KR100964029B1 - 3-dimensional object or environment representation method for multi resolution octree structure - Google Patents

Abstract

The present invention relates to a multi-resolution octree-based object or environment representation method, wherein in representing a three-dimensional data point set obtained from a camera or a three-dimensional measuring device as a volume-based octree, whether or not to divide the next stage of the unit octree cell and The depth of the partitioning step may be determined based on the point density and the point distribution in the unit octree cell.

By having the above-described configuration, it is possible to prevent unnecessary waste of octree cells and increase in octree generation time, thereby enabling more accurate and faster multi-resolution octree representation.

Multi-Resolution, Octry Cell, Three-Dimensional Object, Environment, Point Density, Point Distribution

Description

3-Dimensional object representation environment based on multi-resolution octree {3-DIMENSIONAL OBJECT OR ENVIRONMENT REPRESENTATION METHOD FOR MULTI RESOLUTION OCTREE STRUCTURE}

The present invention relates to a multi-resolution octree-based three-dimensional object or an environment representation method, and more particularly, to determine whether or not to decompose and decompose a unit octree cell in consideration of the point density and the point distribution in the octree cell. The present invention relates to an octree-based object or environment representation method.

In the field of intelligent robots, environment modeling is very important for robots to manipulate their own objects. In this environment modeling, the robot must express objects other than the objects that the robot currently knows so that the robot can manipulate the desired objects by avoiding other objects. For this purpose, the octree structure, which is one of graphic objects or environmental representation methods, is used. The octree structure divides the space into eight cubes (octree cells) and expresses the actual object as a set of cubes, using the midpoint and size of the cube to determine whether or not the points are included.

There are two main methods of decomposing the existing octree cell.

The first method is to divide all octrees by the same size (same resolution).

Using this method, all octrees are commonly divided into the original size, regardless of the spatial distance or the size of the object.

However, there are some problems with this method. In the case of configuring the octree cells of too small size, the number of octree cells increases so much that not only the operation speed of octree avoidance decreases but also the computation time of the octree cells increases. On the contrary, when the size of the octree cell is large, the number of cells will be reduced, but the detail will not be able to express the details. These problems make it difficult for the robot to move in real time and model its environment.

The second method is to divide the octree size according to the distance. In general, this method expresses objects that are relatively far from the camera as large octree cells, and conversely, objects that are close to the camera are represented by small octree cells.

However, the biggest problem with this method is the noise problem. When noise occurs at a distance, a very large octree is created to represent that noise. In this case, accurate robot manipulation and environment modeling becomes difficult. In addition, the objects that are in close proximity are represented by too small octrees, which greatly wastes time for representing objects and manipulating robots.

Most of the problems listed above stem from the inability to properly create multiresolution octrees. Of course, the octree representation of the multi-resolution according to the distance is implemented, but since it is represented by the same octree cell size for the objects at the same distance, it is difficult to say that it is a true multi-resolution octree representation. This multi-resolution representation requires a definite criterion to stop decomposition when creating and decomposing octree cells. However, in the conventional octree decomposition method, the octree decomposition is stopped only when the number of points of one or more points or more is exceeded, which makes it difficult to express the octree in multiple resolutions.

In the conventional octree decomposition method, that is, dividing all octree cells into the same size or generating an octree cell with a different size according to a distance, the reference element for stopping the octree decomposition is simply a point presence / absence. Therefore, when decomposing the octree on the basis of this criterion, not only causes unnecessary octree representation, but also increases the octree generation time.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and in terms of volume-based octree representing a three-dimensional point set obtained from a camera or a three-dimensional measuring device, the unit octree cell is based on the point density and the point distribution in the octree. By deciding whether to decompose and decompose depth, it provides a multi-resolution octree-based object or environment representation method that can reduce the generation time of the cell for octree representation and accurately and efficiently perform the multi-resolution octree representation. There is an object of the invention.

According to an aspect of the present invention, in expressing a set of three-dimensional data points obtained from a camera or a three-dimensional measuring device in a volume-based octree, whether or not the next step of the unit octree cell and the depth of the step are divided within the unit octree cell. There is provided a multi-resolution octree-based three-dimensional object or environment representation method that is determined based on point density and point distribution.

