JP2009276073A - Plane estimating method, curved surface estimating method, and plane estimating device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a plane estimating method which enables accurate estimation of coordinates of a plane without giving precise knowledge beforehand, in regard to the plane estimating method for estimating the coordinates of an unknown plane, based on the coordinates of a plurality of points in a three-dimensional space. <P>SOLUTION: A plane estimating device 20 is installed on a floor surface 10. The plane estimating device 20 acquires the coordinates of observation points in an observation space A and selects, from the observation points, candidate points which may constitute the floor surface 10. Based on the coordinates of the candidate points, a temporary approximate plane is calculated. A processing wherein the candidate points located at positions apart from this approximate plane are removed as outliers and the approximate plane is calculated again, is repeated. Lastly the coordinates of the floor surface 10 are calculated based on the candidate points left after the removal and on the observation points acquired first. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a plane estimation method, a curved surface estimation method, and a plane estimation apparatus, and more particularly to a method for estimating coordinates of an unknown plane or curved surface based on the coordinates of a plurality of points in a three-dimensional space.

  When working with a robot or the like, it is important to recognize the coordinates of a plane such as a floor. In particular, in order to accurately perform operations such as walking, unloading, and unloading, it is necessary to recognize a reference plane serving as a reference for the operation with high accuracy. In order to detect an object installed on a plane such as a floor or a desk, it is necessary to first recognize the installation surface on which the object is placed.

As a conventional method for recognizing a plane, for example, there are a method described in Patent Document 1 and a method described in Patent Document 2.
In the method of Patent Document 1, a road surface is photographed using an in-vehicle camera, and the road surface is recognized based on the photographed image and a parameter specified in advance. Then, the recognition state of the road surface is displayed over the captured image, and the user sees this to determine whether the parameter is appropriate.
In the method of Patent Document 2, the position and the depression angle of the in-vehicle camera are handled as known values. Based on these values, coordinate conversion is performed on the photographed image to calculate a plane that approximates the road surface.

JP-A-2005-62924 JP 2003-178309 A

However, the conventional method has a problem that it is necessary to give precise knowledge beforehand in order to accurately estimate the coordinates of the plane.
In the method of Patent Document 1, since the recognition of the plane depends on a parameter defined in advance, it is necessary to determine this parameter precisely. In addition, since the user determines whether or not the parameter is appropriate, the parameter cannot be automatically adjusted. For this reason, it is difficult to accurately estimate the plane coordinates without precise parameters.
Even in the method of Patent Document 2, it is necessary to know the position and depression angle of the camera, and it is difficult to accurately estimate the coordinates of the plane if such knowledge is not given precisely.

  The present invention has been made to solve such a problem, and can provide a plane estimation method, a curved surface estimation method, and a plane that can accurately estimate the coordinates of the plane without giving precise knowledge beforehand. An object is to provide an estimation device.

  In order to solve the above-described problems, a plane estimation method according to the present invention estimates a plane of an unknown plane based on the coordinates of a plurality of observation points using a plane estimation apparatus in a three-dimensional space. A method of selecting candidate points from observation points; a step of removing an outlier repeated a plurality of times, a sub-step of calculating approximate plane coordinates based on the coordinates of candidate points; and candidate points Outlier removal step including a substep of excluding from the candidate points those whose distance from the approximate plane exceeds a predetermined threshold, and the coordinates of the unknown plane based on the coordinates of the candidate points not excluded Calculating.

An observation point means a point where coordinates can be specified in a three-dimensional space. Observation points include points that constitute an unknown plane, but may also include points that do not constitute an unknown plane. The candidate point means a point that may constitute an unknown plane among the observation points.
In the plane estimation method according to the present invention, a temporary approximate plane is first calculated based on the coordinates of the candidate points. Then, the candidate point at a position away from the approximate plane is removed as an outlier, and the process of calculating the approximate plane again is repeated. In this way, candidate points that are likely to constitute an unknown plane are narrowed down. Finally, the coordinates of the unknown plane are calculated based on the candidate points remaining without being removed and the observation points obtained first.

