewpoints sandwiched by the ones having the largest angle difference formed by the line of sight corresponding to each of the two images used in the recognition estimation processing. And a third recognition estimation processing step for performing the recognition estimation processing using an image from the viewpoint closest to the stereoscopic information acquisition, characterized in that the third recognition estimation processing step is sequentially repeated. Method. 16. An area occupied by the target object in the image is recognized by using images of the target object captured from a plurality of viewpoints and information on the viewpoint positions, and the target is determined based on the area occupied by the target object. A three-dimensional information acquisition method for estimating the three-dimensional shape of an object and acquiring the three-dimensional information of the target object, wherein a set of closed regions closely arranged in a three-dimensional space is set for the target object. Closed region setting step, whether the closed region belonging to the set of closed regions is outside the target object within a predetermined probability range, and whether the closed region exists near the boundary between the inside and outside of the target object. Boundary closed region determination step of determining whether or not, a boundary closed region dividing step of dividing the closed region determined to be a closed region existing near the boundary by the boundary closed region determination step into smaller divided closed regions When, Stereoscopic information characterized by repeating the process of executing the boundary closed region determination step and the boundary closed region dividing step for the divided closed region until the closed region has a predetermined size. Acquisition method. 17. A region occupied by the target object in the image is recognized by using images of the target object photographed by a computer from a plurality of viewpoints and information on the viewpoint positions,
A three-dimensional information acquisition program for estimating the three-dimensional shape of the target object based on the area occupied by the target object and acquiring the three-dimensional information of the target object. Based on the image for capturing the target object from a plurality of viewpoints, the step for setting a closed region that sets a set of closed regions closely arranged in the dimensional space, and each closed region is outside the target object. By determining the probability of existing in the closed area outside the target object to determine whether the closed area exists outside the target object, and the probability that the closed area exists outside the target object is When it is determined in the step for determining the outside of the closed region that the predetermined threshold is exceeded, the closed region is excluded from the determination processing targets in the steps for determining the outside of the closed region. And a program for executing the following, a stereoscopic information acquisition program. 18. An image of a target object captured from a plurality of viewpoints by a computer and information of the viewpoint position are used to recognize an area occupied by the target object in the image,
A three-dimensional information acquisition program for estimating the three-dimensional shape of the target object based on the area occupied by the target object and acquiring the three-dimensional information of the target object, wherein the computer captures images from the plurality of viewpoints. A first recognition estimation process for recognizing a region occupied by the imaging target based on an image captured with a line of sight from a first viewpoint in the image and performing a recognition estimation process for estimating a three-dimensional shape of the target object. For recognizing the area occupied by the target object based on the image captured with the line of sight from the second viewpoint farthest from the first viewpoint, and estimating the three-dimensional shape of the target object. A step for a second recognition estimation process for carrying out a process, and the remaining viewpoints corresponding to images not used in the recognition estimation process, which are used in the recognition estimation process. Among the viewpoints sandwiched by the ones having the largest angle difference formed by the line of sight corresponding to each of the two images, the recognition estimation is performed using the image from the viewpoint closest to the angle at which the angle difference is interpolated. A stereoscopic information acquisition program, comprising a step for performing a third recognition estimation process for performing processing, and a program for executing the step, wherein the steps for the third recognition estimation process are sequentially repeated. 19. An area in which the target object occupies in the image is recognized by using an image obtained by capturing an image of the target object from a plurality of viewpoints by a computer and information of the viewpoint position,
A three-dimensional information acquisition program for estimating the three-dimensional shape of the target object based on the area occupied by the target object and acquiring the three-dimensional information of the target object. A step for setting a closed region that sets a set of closed regions closely arranged in a dimensional space, and a probability that a closed region belonging to the set of closed regions exists outside the target object has a predetermined probability range. Therefore, a step for determining a boundary closed area that determines whether or not the area is a closed area existing near the boundary between the inside and the outside of the target object, and the step for determining the closed area exists near the boundary. A step of dividing the closed region determined to be the closed region into smaller divided closed regions, and a program for executing: On the other hand, a stereoscopic information acquisition program, characterized in that the processing for executing the step for determining the boundary closed area and the step for dividing the boundary closed area is repeated until the closed area reaches a predetermined size. .