In the multi-resolution octree-based three-dimensional object or environment representation method of the present invention, the point density is the focal length of the camera (F), one point size (CPs) on the camera CCD, the camera obtained from the 3D camera and the corresponding The number of points generated in the unit octree cell of the same volume varies according to the distance Z between the points. In addition, the point distribution is calculated using the average and the variance of the coordinate values of the points in the octree. When the point density and the point distribution in the octree cell calculated as above are higher than the predetermined point density reference value or point distribution reference value, the decomposition of the octree cell is stopped, and if none of the conditions is satisfied, the octree decomposition to the next step is determined. . On the other hand, by setting the point density reference value and the point distribution reference value, which are the octree decomposition criteria according to the working environment, the environment can be represented by octree cells of various sizes suitable for the working environment.

In addition, according to another aspect of the present invention, in expressing a set of three-dimensional data points obtained from a camera or a three-dimensional measuring device in a volume-based octree, the size of an area occupied by one point in real space using a camera parameter (RPs) and calculate the point density (Pd) in the octree from the calculated area size (RPs), the size of the octree cell (OCTs) currently divided, and the number of points (N) included in the octree cell. A first step of calculating; A second step of comparing the point density Pd calculated in the first step with a predetermined threshold value and calculating an average and a variance value of coordinates of points in the corresponding octree cell when the point density Pd is equal to or greater than a threshold value ; And a third step of determining a point distribution in the octree cell using the average and the variance calculated in the second step, and determining whether the corresponding octree cell is decomposed according to the distribution. A three-dimensional object or environment representation method is provided.

In the second step, the point density Pd calculated in the first step is compared with a predetermined threshold, and when the point density Pd is smaller than the threshold, determining the decomposition of the octree cell to the next step. It includes more. The third step stops decomposition of the octree cell when the point distribution calculated in the second step is greater than or equal to a predetermined threshold value, and moves to the next step of the octree cell when the calculated point distribution is smaller than the threshold value. It may be provided with a configuration for determining the decomposition.

According to the multi-resolution octree-based three-dimensional object or environment representation method of the present invention, it is necessary to determine whether to decompose the octree cell in the next step based on the point density and the point distribution in the octree cell. And it is possible to prevent the increase in the octree generation time, there is an effect that it is possible to more accurately and quickly express the octree of the multi-resolution.

In addition, according to the present invention, when the robot manipulates or moves an actual object, an object may be represented by generating an octree of the same size regardless of the position of the robot, and by generating an octree of various sizes according to the nature of the work. By doing so, there is an effect that can achieve precise and high speed robot operation.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

Points obtained from a 3D camera or a 3D measuring device are 3D points, and octree cells divided by these 3D points are also generated in real space. In the multi-resolution octree-based three-dimensional object or environment representation method according to the present invention, whether or not actual points are filled in an octree cell in this space is determined to determine whether to continue or stop octree decomposition. In the present invention, a criterion for distinguishing whether or not actual points are filled in an octree cell is defined as a fill factor.

Specifically, in the present invention, the point density and the point distribution in the unit octree cell are adopted as a fill factor that is a reference for decomposition and generation of the octree cell. That is, according to the present invention, when a three-dimensional data point set obtained from a camera or a three-dimensional measuring apparatus is represented as a volume-based octree, whether or not the next step of the unit octree cell is divided according to the point density and the point distribution in the unit octree cell; Configured to determine the depth of the segmentation step.

1 is a flow chart showing a multi-resolution octree-based three-dimensional object or environment representation method according to a preferred embodiment of the present invention.

As shown, the multi-resolution octree-based three-dimensional object or environment representation method according to the present invention calculates the size of an area occupied by one point in real space using camera parameters, and calculates the size of the calculated area and Calculating a point density in an octree cell using the number of points (S101 to S103); A second step (S104 to S105) of comparing the point density calculated in the first step with a predetermined threshold value and calculating an average and a variance value of coordinates of points in the corresponding octree cell when the point density is equal to or greater than a threshold value; And a third step (S106 to S108) of checking the point distribution in the octree using the average and the dispersion value calculated in the second step, and determining whether the corresponding octree is decomposed according to the distribution.