The step of calculating the coordinates of the unknown plane includes substeps in which observation points whose distance from the last calculated approximate plane is less than a predetermined threshold are estimated plane points, and the coordinates of the estimated plane points. And a sub-step of calculating the coordinates of the unknown plane based thereon.
Here, the estimated plane point means a point used for finally estimating an unknown plane. In this way, even if the observation point is not included in the final candidate point, a point that is likely to actually constitute an unknown plane is included in the estimated plane point, and can be trusted as a basis for plane estimation. The point information can be increased.
The plane estimation apparatus may include a stereo camera, and the stereo camera may be stationary with respect to an unknown plane.
The unknown plane may be a plane constituted by the floor or a plane parallel to the plane constituted by the floor.
Here, the plane parallel to the plane formed by the floor means a plane on which a plane estimation device or some other object can be placed or exist, including a floor surface of a building, a ground surface, a desk surface, and the like.
The step of selecting candidate points from among the observation points is the sub-step of classifying each of the observation points into one of a plurality of areas extending in the vertical direction, and the lowest position among the observation points in each of the areas. And sub-steps using a candidate as a candidate point.
You may further include the step which excludes a candidate point from which the position of a perpendicular direction is not contained in a predetermined range among candidate points.
Thus, the estimation accuracy of the coordinates of the unknown plane can be increased by selecting candidate points based on the vertical coordinates and filtering.
The least square method may be used in at least one of the sub-step of calculating the coordinates of the approximate plane and the step of calculating the coordinates of the unknown plane.
You may further include the step which acquires the coordinate of an observation point based on an image.

  The curved surface estimation method according to the present invention is a curved surface estimation method for estimating the shape of an unknown curved surface in a three-dimensional space, the step of dividing the unknown curved surface into a plurality of unknown planes, and the divided unknown For each of the planes, estimating the coordinates using the plane estimation method described above.

  The plane estimation apparatus according to the present invention is a plane estimation apparatus that estimates the coordinates of an unknown plane based on the coordinates of a plurality of observation points in a three-dimensional space, and selects candidate points from the observation points. And repeatedly calculating the coordinates of the approximate plane based on the coordinates of the candidate points and excluding those candidate points whose distance from the approximate plane is equal to or greater than a predetermined threshold from the candidate points. Based on the coordinates of the candidate points, the coordinates of the unknown plane are calculated.

  According to the plane estimation method, the curved surface estimation method, and the plane estimation apparatus according to the present invention, candidate points located away from an unknown plane can be removed with high accuracy as an outlier even if precise knowledge is not given in advance. It is possible to appropriately narrow down candidate points and estimate the coordinates of an unknown plane or curved surface with high accuracy.

Embodiments of the present invention will be described below with reference to the accompanying drawings.
Embodiment 1 FIG.
FIG. 1 shows an arrangement of a floor surface 10 and a plane estimation device 20 installed on the floor surface 10. The plane estimation device 20 is a device that estimates the coordinates of the floor surface 10.
The plane estimation device 20 includes a stereo camera. That is, the plane estimation apparatus 20 includes a plurality of imaging means such as a digital camera, and can capture images of a certain observation space from different angles at a single point and generate a plurality of images. Further, the plane estimation device 20 has a function as a coordinate acquisition unit that acquires coordinates representing the position of a point in the observation space based on a plurality of captured images. Furthermore, the plane estimation device 20 has a function as plane calculation means for estimating coordinates representing the position of the floor surface 10 based on the coordinates of the acquired points.

The plane estimation device 20 is not given any knowledge about the coordinates representing the position of the floor surface 10 in the three-dimensional space in advance, and the floor surface 10 is an unknown plane for the plane estimation device 20.
An X axis, a Y axis, and a Z axis in FIG. 1 indicate axes of an orthogonal coordinate system used when the plane estimation apparatus 20 recognizes a structure in the three-dimensional space. The plane estimation device 20 is installed such that the XZ plane is parallel or substantially parallel to the floor surface 10. In the example of FIG. 1, the XZ plane and the floor surface 10 are not strictly parallel, and form a certain angle as an error. Moreover, the part which the stereo camera of the plane estimation apparatus 20 image | photographs among the three-dimensional spaces containing the floor surface 10 of FIG.
Note that the plane estimation apparatus 20 handles coordinates in the observation space A by normalizing them in order to cope with a change in the size of the observation space A due to a change in the installation position of the plane estimation apparatus 20 or the like. For example, the minimum value is set to 0 and the maximum value is set to 1.0 for each of the X-axis direction, the Y-axis direction, and the Z-axis direction observed in the observation space A.

