JP2017167671A - Information processing device, information processing method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To efficiently and properly remove noises and errors of point group data without receiving a restriction relating to the coordinates.SOLUTION: An information processing device includes: dividing means for dividing a group of three-dimensional coordinate points into a plurality of clusters; coordinate value calculation means for defining a local coordinate system with a coordinate axis determined based on the dispersion of each coordinate point included in the cluster and calculating a coordinate value in the local coordinate system of each coordinate point; revolution curved surface calculation means for calculating a revolution curved surface corresponding to the cluster based on the coordinate value calculated by the coordinate value calculation means; and correction means for correcting a component corresponding to the coordinate axis determined based on the dispersion of each coordinate point by the coordinate calculation means from among coordinate values in the local coordinate system of each coordinate point included in the cluster based on the revolution curved surface calculated by the revolution curved surface calculation means.SELECTED DRAWING: Figure 3

Description

  The present invention relates to a process for correcting point cloud data.

  Conventionally, a technique for removing noise and error coordinate points (hereinafter also simply referred to as noise and error) included in a set of three-dimensional coordinate points (hereinafter referred to as point cloud data) representing the shape of a subject. It has been known. In general, methods for acquiring point cloud data are classified into an active method in which measurement light is projected and reflected light from a subject is measured, and a passive method in which estimation is performed only from a plurality of images. The active method is affected by interference from ambient light and reflection characteristics of the measurement light of the subject, and noise and errors are generated. The passive method is affected by the gloss and shielding of the subject, and noise and errors are generated. Therefore, it is necessary to remove noise and errors after obtaining the point cloud data. Patent Document 1 proposes a method of smoothing point cloud data using a fourth-order difference quotient.

Japanese Patent No. 3610295

  However, in the method proposed in Patent Document 1, it is assumed that the point cloud data is arranged on a two-dimensional equidistant orthogonal lattice. Furthermore, in the method described in Patent Document 1, there may be a case where the optionality remains in the correction pattern (a quintic Bezier curved surface). For example, in the example shown in FIG. 4 of Patent Document 1, there are a plurality of correction patterns in which the smoothing reference value F of the entire point sequence is minimized. In that case, a uniquely optimal smoothing cannot be determined. Accordingly, an object of the present invention is to efficiently and appropriately remove noise and errors of point cloud data without being restricted by coordinates.

  An information processing apparatus according to the present invention is configured to divide a set of three-dimensional coordinate points representing a shape of a subject into a plurality of clusters, and to determine for each cluster based on a variance of each coordinate point included in the cluster. A local coordinate system having the coordinate axis as one coordinate axis, a coordinate value calculating means for calculating the coordinate value of each coordinate point in the local coordinate system, and a coordinate value calculated by the coordinate value calculating means. Based on the regression surface calculated by the regression surface calculation means and the regression surface calculated by the regression surface calculation means, out of the coordinate values in the local coordinate system of each coordinate point included in the cluster, the coordinate value calculation means Correction means for correcting a component corresponding to the coordinate axis determined based on the dispersion of the points.

  According to the present invention, noise and errors in point cloud data can be efficiently and appropriately removed without being restricted by coordinates.

The block diagram which shows the structure in the 1st Example of the information processing apparatus by this invention. The functional block diagram of the information processing apparatus by this invention. The flowchart which shows the flow of a process in 1st Example of the information processing apparatus by this invention. The figure which shows a mode that the coordinate point contained in a cluster is correct | amended. The figure which shows the example of the process result of the point group correction process of 1st Example. 10 is a flowchart showing a flow of processing in the information processing apparatus according to the second embodiment. 10 is a flowchart showing a flow of processing in the information processing apparatus according to the third embodiment.

  Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. The configurations shown in the following embodiments are merely examples, and the present invention is not limited to the illustrated configurations.

[Example 1]
In the first embodiment, an example in which efficient and accurate point cloud data correction is realized by performing local regression analysis will be described. First, the configuration of the information processing apparatus according to the first embodiment will be described.