Multi-resolution octree-based three-dimensional object or environment representation method according to the present invention is an image input device consisting of a three-dimensional measuring device or a 3D camera, and the like, parameters for the image input device and 3 transmitted from the image input device It can be executed by data processing means for performing a process for determining whether to decompose an octree cell using the image data of the dimensional representation. In addition, the data processing means generates a two-dimensional image signal from the image data of the three-dimensional representation by generating an octree cell according to the method defined by the present invention, the generated two-dimensional image is to be displayed by the display device desirable.

In the first step, the point density is the unit octree of the same volume according to the focal length (F) of the camera, one point size (CPs) on the camera CCD, and the distance (Z) between the camera and the corresponding point obtained from the 3D camera. It is calculated based on the fact that the number of points generated in the cell varies.

Specifically, the first step is to calculate the point density, the focal length (F) of the camera, one point size (CPs) on the camera CCD, and the distance (Z) between the camera and the point obtained from the 3D camera And calculating the size (RPs) of the area occupied by one point in the real space from the proportional relation between the size (RPs) of the area occupied by one point in the real space (S101, S102), and the calculated actual Computing the point density Pd in the octree by using the size RPs of the area occupied by one point in the space and the number of points included in the octree cell (S103).

In order to calculate the size (RPs) of the area occupied by one point in the real space, the relationship between the camera parameters shown in FIG. 2 may be used. As shown in Fig. 2, the focal length F of the camera, one point size CPs on the camera CCD, the distance Z between the camera and the corresponding point obtained from the 3D camera, and one in real space A proportional relationship, as shown in Equation 1 below, may be established between the size RPs of the area occupied by.

CPs: RPs = F: Z

The size (RPs) of the area occupied by one point in the real space is calculated from Equation 1 below.

RPs = (CPs × Z) / F

As shown in Equation 2, after using the camera parameters how much space each point actually occupies, that is, the size (RPs) of the area occupied by one point in the real space is calculated, The point density Pd, which is a measure of whether the actual points are occupied by the octree cell area divided in the real space, is calculated.

Before calculating the point density according to the present invention, first, the problems of the prior art of determining the generation and decomposition of the octree using only the number of points will be briefly described with reference to FIGS. 3 and 4.

FIG. 3 is a diagram illustrating a change in size of an area occupied by one point in a real space according to distance, and FIG. 4 is a diagram illustrating a point array in a two-dimensional square area having one side length X at distances Z1 and Z2 of FIG. 3, respectively. Excerpted and illustrated. 3 illustrates the arrangement of points in a two-dimensional square region in which the length of one side at the distance Z1 is X. In the case of the right side, the points in the two-dimensional square region in which the length of one side at the distance Z2 is X. The figure illustrates an arrangement.

3 and 4, it can be seen that the size of an area occupied by one point in the actual space varies according to the distance. Therefore, when the decomposition of the octree cell is stopped using only the number of points, the octree cell may be generated for the object close to the camera, but the octree cell may not be generated for the object far from the camera.

Therefore, the present invention adopts a method of calculating the point density (Pd) in the octree cell in consideration of the size and number of areas occupied by one point in the real space at the same time. First, when the density (U _d ) of the unit point in the octree cell can be calculated as shown in Equation 3 below.

In Equation 3, the OCTs represent sizes of octree cells divided in real space. Since the actual octree is decomposed by dividing into eight equal parts from the cell of the initial cube, it is easy to know what the size of the current octree cell.

On the other hand, an object existing in a three-dimensional space actually exhibits a three-dimensional form, but an image obtained by a camera is a two-dimensional or 2.5-dimensional form rather than a three-dimensional form. Therefore, in the present invention, when determining the size of the octree cell, the octree cell is determined not as the entire volume of the octree cell but as one side of the octree cell, that is, in the form of a two-dimensional square. Therefore, in Equation 3, the denominator represents the size of an area occupied by one point in the real space, and the molecule represents the size of the octree cell. That is, the unit point density Ud of Equation 3 represents the ratio of each point corresponding to the size of the octree cell in which the point is included. When the unit point density Ud of Equation 3 is multiplied by the number N of points included in the octree cell, as shown in Equation 4, the actual point density indicating how full the points are in the octree cell is expressed as follows. Pd) is calculated.

