JPH1096607A - Object detector and plane estimation method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve accuracy of detecting an object by exactly estimating a three dimensional position of a plane object without being affected by an object other than the plane object. SOLUTION: By using a stereo image photographed with a plurality of photographing devices 1 and 2, parallax data as a kind of three dimensional data of the object photographed with a corresponding processor 5 is gained. At the initiation of the processing, the parallax data is input in a plane estimating part 50 and from the parallax data partially measured, the position of the plane object such as road and floor is estimated by means of Hough transform. After the initiation of processing, a new object put on the plane object is detected with a parallax change detector 8 using the reference of the position of the estimated plane object. At the time after the initiation of processing, the position information of the plane object already estimated by the corresponding processor 5 is input, and by means of the position information, the measuring range of three dimensional data such as parallax is limited.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to an object detection apparatus using stereo image processing applied to monitoring, measurement, and the like, and a plane estimation method.

[0002]

2. Description of the Related Art A conventional object detection device has an operation function as shown in a flowchart of FIG. The flow of the operation will be described below with reference to FIG.

In FIG. 11, reference numerals 1 and 2 denote two right and left imaging devices arranged so that their optical axes are parallel to each other, and reference numerals 3 and 4 denote image memories for storing image data from the respective imaging devices 1 and 5;
Is an associating processing unit that measures parallax as described later, 6
Is a parallax image memory that stores parallax image data from the association processing unit 5, 7 is an initial parallax image memory that stores parallax image data at the start of processing, and 8 is a parallax image memory 6 and an initial parallax image memory 7. Based on the stored parallax image data, a parallax change detection unit 9 for determining the presence or absence of an object in the image as described later, and an object detection image memory 9 for storing image data of the object detected by the parallax change detection unit 8 It is.

In the above-mentioned apparatus, a left image is used as a reference among the images taken by the two left and right imaging devices 1 and 2 arranged so that the optical axes are parallel to each other. Figure 12 on the left image
Then, the image is divided into rectangular areas in the horizontal direction M and the vertical direction N as described above, and the correspondence processing unit 5 measures the shift of the image position of the same object in both images, that is, the parallax. Parallax is inversely proportional to the distance to an object.

[0005] As an example of the stereo image processing method and the associating method, a method described in Keiji Mikiyoshi et al., "Driving support system using three-dimensional image recognition technology" (Preprint 9241992-10, Academic Lecture Meeting of the Society of Automotive Engineers of Japan) This will be described with reference to FIG.

Hereinafter, a rectangular area in the left image in FIG. 12A is referred to as a block 11, and as shown in FIG. 12B, each block is composed of m × n pixels. Let the brightness of the i-th pixel inside the block 11 be Li.

In FIG. 13, a rectangular area 22 of m × n pixels is set in the right image 20 in the same manner as described above, and the brightness of the i-th pixel in the rectangular area 22 is defined as Ri.

The similarity evaluation value C of the rectangular area between these left and right images is given by (Equation 1).

[0009]

(Equation 1)

The right image 20 in which a corresponding area may exist
In the middle parallax search range 21, the similarity evaluation value C is calculated by moving the rectangular area 22 by one pixel in the horizontal direction, and an area where this value is minimum is defined as a corresponding area. In this method, the corresponding area can be determined for each block 11 in the left image, and when the corresponding area is determined, the parallax S can be immediately obtained from (Equation 2) from the coordinate position of the corresponding area. .

[0011]

S = X−X _R where X _R is the X coordinate of the rectangular area in the right image X is the X coordinate of the rectangular area in the left image From the parallax S obtained for each block 11 in the left image as a reference, The distance K to the object can also be obtained using the following formula (Equation 3) which is generally known.

[0012]

(Equation 3)

However, 2a: the distance between the left and right cameras f: the lens focal length In order to distinguish the parallax for each position of each block 11, the parallax to the object measured for each block 11 is expressed by the following equation (4).

[0014]

## EQU4 ## Parallax: S (X, Y) 1 ≦ X ≦ M, 1 ≦ Y ≦ N, and an image having parallax information is referred to as a parallax image. In particular, the parallax image at the processing start time t _start is called an initial parallax image, and is distinguished from the case at time t (t> t _start ). The initial parallax image is stored in the initial parallax image memory 7, and the parallax image at the current time is stored in the parallax image memory 6.
Is stored.

The parallax change detection unit 8 compares parallax between the same block in the parallax image at time t (present) and the initial parallax image at time _tstart (past), and a parallax change equal to or greater than a certain threshold value occurs. It is determined that a new object exists or has disappeared in the rectangular area.