  FIG. 1 is a diagram illustrating an example of the configuration of the information processing apparatus according to the first embodiment. An information processing apparatus 100 (hereinafter referred to as a processing apparatus 100) of the first embodiment includes a CPU 101, a RAM 102, a ROM 103, a secondary storage device 104, an input interface 105, and an output interface 106. Each component of the processing apparatus 100 is connected to each other by a system bus 107. In addition, the processing device 100 is connected to the external storage device 108 and the operation unit 110 via the input interface 105. The processing device 100 is connected to an external storage device 108 and a display device 109 via the output interface 106.

  The CPU 101 executes a program stored in the ROM 103 using the RAM 102 as a work memory, and comprehensively controls each component of the processing apparatus 100 via the system bus 107. Thereby, various processes described later are executed. The secondary storage device 104 is a storage device that stores various data handled by the processing device 100, and an HDD is used in this embodiment. The CPU 101 writes data to the secondary storage device 104 and reads data stored in the secondary storage device 104 via the system bus 107. In addition to the HDD, various storage devices such as an optical disk drive and a flash memory can be used for the secondary storage device 104.

  The input interface 105 is a serial bus interface such as USB or IEEE1394. The processing device 100 inputs data, instructions, and the like from an external device via the input interface 105. In this embodiment, the processing apparatus 100 acquires data from an external storage device 108 (for example, a storage medium such as a hard disk, a memory card, a CF card, an SD card, and a USB memory) via the input interface 105. In this embodiment, the processing apparatus 100 acquires a user instruction input to the operation unit 110 via the input interface 105. The operation unit 110 is an input device such as a mouse or a keyboard, and inputs a user instruction.

  The output interface 106 is a serial bus interface such as USB or IEEE 1394, like the input interface 105. The output interface 106 may be a video output terminal such as DVI or HDMI (registered trademark). The processing device 100 outputs data and the like to an external device via the output interface 106. In this embodiment, the processing device 100 outputs data (for example, image data) processed by the CPU 101 to the display device 109 (various image display devices such as a liquid crystal display) via the output interface 106. In addition, although the component of the processing apparatus 100 exists besides the above, since it is not the main point of this invention, description is abbreviate | omitted.

  Hereinafter, processing performed by the processing apparatus 100 according to the first embodiment will be described with reference to a functional block diagram shown in FIG. 2 and a flowchart shown in FIG. FIG. 2 is a functional block diagram of the processing apparatus according to the present invention. As illustrated in FIG. 2, the processing device 100 includes a data acquisition unit 201, a cluster determination unit (also referred to as a cluster division unit) 202, a coordinate conversion unit (also referred to as a coordinate value calculation unit) 203, a regression surface calculation unit 204, and a coordinate It has a function as the correction unit 205. When the CPU 101 reads and executes the control program stored in the ROM 103, the functions of the above units are realized. Note that the processing apparatus 100 may be configured to include a dedicated processing circuit corresponding to each unit. Hereinafter, the flow of processing performed by each unit will be described. FIG. 3 is a flowchart showing the flow of processing in the first embodiment of the processing apparatus according to the present invention.

  In step S <b> 301, the data acquisition unit 201 acquires processing-target point cloud data from an external device or the secondary storage device 104 via the input interface 105. Then, the data acquisition unit 201 outputs the acquired point cloud data to the cluster determination unit 202. The point cloud data is, for example, data indicating the three-dimensional shape of the subject, and includes a set of three-dimensional coordinates.

  In step S302, the cluster determination unit 202 divides the point cloud data into a plurality of clusters. Various known methods can be used for cluster division. As a general method, a nearest point search algorithm using a kd tree is used to search for k points closest to each point constituting a point group, and the searched points are accumulated to form a cluster. There is a method of forming. Duplication between clusters may be allowed, or may be formed exclusively. At this time, if the cluster determining unit 202 searches for the nearest neighbors collectively for all points of the point cloud data, it becomes difficult to form a completely exclusive cluster. Therefore, the cluster determination unit 202 may determine the clusters gradually as follows. As a first method, after determining one or a plurality of clusters, the nearest neighbor point search is performed again only for the point group that does not yet belong to any cluster, and a new cluster is obtained based on the result of the nearest neighbor point search. There is a method of iterating a series of processes of determining. As a second method, after determining one or more clusters, the nearest neighbor point search is performed again, and corrections such as excluding points included in the determined cluster are performed on the result of the nearest neighbor point search. There is a method of repeating a series of processes of determining a new cluster based on the result of the nearest neighbor search updated by the correction. The number of points included in each cluster may be the same or may be adaptively changed. When it fluctuates, the cluster determination unit 202 may determine the number of points constituting each cluster based on, for example, the distance between points and the statistics thereof.