FIG. 2 is a flowchart for explaining the flow of processing executed when the plane estimation device 20 estimates the coordinates of the floor surface 10.
Hereinafter, the operation of the plane estimation apparatus 20 will be described with reference to FIGS. 1 and 2.
First, the plane estimation device 20 functions as a coordinate acquisition unit. Here, the plane estimation apparatus 20 acquires the coordinates of a plurality of observation points included in the observation space A based on a plurality of images obtained by photographing the observation space A (step S1). A known method is used as a method for acquiring the coordinates of the point in the three-dimensional space. For example, there is a method of detecting an edge that is common to two images and using the end as an observation point. A person skilled in the art can determine how to use these methods to obtain coordinates. This is a matter that can be designed.
In the example of FIG. 1, there are 10 observation points for the observation space A. That is, they are point P ₁₁ , point P ₁₂ , point P ₁₃ , point P ₂₁ , point P ₂₂ , point P ₂₃ , point P ₃₁ , point P _31B , point P ₃₂ , and point P ₃₃ . Among these observation points, the point P ₁₂ , the point P ₁₃ , the point P ₂₁ , the point P ₂₃ , and the point P ₃₂ constitute the floor surface 10, but the point P ₁₁ , the point P ₂₂ , and the point P _{31 are included.} The point P _31B and the point P ₃₃ are points that do not constitute the floor surface 10.

  After acquiring the coordinates of the observation point, the plane estimation device 20 functions as plane estimation means. Here, first, the plane estimation device 20 selects a point that may constitute the floor surface 10 as a candidate point from the observation points acquired in step S1 (steps S2 and S3).

In step S2, the plane estimation apparatus 20 divides the observation space A into a plurality of regions, and classifies each observation point as one of the regions. In the example of FIG. 1, the observation space A is divided into a region A ₁₁ , a region A ₁₂ , a region A ₁₃ , a region A ₂₁ , a region A ₂₂ , a region A ₂₃ , a region A ₃₁ , a region A ₃₂ , and a region A _33. The Here, each region is assumed to extend in the vertical direction. Thus, for example region A ₁₁ includes a portion included in the area A ₁₁ of the floor 10, and the space above them, a space thereunder. Each region divides the observation space A at equal intervals along the X axis and the Z axis. In the example of FIG. 1, the observation space A is divided into three along the X axis and the Z axis, and as a result, nine regions A _{11 to} A ₃₃ are configured.
In FIG. 1, two observation points P ₃₁ and P _31B are classified only in the region A ₃₁ , and observation points are classified one by one in the other regions. In the example of FIG. 1, for convenience of explanation, ten observation points are used. However, actually, a larger number of observation points may exist, and a large number of observation points may be classified in each region.

In step S3, the plane estimation apparatus 20 selects the observation point at the lowest position in each region as a candidate point. In the example of FIG. 1, an observation point with the smallest Y coordinate is a candidate point.
In the example of FIG. 1, in the area _{A 31,} it is in the lower position towards the point _{P 31} than the point _{P 31B,} that is, the Y coordinate is small, only the point _{P 31} is the candidate point. Since there is only one observation point in the other region, each observation point is a candidate point.
In this way, among the ten observation points, the point P ₁₁ , the point P ₁₂ , the point P ₁₃ , the point P ₂₁ , the point P ₂₂ , the point P ₂₃ , the point P ₃₁ , the point P ₃₂ , and the point P ₃₃ are Although it is a candidate point, the point P _31B is not a candidate point.