Next, the Hough transform will be described with reference to FIG. In the straight line 30 on the xy plane, if the length of the perpendicular 31 drawn from the coordinate origin is ρ ₀ , and the angle between the perpendicular 31 and the x axis is θ ₀ , the straight line 30 is expressed as (Equation 5).

[0017]

Equation 5] ρ ₀ = xcosθ ₀ + ysinθ ₀ Then, x-y point sequence on the plane _{(x i, y i) {} i = 0,1,
Calculate 2｝ with (Equation 6),

[0018]

Ρ = x _i cos θ ₀ + y _i sin θ ₀ is expressed as a composite triangular function as shown in FIG. 14 when expressed on the ρ-θ plane. This sine curve is called a Hough curve. This Hough curve shows all straight lines passing through the point (x _i , y _i ). x-y sequence of points (x _i, y _i) on the plane Hough curves in the case, the point sequence (x _i, y _i) {i = 0, 1, 2} sequentially [rho-theta plane in a straight line Then, as shown in FIG.
At the intersection. Point 35 where the most curves intersect
A straight line 30 (shown in (Expression ₅ )) on the xy plane is determined by (ρ ₀ , θ ₀ ).

The above operation is called a straight line fitting by Hough transform. Since the straight line fitting by the Hough transform is a straight line detection method for grasping the global situation of a data point sequence, even if points that are not on a straight line are mixed, the straight line can be detected without being affected by this.

[0020]

In the above-mentioned conventional object detection device, it is desired to accurately estimate the position in a three-dimensional space with respect to a relatively large planar object such as a road, a floor, or a wall, which is a reference for detection determination. It is rare.

Therefore, in the present invention, when estimating the position of a plane, the Hough transform, which is a global linear approximation method, is used to accurately determine the three-dimensional position of the plane object without being affected by objects other than the plane object. The first purpose is to estimate.

Further, in the conventional object detection device, when associating a rectangular area between the left and right images with the associating processing unit, the position where the similarity evaluation value C is minimum is used as the corresponding area. Since there are few features in the region, the minimum point of the similarity evaluation value C cannot be clearly obtained, and thus parallax data may not be obtained.

This means that the conventional object detection device, on the premise that almost all of the parallax data in the initial parallax image and the parallax image at the current time have been obtained, cannot perform accurate detection.

Therefore, in the present invention, the position of a planar object is estimated, and undetermined parallax data in the same plane is interpolated by calculation.
A second object is to enable object detection with high accuracy.

Further, in the conventional association processing, FIG.
As shown in FIG. 7, the parallax distance measurement range 40 was set in accordance with the object farthest from the camera 42 as the imaging device and the object closest to the camera 42, and was the same regardless of the position in the image. However, in most cases, the object to be detected is located closer to the camera 42 than the plane object 41 such as a road, a floor, or a wall, and the plane object 41 has an angle other than a right angle with respect to the optical axis L of the camera 42. In most cases, the parallax is searched for a useless range in which there is no object whose distance is to be measured, so that the calculation amount is large.

Therefore, in the present invention, the search range of the parallax is optimized by the block position of the image based on the position information of the estimated plane to reduce the amount of calculation of the correspondence.
The purpose is.

[0027]

In order to achieve the above object, an object detecting apparatus according to the present invention comprises a plurality of image pickup means arranged at a predetermined interval and a triangular image using a plurality of images output from these image pickup means. An association processing unit that measures three-dimensional data of an object based on the principle of surveying, and a plane position estimation that estimates a position of the planar object in a three-dimensional space by performing a Hough transform process on the measured three-dimensional data With this configuration, the positions of planar objects such as roads, floors, and walls that occupy most of the image in the image are globally estimated by Hough transform processing based on the measured three-dimensional data. Thus, the position of the planar object can be accurately estimated without affecting the three-dimensional data of the object other than the planar object.

In the object detection apparatus, the plane position estimating unit estimates the position of a plane object in an image in a three-dimensional space, and determines the position of a rectangular area in which three-dimensional data cannot be obtained in the same plane. The three-dimensional data is calculated and interpolated from the position of the planar object which has already been estimated, and an object located at a position different from the planar object is detected on the basis of the imaging means. By estimating the position of the plane object and interpolating the unmeasured three-dimensional data in the same plane by calculation from the estimated plane position, it is possible to prevent omission of detection due to unmeasured parallax data. As a result, it is possible to improve the detection accuracy when detecting an object at a position distant from the planar object.