  In step S303, the cluster determination unit 202 selects one cluster. The cluster selected here may include a point on which coordinate correction described later is performed. As an example of such a case, for example, a case where a plurality of clusters share the same point and one of the clusters has already performed coordinate correction. In addition, for example, there may be a case where coordinate correction is repeatedly performed a plurality of times on all or part of the point cloud data.

  In step S304, the coordinate conversion unit 203 defines a local coordinate system to be applied to only one selected cluster, and calculates coordinate values in the local coordinate system of all points included in the cluster. It is assumed that the input point group data includes the coordinate values of all points in the same global coordinate system. This global coordinate system is not necessarily the same as the local coordinate system. As an example of a method for determining the local coordinate system, for example, there is a method in which principal component analysis is performed on the three-dimensional coordinates of the points included in the cluster in the global coordinate system to determine the direction of the coordinate axis of the local coordinate system. In this method, a matrix of N rows and 3 columns combining three-dimensional coordinates in a global coordinate system of N points included in a cluster is A, and a 3 rows and 3 columns combining direction vectors of three coordinate axes of the local coordinate system. If the matrix is U, these can be related by equation (1).

  Here, C is a matrix of N rows and 3 columns obtained by repeatedly combining N three-dimensional row vectors representing cluster center coordinates. T is the transpose of the matrix. Σ is a 3-by-3 diagonal matrix representing eigenvalues. For C, the center of gravity of the cluster or the three-dimensional coordinates of one point included in the cluster can be used. The center of gravity is a three-dimensional vector obtained by calculating a weighted average of three-dimensional coordinates in a global coordinate system of N points. In Equation (1), a specific method for numerically calculating U from the left side to the right side is matrix singular value decomposition. Each column of the matrix U calculated in this way is a three-dimensional vector representing the directions of the three coordinate axes of the local coordinate system. This local coordinate system has a property that the plane formed by the two axes corresponding to the first and second columns of the matrix U coincides with the plane that best approximates the cluster. The closest approximation means that the variance value of the distance between the plane and each point is minimized. The coordinate conversion unit 203 defines the z axis in a direction perpendicular to the plane, that is, a normal direction. The z-axis is a coordinate axis that minimizes the dispersion value of the distance between each point of the point group included in the cluster among the coordinate axes of the local coordinate system and the plane formed by the other two axes. Note that the method of determining the direction of the coordinate axis of the local coordinate system is not limited to principal component analysis, and various known dimensionality reduction techniques such as Locally Linear Embedding and locality preserving projection can be applied.

  Thus, if the N-dimensional column vectors representing the x, y, and z coordinates in the local coordinate system of N points included in the cluster are X, Y, and Z, respectively, these are determined by Equation (2).

  Equation (2) means a set of coordinate values obtained by projecting N points included in the cluster onto each coordinate axis of the local coordinate system with the center of the cluster as the origin. Further, the left side of Equation (2) is a matrix obtained by combining N-dimensional column vectors in the column direction. Calculation of a regression surface, which will be described later, is performed based on coordinates X, Y, and Z on the local coordinate system. Hereinafter, calculating X, Y, and Z from A in this way is referred to as coordinate transformation. The reason why the coordinate conversion is necessary is to minimize the change in the density of the point group by moving each point in the normal direction in the coordinate correction described later. Note that the coordinate transformation in step S304 is not necessarily performed, and the subsequent processing may be performed directly on the three-dimensional coordinates in the global coordinate system, that is, the matrix A. However, there is a concern that the correction error tends to be larger than when coordinate conversion is performed, and the density of the point group changes more greatly.

In step S305, the regression surface calculation unit 204 calculates a regression surface based on the coordinates X, Y, and Z of the N points included in the cluster selected in step S303. An arbitrary function can be used as the regression surface, but hereinafter, a polynomial of P order (P = 1, 2, 3...) Is used. When the three-dimensional coordinate of one point in the local coordinate system is (x, y, z), the regression surface is expressed by equation (3) using the real constant a _ij .

In this embodiment, the regression surface is described as a quadratic polynomial expressed by Expression (4).