Next, the plane estimation apparatus 20 filters the candidate points based on the deviation value of the Y coordinate, and excludes those whose Y coordinate is not included in the predetermined range from the candidate points (step S4). Here, the “Y-coordinate deviation value” means the deviation value of the Y-coordinate of each candidate point when a set of Y-coordinates of all candidate points is used as a mother set.
This predetermined range is a range in which the deviation value is 40 or more and 50 or less, for example. In a general observation space, an object exists on the floor surface, and the candidate points may be distributed on both the floor surface and the object. Therefore, the average of the Y coordinates of the candidate points (that is, corresponding to the deviation value 50). (Value) is slightly above the Y coordinate corresponding to the floor surface 10. That is, the deviation value of the Y coordinate of the candidate points constituting the floor surface 10 is a value slightly smaller than 50. Therefore, there is a high possibility that the candidate points included in the range of the deviation values 40 to 50 actually constitute the floor surface 10, and such filtering leads to an improvement in accuracy of the entire process.
In the example of FIG. 1, for example, assuming that the Y coordinate of the point P ₃₃ has a deviation value of 55, and the Y coordinates of other candidate points are within the range of the deviation value of 40 to 50, in step S3 described above, the candidate point _{P 33} point has been considered is excluded from the candidate points in step S4, the candidate point is the point _{P 11} that remains, the point _{P 12,} the point _{P 13,} the point _{P 21,} the point _{P 22,} the point _{P 23,} and the point _{P 31,} There are 8 points P ₃₂ .

Next, the plane estimation device 20 determines that a candidate point that satisfies a predetermined condition is an outlier, and excludes it from the candidate point (steps S5 to S7).
In step S5, the plane estimation apparatus 20 calculates the coordinates of the approximate plane based on the coordinates of the candidate points. This calculation is performed using the least square method. For example, in the example of FIG. 1, when step S5 is first executed, the coordinates of the approximate plane 11 are calculated by the least square method based on the above eight candidate points.
In step S <b> 6, the plane estimation apparatus 20 determines whether there is a candidate point whose distance from the approximate plane 11 exceeds a predetermined threshold T, that is, whether an outlier exists.
If it is determined in step S6 that an outlier exists, the plane estimation apparatus 20 excludes all the outliers from candidate points in step S7.

For example, in FIG. 1, it exceeds the distance threshold value T of the point P ₂₂ and the approximate plane 11, and the distance between the approximate plane 11 and the other candidate points is less than the threshold value T. In such a case, steps S4 _{P 22} point is a candidate point is excluded from the candidate points in step S7, the candidate point is the point _{P 11} that remains, the point _{P 12,} the point _{P 13,} the point _{P 21,} the point P ₂₃ , point P ₃₁ , and point P ₃₂ . Then, the process of the plane estimation apparatus 20 returns to step S5.
As described above, steps S5 to S7 constitute an outlier removal step, and this outlier removal step is repeated a plurality of times according to the conditional branch in step S6. Here, since the approximate plane 11 is recalculated each time the outlier is excluded, the coordinates of the approximate plane 11 change each time. In such an iterative process, the point P ₁₁ and the point P ₃₁ that do not constitute the floor surface 10 are sequentially excluded from the candidate points as outliers, and accordingly the accuracy of approximation gradually improves.

As described above, this outlier removal step improves the accuracy of approximation as a whole, but in some cases, there is a slight possibility that a point that actually constitutes the floor surface 10 is determined as an outlier. Exist. For example, in practice in Fig. 1 P ₃₂ points constituting the floor surface 10 is determined to outliers, and were excluded from the candidate point. If other candidate points are not excluded in the subsequent iterations, the remaining candidate points are four points of point P ₁₂ , point P ₁₃ , point P ₂₁ , and point P ₂₃ .

When it is determined in step S6 that no outlier exists, the plane estimation device 20 calculates the coordinates of the floor surface 10 based on the coordinates of candidate points that have not been excluded so far (steps S8 and S9).
The approximate plane 11 based on the candidate points that have not been excluded until that time, that is, the approximate plane 11 calculated last is already determined at the time when the above step S5 is finally executed. In step S8, the plane estimation device 20 selects all the observation points whose distance from the approximate plane 11 is less than the threshold value T, and finally estimates the floor surface 10 for the selected observation points. Are estimated plane points.