In the object detecting apparatus, the plane position estimating unit estimates the position of the plane object in the image in a three-dimensional space, and for an object that is farther from the plane object with respect to the imaging means, 3 by attachment processing part
By configuring so as not to measure the three-dimensional data, the calculation of the associating process is performed so that the three-dimensional data is not measured at a position farther than a plane such as a road, a floor, or a wall estimated from the imaging device. The amount can be reduced.

Further, according to the plane estimation method of the present invention, a rectangular small area obtained by dividing an image into a horizontal direction and a vertical direction using a plurality of images output from a plurality of image pickup means arranged at predetermined intervals. For each large area obtained by grouping the rectangular small areas in the horizontal direction, a plane-passing straight line is applied by a Hough transform from a plurality of three-dimensional data included in the large area, and the arithmetic processing is performed. This is a method of estimating the position of a planar object existing in a three-dimensional space by performing the above processing on all the large regions. By this method, when approximating a straight line passing through a plane, a Hough transform process is performed. Since it is used, the plane position can be globally estimated without being affected by objects other than the plane object.

In the above-mentioned plane estimating method, the slope of the total number of each straight line passing through the planes is sorted by the magnitude using the slope of the straight line which is one of the straight line parameters of the straight line passing through the plane applied to each large area. The position of a plane object existing in a three-dimensional space is obtained by using the corrected plane-passing straight line obtained by replacing the gradients of all the plane-passing straight lines with the above-mentioned median. Is estimated, the slope, which is one of the straight line parameters, should originally be the same for all plane-passing straight lines. Therefore, by replacing the slope of each plane-passing straight line with the most reliable slope, The reliability of the inclination of the straight line passing through the plane can be improved.

In the above-mentioned plane estimating method, a distance correction straight line is fitted by Hough transform using a distance from the origin, which is a straight line parameter of the plane-passing straight line whose inclination has been corrected, and the slope is corrected so as to intersect the distance correction straight line. By estimating the position of a planar object existing in a three-dimensional space using a group of straight lines obtained by translating the corrected straight line, the condition that the straight line passing through the plane should be in the same plane is obtained. The distance correction straight line is obtained by performing Hough transform on the distance from the origin, which is another straight line parameter of each plane passing straight line, so that each plane passing straight line intersects the distance correction straight line. Since the distance from the origin of each straight line passing through the plane is corrected, the reliability of the distance from the origin of each straight line passing through the plane can be improved.

[0033]

Preferred embodiments of the present invention will be described below with reference to the drawings. Members corresponding to the members described in FIGS. 11 to 16 are denoted by the same reference numerals, and detailed description is omitted.

FIG. 1 is a block diagram showing the configuration of an object detection device for explaining an embodiment of the present invention. In the imaging devices 1 and 2, the left image is divided into blocks in the horizontal direction M and the vertical direction N. Then, the parallax processing unit 5 acquires the parallax S (X, Y) as three-dimensional data of the object for each block, and the plane estimating unit 50 estimates the plane position using the measured parallax.
The parallax change detection unit 8 stores the parallax image stored as image data in the initial parallax image memory 7 at the processing start time t _start as image data in the parallax image memory 6 at time t (t> t _start ). A block in which the parallax in the displayed parallax image has changed by the threshold value TH or more is extracted, and it is determined that a new object exists at a position indicated by the block.

The operation of the object detection device configured as described above will be described with reference to FIG. Time t _start is the processing start time, and an image is input at time t _start ;
The operation will be described separately when an image is input at time t (t> t _start ).

First, the image captured at the time t _start is correlated between the left and right images by the correlating processor 5.
The positional information on the image, that is, parallax S, is output. Parallax is a type of three-dimensional data. The disparity S (X, Y) for each block, which is formed by dividing the left image and stored in the left image memory 3, is input to the plane estimating unit 50, and the position of the plane in the real space is calculated using the measured disparity. Is estimated. Output from the plane estimating unit 50 is a parallax S (X, Y) for each block.
Is the stored initial parallax image.

Next, the flow of processing at the time t (t> t _start ) will be described. The operation up to the association processing unit 5 is the same as that at the time t _start described above. The parallax change detection unit 8 compares the initial parallax image at time t _start with time t (t> t
The parallax S (X, Y) of each block of the parallax image of ( _start ) is compared, and a parallax change equal to or larger than the threshold value TH is detected. When the parallax changes with time, it means that an object has entered or exited. Therefore, a block in which the parallax has changed is output to the object detection image memory 9 and stored as an object detection image by marking or the like. Is done.