Next, in the second step (S104) and comparing the point density (Pd) calculated in the first step with a predetermined density reference value, and the point density (Pd) calculated as a result of the comparison is a predetermined reference value or more, A step S105 of calculating an average and a variance value of the coordinate values of the points in the corresponding octree cell is performed. On the other hand, as a result of the comparison, if the calculated point density (Pd) is less than a predetermined reference value step (S108) to determine the decomposition to the next step of the corresponding octree cell is performed.

That is, in the second step, when the point density condition in the octree cell is satisfied, the average and variance values are calculated to determine how evenly distributed points are in the octree cell, and when the density condition is not satisfied, It operates to determine the decomposition to the next step.

Finally, in the third step, confirming the distribution of the points in the cell from the mean and the variance of the coordinate values of the points calculated in the second step (S106), and comparing the distribution of the points with a preset reference value. In operation S107, when the point distribution is greater than or equal to a predetermined reference value, the decomposition of the corresponding octree cell is terminated (S109). Meanwhile, as a result of the comparison, when the calculated point distribution is smaller than a predetermined reference value, the step S108 of determining the decomposition of the corresponding octree cell is performed.

As such, in the third step, even when the point density condition in the octree cell is satisfied, the method determines whether the points are evenly distributed in the corresponding octree cell, and if the points are not evenly distributed and are biased to one side, the octree is again distributed. Operate to disassemble the cell.

Meanwhile, the point density reference value in the octree cell in the second step and the point distribution reference value in the third step may be variously set according to the working environment. That is, by generating the octree cells of various sizes to express the environment according to the working characteristics of the robot, it is possible to achieve a precise and high speed robot operation.

5 shows the results of the octree generation when the multi-resolution octree-based three-dimensional object or environment representation method of the present invention is applied and when the existing octree generation method is used. As a camera, a structured light camera was used, and a Pentium-4 pc was used as a data processing means.

Fig. 5 (a) shows the original image of the object, and Fig. 5 (b) shows the same resolution of the octree decomposition with the octree size of 3.32 cm when the distance from the object is 80 cm, 110 cm, and 140 cm. c) shows the result of octree decomposition of the same resolution with an octree size of 1.66 cm for the same distance change, and FIG. 5 (d) shows the results of octree decomposition of multiple resolutions according to the present invention for the same distance change.

As shown in (b) and (c) of FIG. 5, it can be seen that the octree cell is generated differently according to the distance according to the existing octree generation method. On the other hand, according to the present invention it was confirmed that the octree cells are generated in the same form regardless of the distance change.

As described above, according to the multi-resolution octree-based three-dimensional object or environment representation method of the present invention, it is determined whether or not to decompose the octree cell in the next step based on the point density and the point distribution in the corresponding octree cell. By stopping the octree cell decomposition only when both the and distribution conditions are satisfied, it is possible to prevent unnecessary octree cell waste and increase the octree generation time, and to more accurately and quickly express the octree in multiple resolutions. .

In the above, the present invention has been shown and described with respect to certain preferred embodiments. However, the present invention is not limited to the above-described embodiments, and those skilled in the art to which the present invention pertains can vary as many without departing from the spirit of the technical idea of the present invention described in the claims below. Changes may be made.

1 is an operation flowchart showing a multi-resolution octree-based three-dimensional object or environment representation method according to an embodiment of the present invention,

2 is a diagram showing the relationship between the size of the area occupied by one point in the real space (RPs) and the camera parameters;

3 is a diagram illustrating a change in size of an area occupied by one point in a real space according to distance;

FIG. 4 is a diagram illustrating an excerpt of an array of points in a two-dimensional square region having a length X of one side at distances Z1 and Z2 of FIG. 3;

FIG. 5 is a view illustrating an octree generation result when a multi-resolution octree-based three-dimensional object or an environment representation method of the present invention is applied and an existing octree generation method is used. FIG.