Here, the estimated plane points are selected not only from candidate points that are not excluded until the end but also from all observation points. That is, observation points that have not become candidate points in step S3 and observation points that have been excluded from candidate points in steps S4 and S7 may also be estimated plane points. For example, in FIG. 1, it is assumed that the approximation accuracy of the approximate plane 11 finally increases, and as a result, the distance between the observation point P ₃₂ once excluded and the approximate plane 11 is less than the threshold T. In this case, the estimated plane points are five points of point P ₁₂ , point P ₁₃ , point P ₂₁ , point P ₂₃ , and point P ₃₂ .
Here, since the approximation accuracy of the approximate plane 11 is increased by repeating Steps S <b> 5 to S <b> 7, it can be said that a point close to the approximate plane 11 is highly likely to actually constitute the floor surface 10. As described above, even if the observation point is not included in the final candidate point, a point that is likely to actually constitute the floor surface 10 is included in the estimated plane point. Can increase information.
In step S9, the plane estimation device 20 calculates the coordinates of the floor 10 based on the coordinates of the estimated plane point. This calculation is performed using the least square method.

As described above, according to the plane estimation apparatus 20 according to the first embodiment, the floor approximation by the least square method and the outlier removal are repeatedly performed even if precise knowledge about the floor surface 10 is not given in advance. Candidate points located away from the surface 10 can be removed as an outlier with high accuracy, so that the candidate points can be appropriately narrowed down and the coordinates of the floor surface 10 can be estimated with high accuracy.
In addition, by applying such plane estimation, the accuracy of the reference plane recognition required for operations such as walking, unloading, and unloading of robots is improved, so that devices such as robots operate more appropriately. It can be performed.

  In addition, even if the observation point is not included in the final candidate point, the point near the last calculated approximate plane 11 is the estimated plane point as a point that is likely to actually constitute the floor surface 10. Since it is included, the information of the reliable point can be increased and the estimation accuracy of the coordinates of the floor surface 10 can be further increased.

  In addition, since the plane estimation apparatus can use a stereo camera that is stationary with respect to the floor 10 as an imaging unit, there is no need to acquire a plurality of images at different time points while moving the imaging apparatus. Processing contents can be simplified. Of course, the plane estimation device 20 or the stereo camera may be moved, and the image is taken at different positions while moving with respect to the floor surface 10, and the coordinates of the observation point are acquired based on the result. May be.

  Further, since the plane estimation device 20 is installed so that the XZ plane is substantially parallel to the floor surface 10, “the floor surface 10 exists at a position lower than an object placed on the floor surface 10”. Knowledge can be used. That is, the Y coordinate can be used as a coordinate representing the height of the point, and the estimation accuracy of the coordinate of the floor surface 10 can be increased by selecting candidate points based on the Y coordinate and filtering. This effect is sufficiently obtained even when the XZ plane is not strictly parallel to the floor surface 10 and has a slight inclination as shown in FIG. 1, and precise positioning is performed when the plane estimation device 20 is installed. Is not required. For this reason, the process at the time of installing the plane estimation apparatus 20 can be simplified.

Note that the knowledge that “the floor surface 10 exists at a position lower than the object placed on the floor surface 10” is that the closer the XZ plane is to the floor surface 10, that is, the smaller the angle formed between them. Can be used efficiently. However, in principle, if the direction that the plane estimation device 20 recognizes as “upward”, that is, the direction in which the Y coordinate increases, is higher than the horizontal direction, it is compared with a method that does not use prior knowledge at all. The estimation accuracy may be improved. In other words, if the plane estimator 20 is installed so that it does not turn upside down, the estimation accuracy may be improved even if the vertical axis of the plane estimator 20 is in a state of being nearly lying down. .
However, such a configuration may not be used, and for example, step S3 and step S4 in FIG. 2 may not be executed. In this case, more general plane estimation can be performed without being limited to the horizontal plane in exchange for a decrease in estimation accuracy.

In the above-described first embodiment, the observation space A is divided into nine regions A _{11 to} A ₃₃ , but the number of regions may be different from this. The number of regions to be divided may be appropriately determined according to the accuracy of plane estimation and the time required for the entire calculation.

  In the first embodiment, the range of the deviation value in step S4 is 40 or more and 50 or less. However, the upper limit and the lower limit of this range may be different values, and are determined based on, for example, experiments. Also good. Moreover, the value used for determination may not be a deviation value as long as it is a value calculated based on the Y coordinate, and may be, for example, the difference between the average of the Y coordinates of all points and the Y coordinate of each point.

  In the first embodiment, the predetermined threshold value T is set to an appropriate value in consideration of an error included in the coordinates of the observation point. Here, when the threshold value T is increased, a large number of estimated plane points can be secured, and an improvement in accuracy according to this can be expected. On the other hand, if the threshold value T is reduced, the point that does not actually constitute the floor surface 10 can be more reliably excluded, and an improvement in accuracy corresponding to this can be expected. The threshold value T may be determined in consideration of the balance of these effects.