Hereinafter, the left image used as the reference image is defined as a horizontal image M
And N blocks in the vertical direction are divided into blocks.
Assuming that (X, Y) has been obtained, a plane estimation method will be described.

FIG. 2 is a flowchart relating to an operation for performing plane estimation. First, in step (S1),
The following fitting of the straight line passing through the plane is performed.

Here, as shown in FIG. 12, for convenience of explanation, the blocks are represented as (Equation 7).

[0041]

## EQU00007 ## Block: BL (X, Y), 1≤X≤M, 1≤Y≤N In the description of the plane estimation method, the disparity data S (X, Y) is added to the indexes X, Y. 3 as in Figure 3
A dimensional SXY reference coordinate system is used.

In FIG. 12, an area 12 surrounded by a thick line is an area generated by fixing the index Y of the block BL (X, Y) and changing only the index X. This area 12 is called GL, and each GL is described as in (Equation 8).

[0043]

## EQU8 ## Region: GL (Y), 1 ≦ Y ≦ N GL includes a maximum of M pieces of parallax data S. GL
As shown in FIG. 4, the disparity data S (X, Y) of the block is plotted on the vertical axis and the block index X is plotted on the horizontal axis for each block included therein. When applying the straight line passing through the plane, since the index Y is fixed, a description will be given by projecting a point on the SXY reference coordinate space onto the SX plane. One Huff curve corresponding to each plotted point (S (X, Y), X) is obtained. Each point (S (X,
When the Hough curves corresponding to all of Y) and X) are plotted on the ρ-θ plane, they are as shown in FIG.

[0044]

Equation 9] ρ = Xcosθ + S (X, Y) sinθ, θ start ≦ θ ≦ θ
_{end The} point at which the Hough curve intersects most in the ρ-θ plane (ρ
_K (Y), θ _K (Y)) and (Equation 9) give a straight line 70 on the SX plane.
It is determined as in (Equation 10).

[0045]

(Equation 10)

As a method of searching for the point where the Hough curve intersects most in the ρ-θ plane, each point on the ρ-θ plane is regarded as a two-dimensional array HG (ρ _j , θ _j ) as shown in FIG. ,point
Each time the Hough curve passes through (ρ _j , θ _j ), the two-dimensional array HG
(ρ _j , θ _j ) is incremented. By searching the two-dimensional array HG (ρ _j , θ _j ) having the highest frequency, the point on the ρ-θ plane through which the Hough curve passes most often
(ρ _K (Y), θ _K (Y)) are obtained.

The same processing is performed on the parallax data included in all the areas GL (Y), 1 ≦ Y ≦ N, and a total of N plane-passing straight lines are obtained. FIG. 6 shows each straight line passing through the plane using SXY reference coordinates. Vertical inclination θ _K (Y)
And the distance from the Y-axis ρ _K (Y) to the error θ _E (Y), ρ
_{Since E} (Y) is included, each highly reliable value θ _R ,
Equation (11) for ρ _R (Y).

[0048]

Θ _K (Y) = θ _R + θ _E (Y), ρ _K (Y) = ρ _R (Y) + ρ _E (Y) Therefore, in the processing so far, the straight line passing through the plane is not necessarily in the same plane. There is no, but it will be corrected in subsequent processing.

Correction of inclination of straight line passing through plane (step (S2))
Description.

That is, assuming that the plane-passing straight lines are all in the same plane, the angle θ _K (Y) between the perpendicular to the Y-axis and the X-axis of all the plane-passing straight lines should be the same.
Therefore, each plane passing straight line is replaced with the inclination from the X axis. This makes it possible to unify the inclination of the perpendiculars of the straight lines passing through the planes from the X axis with a more reliable inclination θ _R.

When this operation is expressed by an equation, the inclination θ _K (Y) of the perpendicular from the X axis contains an error component for each straight line passing through the plane as shown in (Equation 12). Through the processing, the error component θ _E (Y) can be removed from the inclinations of all the straight lines passing through the plane as shown in (Equation 13), and the inclination θ _R can be replaced with the most reliable inclination.

[0052]

(Equation 12)

[0053]

(Equation 13)

By this operation, the straight line passing through the plane, which had a different inclination from the inclination of the plane around the Y axis, can be corrected for the inclination. When the straight lines passing through the plane at the current step are represented by SXY reference coordinates, as shown in FIG. 7, all the straight lines passing through the plane have the same inclination.