<Explanation of symbols for the main parts of the drawings>

CPs: Point size on camera CCD F: Camera focal length

P1, P2: Point

Z, Z1, Z2: Distance from the camera to the point

RPs: the size of an area occupied by one point in real space

Claims (10)

In expressing a set of three-dimensional data points obtained from a camera or a three-dimensional measuring device as a volume-based octree, Calculate the size (RPs) of the area occupied by one point in the real space using the camera parameter, and include the calculated size of the area (RPs) with the octree cell size (OCTs) currently being divided and the octree cell. Calculating a point density Pd in the octree cell from the number of points N thus obtained; A second step of comparing the point density Pd calculated in the first step with a predetermined threshold value and calculating an average and a variance value of coordinates of points in the corresponding octree cell when the point density Pd is equal to or greater than a threshold value ; The point distribution in the octree cell is identified using the mean and the variance calculated in the second step. When the point distribution calculated in the second step is greater than or equal to a predetermined threshold, the decomposition of the octree cell is stopped to determine the point distribution. A third step of determining whether to disassemble, The size (RPS) of the area occupied by one point in the real space is Using the focal length (F) of the camera, one point size (CPs) on the camera CCD, and the distance (Z) between the camera and the point obtained from the 3D camera, RPs = CPs × Z / F Calculated according to The point density Pd in the octree cell is Using the size (RPs) of the area occupied by one point in the real space, the size of the octree cell (OCTs) currently being divided, and the number of points (N) included in the octree cell,

Multi-resolution octree-based three-dimensional object representation, characterized in that calculated according to the equation. The method of claim 1, A multi-resolution octree-based three-dimensional object representation method characterized by expressing an environment with an octree cell of various sizes suitable for the working environment by setting the point density reference value and the point distribution reference value which are the octree decomposition criteria according to the working environment. The method of claim 1, wherein the second step, Comparing the point density Pd calculated in the first step with a predetermined threshold value, and determining the decomposition of the octree cell to the next step when the point density Pd is smaller than the threshold value. Characterized by multi-resolution octree-based three-dimensional object representation. The method of claim 1, wherein the third step, If the point distribution calculated in the second step is smaller than the threshold value, multi-resolution octree-based three-dimensional object representation method characterized in that for determining the decomposition to the next step of the corresponding octree cell. delete In expressing a set of three-dimensional data points obtained from a camera or a three-dimensional measuring device as a volume-based octree, Calculate the size (RPs) of the area occupied by one point in the real space using the camera parameter, and include the calculated size of the area (RPs) with the octree cell size (OCTs) currently being divided and the octree cell. Calculating a point density Pd in the octree cell from the number of points N thus obtained; A second step of comparing the point density Pd calculated in the first step with a predetermined threshold value and calculating an average and a variance value of coordinates of points in the corresponding octree cell when the point density Pd is equal to or greater than a threshold value ; The point distribution in the octree cell is identified using the mean and the variance calculated in the second step. When the point distribution calculated in the second step is greater than or equal to a predetermined threshold, the decomposition of the octree cell is stopped to determine the point distribution. A third step of determining whether to disassemble, The size (RPS) of the area occupied by one point in the real space is Using the focal length (F) of the camera, one point size (CPs) on the camera CCD, and the distance (Z) between the camera and the point obtained from the 3D camera, RPs = CPs × Z / F Calculated according to The point density Pd in the octree cell is Using the size (RPs) of the area occupied by one point in the real space, the size of the octree cell (OCTs) currently being divided, and the number of points (N) included in the octree cell,

3D environment representation method based on multi-resolution octree, characterized in that calculated according to the equation. The method of claim 6, A multi-resolution octree-based three-dimensional environment representation method characterized by expressing the environment in various sizes of octree cells suitable for the work environment by setting the point density reference value and the point distribution reference value which are the octree decomposition criteria according to the work environment. delete The method of claim 6, wherein the second step, Comparing the point density Pd calculated in the first step with a predetermined threshold value, and determining the decomposition of the octree cell to the next step when the point density Pd is smaller than the threshold value. A multi-resolution octree-based three-dimensional environment representation method. The method of claim 6, wherein the third step, If the point distribution calculated in the second step is less than the threshold value, multi-resolution octree-based three-dimensional environment representation method characterized in that for determining the decomposition to the next step of the corresponding octree cell.

Images