In the first embodiment, the plane estimation device 20 functions as a coordinate acquisition unit and a plane calculation unit. As a modification, the coordinate acquisition unit or the plane calculation unit may be configured as a device external to the plane estimation device 20, or may be configured to be distributed among a plurality of devices. For example, in the configuration in which the coordinate acquisition unit is not included in the plane estimation device 20, the plane estimation device 20 receives the coordinates of the observation point from the outside instead of acquiring the coordinates of the observation point based on the image in step S1. Alternatively, the coordinates of observation points recorded externally may be read. Thereafter, the processing after step S2 may be performed based on the coordinates of these observation points.
Further, the plane estimation device 20 does not need to include a stereo camera, and may include other well-known means as means for acquiring the coordinates of the observation point in the observation space A.

  The first embodiment assumes that the floor surface 10 is a flat surface or a substantially flat surface, but the same processing as in the first embodiment can be performed even when the floor surface 10 is a curved surface. In this case, the plane estimation device 20 calculates the coordinates of the plane that approximates the floor curved surface as the coordinates of the floor curved surface.

Embodiment 2. FIG.
The second embodiment is different from the first embodiment in that the unknown plane whose coordinates are to be estimated is different from the plane on which the plane estimation device 20 is installed. Hereinafter, differences from the first embodiment will be described.
FIG. 3 shows the positional relationship between the floor surface 10, the plane estimation device 20 installed on the floor surface 10, and the desk surface 30 which is the surface of the desk installed on the floor surface 10. In the second embodiment, the desk surface 30 is an unknown plane whose coordinates should be estimated.
Of the three-dimensional space including the desk surface 30, a portion captured by the plane estimation device 20 in the second embodiment is represented as an observation space B. Note that FIG. 3 shows that the observation space B is divided into nine regions.
The plane estimation device 20 estimates the coordinates of the desk surface 30 included in the observation space B based on the coordinates of the observation points in the observation space B. This estimation is performed according to the flowchart of FIG. 2 in the same manner as the processing for estimating the coordinates of the floor surface 10 in the observation space A in the first embodiment.

In the second embodiment, the floor surface 10 on which the plane estimation device 20 is installed is different from the desk surface 30 that is an unknown plane, but the plane estimation device 20 according to the second embodiment is a table surface. By setting the observation space B on 30, the coordinates of the desk surface 30 can be estimated with the same accuracy as in the first embodiment.
The unknown plane is not limited to the desk surface 30 and may be any plane as long as the plane is parallel or substantially parallel to the plane on which the plane estimation device 20 is installed.

Embodiment 3 FIG.
The third embodiment is configured to apply the estimation of the plane in the first embodiment to estimate the shape of an unknown curved surface. Hereinafter, differences from the first embodiment will be described.
FIG. 4 shows an arrangement of the floor curved surface 15 and the curved surface estimation device 25 installed on the floor curved surface 15. The shape and position of the floor curved surface 15 are unknown to the curved surface estimation device 25. The curved surface estimating device 25 has the function of the flat surface estimating device 20 in the first embodiment, and estimates the shape of the floor curved surface 15 using this function.

  The curved surface estimation device 25 executes the processing shown in the flowchart of FIG. However, a step of dividing the floor curved surface 15 into a plurality of planes is executed before step S1 of FIG. In the example of FIG. 4, the entire observation space including the floor curved surface 15 is divided into a plurality of smaller observation spaces, and each divided observation space includes one plane. Thereby, the floor curved surface 15 is divided into a plurality of planes (in FIG. 4, only four observation spaces C1 to C4 corresponding to the four planes are illustrated). This division divides the observation space corresponding to the floor curved surface 15 at equal intervals along the X axis and the Z axis.