The distance correction of the straight line passing through the plane (step (S3))
Description.

If there are two planes in the three-dimensional space, and the planes are not parallel to each other, the intersection of the two planes should be a straight line. The straight line 103 that planar object 100 and Y-[rho _K plane 102 that estimates shown in FIG. 8 by utilizing this fact intersect fit using Hough transform. The ρ _K axis is an axis inclined by θ _{R with} respect to the X axis.

The distance of each straight line passing from the Y axis, that is, ρ _K (Y) is plotted on the Y-ρ _K plane as shown in FIG. Points plotted on Y-[rho _K plane _{102 (Y, ρ K (Y} ))
Is subjected to Hough transform using (Equation 14).

[0058]

Ρ ′ = Y cos θ ′ + ρ _K (Y) sin θ ′, θ _start ≦ θ ′ ≦ θ ′ _{end The} point where the Hough curve intersects most in the ρ′−θ ′ plane
(ρ ′ _S , θ ′ _S ) is obtained, and the straight line 110 in FIG. The straight line 110 indicates a distance from the Y axis which is considered most suitable for each plane straight line, and will be referred to as a distance correction straight line. The distance correction straight line 110 is SX-
At the Y coordinate, there is a distance correction straight line 103 in FIG. Further, this distance correction straight line 103 is represented as (Equation 15).

[0059]

(Equation 15)

When the distance from the Y axis of each straight line passing through the plane is corrected to the optimum value ρ _R (Y) using the distance correction straight line 110,
5) becomes (Equation 16),

[0061]

(Equation 16)

The error component is completely removed. When the straight lines passing through each plane are represented by SXY reference coordinates, they all exist on the same plan view as shown in FIG.

Description of determination of plane (step (S4)).

With the above-described phases, all the straight lines passing through the plane are present in the same plane, and the plane is approximated by a group of straight lines passing through the plane. Parallax data S (X, Y) on a plane
Is not determined, a straight line passing through the plane corresponding to GL (Y)
Interpolation can be performed using (Equation 16).

By the way, the estimated planar object cannot physically be imaged at a longer distance such as a road, a floor, or a wall. Therefore, it is associated with an invisible object present at a greater distance than the planar object. It is useless to perform the processing.

To cope with this, the time t
_{At the start} , the maximum value d _max and the minimum value d _min of the parallax measurement range are set to appropriate values according to the distance to the object, and the parallax is calculated between the maximum value d _max and the minimum value d _min. Search (where the maximum value d _max and the minimum value d
The interval between _min is called a parallax search range). Then, at time t (t
> T _start ), it is possible to optimize the maximum value d _max and the minimum value d _min for each block based on the position and the parallax of the planar object estimated at t _start .

In the associating process, the minimum value d _min of the parallax search range is set as the parallax of the planar object. In other words, the maximum value d _max of the distance measurement range is optimized with the distance to the planar object, and the matching process is performed in the optimized parallax search range at time t (t> t _start ).

Conventionally, the parallax search range has been set from the maximum value and the minimum value of the distance to the object to be measured in the photographed real space, and has been uniform in all blocks of the image. However, according to the present invention, in the case of the positional relationship between the camera 42 and the plane object 41 as shown in FIG. 16, distance measurement is not performed on an object located behind the plane object 41 estimated based on the camera 42. As a result, the amount of calculation can be greatly reduced.

In the present embodiment, an example in which the parallax of an object is used as three-dimensional data has been described. However, the same operation can be performed by replacing the distance to the object.

As described above, the object detecting apparatus and the plane estimating method described in the present embodiment are for estimating the position of a planar object from the measured parallax data using Hough transform processing.

Therefore, by using the Hough transform, which is a global linear approximation method, the three-dimensional position of a plane object can be accurately estimated without being affected by objects other than the plane object.

In addition, by interpolating the undetermined parallax data in the same plane by calculation from the position of the estimated plane, the object can be detected with high accuracy.

Further, since the parallax search range is optimized based on the position of the image based on the estimated plane position information, the amount of calculation for association can be reduced.

[0074]

As described above, according to the object detecting apparatus and the plane estimating method according to the present invention, the Hough transform processing is used to globally and accurately detect three-dimensional objects such as roads and floors.
The position in the dimensional space can be estimated.

Further, since the parallax data of the unmeasured area can be estimated from the estimated position of the planar object,
Object detection can be performed with high accuracy using the position of the planar object as a comparison reference.