After dividing the floor curved surface 15 into a plurality of planes in this way, the curved surface estimation device 25 executes the plane estimation method of the first embodiment for each plane and estimates coordinates. FIG. 4 shows that the planes corresponding to the observation spaces C1 to C4 are each divided into nine regions in the same manner as in the first embodiment.
Furthermore, the curved surface estimation device 25 estimates the shape of the floor curved surface 15 based on the estimated coordinates of each plane. For example, the shape of the floor curved surface 15 is estimated to be a shape combining all estimated planes. Further, some interpolation processing may be performed at the boundary between different planes. This interpolation processing may be processing for estimating only the portion corresponding to the boundary as an observation space and estimating the coordinates of the plane included therein according to the flowchart of FIG.
As described above, the curved surface estimation apparatus 25 according to the third embodiment estimates the shape of the floor curved surface 15 based on the coordinates of the plane estimated with high accuracy, so the shape of the floor curved surface 15 is estimated with high accuracy as a whole. can do.

In Embodiment 1, it is a figure which shows arrangement | positioning with a floor surface and the plane estimation apparatus installed in the floor surface. It is a flowchart explaining the flow of the process performed when a plane estimation apparatus estimates the coordinate of a floor surface. In Embodiment 2, it is a figure which shows the positional relationship with the floor surface, the plane estimation apparatus installed in the floor surface, and the surface of the desk installed in the floor surface. In Embodiment 3, it is a figure which shows arrangement | positioning with a floor curved surface and the curved surface estimation apparatus installed in the floor curved surface.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Floor surface (Unknown plane, plane which a floor comprises), 11 Approximation plane, 15 Floor curved surface (Unknown curved surface), 20 Plane estimation apparatus (coordinate acquisition means, plane calculation means), 25 Curved surface estimation apparatus, 30 Desk surface (Unknown plane, plane parallel to the plane formed by the floor), A _{11 to} A ₃₃ region, P _{11 to} P ₃₃ points (observation point, candidate point), P _31B point (observation point).

Claims (10)

A plane estimation method for estimating coordinates of an unknown plane based on coordinates of a plurality of observation points using a plane estimation device in a three-dimensional space,
Selecting candidate points from the observation points;
An outlier removal step that is repeated multiple times,
A sub-step of calculating the coordinates of the approximate plane based on the coordinates of the candidate points;
An outlier removal step comprising: substeps of excluding candidate points whose distance from the approximate plane exceeds a predetermined threshold from the candidate points;
Calculating the coordinates of the unknown plane based on the coordinates of the candidate points that were not excluded. Calculating the coordinates of the unknown plane,
Among the observation points, a sub-step in which an estimated plane point is a distance that is less than the predetermined threshold value from the last calculated approximate plane;
The plane estimation method according to claim 1, further comprising a sub-step of calculating coordinates of the unknown plane based on coordinates of the estimated plane point.   The plane estimation method according to claim 1, wherein the plane estimation device includes a stereo camera, and the stereo camera is stationary with respect to the unknown plane.   The plane estimation method according to any one of claims 1 to 3, wherein the unknown plane is a plane constituted by a floor or a plane parallel to a plane constituted by the floor. The step of selecting candidate points from the observation points includes:
A sub-step of classifying each of the observation points into one of a plurality of regions extending in a vertical direction;
5. The plane estimation method according to claim 1, further comprising: a substep in which a candidate point is the lowest observation point in each of the regions.   The plane estimation method according to claim 5, further comprising the step of excluding from the candidate points those candidate points whose vertical positions are not included in a predetermined range.   The least square method is used in at least one of the sub-step of calculating the coordinates of the approximate plane and the step of calculating the coordinates of the unknown plane, according to any one of claims 1 to 6. Plane estimation method.   The plane estimation method according to claim 1, further comprising a step of acquiring coordinates of the observation point based on an image. A surface estimation method for estimating the shape of an unknown curved surface in a three-dimensional space,
Dividing the unknown curved surface into a plurality of unknown planes;
A surface estimation method including a step of estimating coordinates for each of the divided unknown planes using the plane estimation method according to claim 1. A plane estimation device that estimates the coordinates of an unknown plane based on the coordinates of a plurality of observation points in a three-dimensional space,
A candidate point is selected from the observation points,
Repeatedly calculating the coordinates of the approximate plane based on the coordinates of the candidate point, and excluding those candidate points whose distance from the approximate plane is equal to or greater than a predetermined threshold from the candidate points,
A plane estimation apparatus that calculates coordinates of the unknown plane based on the coordinates of the candidate points that are not excluded.

Images