Further, since the parallax search range of the associating process can be optimized by using the position of the planar object, the amount of calculation can be reduced.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of an object recognition device for describing an embodiment of an object detection device and a plane estimation method according to the present invention.

FIG. 2 is a flowchart of a plane estimation process according to the embodiment of the present invention.

FIG. 3 is an explanatory diagram for explaining a three-dimensional coordinate system serving as a reference in the embodiment of the present invention.

FIG. 4 is an explanatory diagram illustrating fitting of a plane-passing straight line according to the embodiment of the present invention.

FIG. 5 is an explanatory diagram for explaining linear approximation by Hough transform in the embodiment of the present invention.

FIG. 6 is an explanatory diagram showing a result of a process (step S1 in FIG. 2) of the plane estimation method in the embodiment of the present invention.

FIG. 7 is an explanatory diagram showing a result of a process (step S2 in FIG. 2) of the plane estimation method in the embodiment of the present invention.

FIG. 8 is an explanatory diagram showing a distance correction straight line according to the embodiment of the present invention.

FIG. 9 is an explanatory diagram showing a state of fitting a distance correction straight line according to the embodiment of the present invention.

FIG. 10 is an explanatory diagram showing a result of a process (step S3 in FIG. 2) of the plane estimation method according to the embodiment of the present invention.

FIG. 11 is a block diagram showing a configuration of a conventional object recognition device.

FIG. 12 is an explanatory diagram showing a block arrangement in a left image serving as a reference.

FIG. 13 is an explanatory diagram showing a parallax search range of a right image in a conventional technique.

FIG. 14 is an explanatory diagram illustrating a straight line parameter used for the Hough transform according to the related art.

FIG. 15 is an explanatory diagram illustrating Hough transform in a conventional technique.

FIG. 16 is an explanatory diagram illustrating a general positional relationship between an imaging device and a planar object.

[Explanation of symbols]

1, 2, imaging device 3, 4, image memory 5, association processing unit 6, parallax image memory 7, initial parallax image memory 8, parallax change detection unit 9, object detection image memory 50 Plane estimator, 60, 81, 91, 100, 121 ...
Plane object, 82, 92, 122 ... straight line passing through the plane, 102 ... Y-
ρ _K plane, 103,110 ... Distance corrected straight line.

Claims (6)

[Claims] A plurality of image pickup means arranged at a predetermined interval;
By performing a Hough transform process on the measured three-dimensional data, and a correspondence processing unit that measures three-dimensional data of the object based on the principle of triangulation using a plurality of images output from these imaging units. An object detection device, comprising: a plane position estimating unit that estimates a position of a planar object in a three-dimensional space. 2. The planar position estimating unit estimates the position of a planar object in an image in a three-dimensional space, and converts the three-dimensional data of a rectangular area for which no three-dimensional data was obtained in the same plane. 2. The object detection apparatus according to claim 1, wherein the object detection apparatus calculates and interpolates from the estimated position of the planar object, and detects an object located at a position different from the planar object on the basis of the imaging means. 3. The planar position estimating unit estimates the position of a planar object in an image in a three-dimensional space, and for an object located farther from the planar object with respect to an imaging unit, the three-dimensional object by the association processing unit. 2. The object detection device according to claim 1, wherein data measurement is not performed. 4. Using a plurality of images output from a plurality of imaging means arranged at a predetermined interval, three-dimensional data is obtained for each rectangular small area obtained by dividing the image into a horizontal direction and a vertical direction, For each large area obtained by grouping the rectangular small areas in the horizontal direction, a plane-passing straight line is applied by a Hough transform from a plurality of three-dimensional data included in the large area, and the above-described arithmetic processing is performed on all the large areas. A plane estimating method characterized by estimating a position of a planar object existing in a three-dimensional space by performing the method. 5. Using a slope of a straight line, which is one of the straight line parameters of the straight line passing through the plane, applied to each of the large areas, and sorting the slopes of the total number of straight lines passing through the respective planes by magnitude, thereby obtaining the slope of the straight line. The median value is obtained, and the inclination of all the straight line passing through the plane is replaced by the median value.
5. The plane estimation method according to claim 4, wherein the position of the plane object existing in the dimensional space is estimated. 6. A distance correction straight line is applied by a Hough transform using a distance from the origin, which is a straight line parameter of the straight line passing through the plane whose inclination has been corrected, and a straight line whose inclination has been corrected so as to intersect the distance correction straight line. 6. The plane estimating method according to claim 4, wherein the position of the planar object existing in the three-dimensional space is estimated using a straight line group obtained by performing the parallel movement.