  As a method of calculating a regression surface of data arranged at unequal intervals, there is a least square estimation described in US Pat. No. 7,889,950. The regression surface calculation unit 204 determines the coordinates of each point after moving the N points included in the cluster onto the regression surface based on the function representing the regression surface. For example, when moving each point on the regression surface along the z-axis (normal direction), the x and y coordinates of each point may be substituted into a function representing the regression surface. Thereby, the z coordinate on the regression surface is derived. Such movement of each point on the regression surface is hereinafter referred to as correction. By performing the correction along the z-axis, the change in the density of the point group due to the correction can be minimized. However, the direction of correction is not necessarily limited to the z-axis.

  As an example of an efficient correction method, by performing the linear algebra calculation of Expression (5), it is possible to collectively correct the z coordinates of N points without iterative processing.

  Here, Z ′ is an N-dimensional column vector in which z coordinates of the N points after correction in the local coordinate system are arranged. In addition, the i-th row of the matrix E is defined as follows using Expression (26) described in Reference Document 1 below.

[Reference 1]
Hiroyuki Takeda, Student Member, IEEE, Sina Farsiu, Member, IEEE, and Peyman Milanfar, Senior Member, IEEE, “Kernel Regression for Image Processing and Reconstruction”, IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 16, NO. 2, FEBRUARY 2007 , p349-366
In Expression (6), e is an N-dimensional column vector in which the first element is 1 and all other elements are 0. When the regression surface is a quadratic polynomial shown in Expression (4), the matrix D is an N × 6 matrix defined by Expression (7).

Here, 1 in bold is an N-dimensional column vector with all elements being 1.・ Represents the product of each element. Further, the right side of the equation (7) corresponds to the powers of the variables x and y in each term of the equation (4). When the regression surface is a P-order polynomial, the matrix D is similarly defined by making the right side of Equation (7) correspond to Equation (3).

  The matrix W is a diagonal matrix determined from the function K of the distance between two points, and is defined by Expression (8).

Here, diag is an operator that generates a diagonal matrix having a vector of arguments as diagonal components. The function K is an arbitrary function. As the function K, a two-dimensional Gaussian function can be used.

  The method for obtaining Z ′ is not limited to this method. In step S305, it is not always necessary to derive Z ', and only the coefficient a in the equation (3) or the equation (4) may be obtained. In that case, the coordinate correcting unit 205 may calculate Z ′ in step S306.

  In step S306, the coordinate correction unit 205 corrects the positions (coordinate values) of N points included in the cluster based on the result of step S305. When Z ′ described above is calculated, an N × 3 matrix A ′ obtained by combining the three-dimensional coordinates of the corrected N points in the global coordinate system is determined by Expression (9).

Here, u _z is the third column of the matrix U and is a three-dimensional column vector representing a normal vector. (Z′−Z) U _z ^T represents a correction amount in the global coordinate system. FIG. 4 is a diagram schematically showing how the coordinate points included in the cluster are corrected by the processing in steps S304 to S306. A black coordinate point shown in FIG. 4A is a cluster selected by the cluster determination unit 202 and a cluster before correction. White coordinate points are coordinate points not included in the cluster. The gray coordinate point is the center of gravity of the cluster. A coordinate system composed of GxGyGz axes is a coordinate axis of the global coordinate system. A coordinate system composed of xyz axes is a local coordinate system corresponding to a cluster. FIG. 4B shows a scatter diagram obtained by projecting the coordinate points included in the cluster onto the xz plane. The broken line shown in FIG. 4B is a regression surface corresponding to the cluster calculated by the process of step S305. The arrows shown in FIG. 4B represent a state in which each coordinate point moves on the regression surface along the z axis (normal direction) by the process of S306.

  In step S307, the processing apparatus 100 determines whether processing has been completed for all points included in the point cloud data acquired by the data acquisition unit 201, that is, whether correction has been completed for all clusters. If there is a cluster that has not been corrected (NO in step S307), the processing device 100 returns to step S303. Then, a different cluster is selected by the cluster determination unit 202, and the processes in steps S303 to S306 are performed on the cluster. When the processing is completed for all the points included in the point cloud data (YES in step S307), the processing device 100 outputs the corrected point cloud data and ends the processing.

  The above is the processing performed by the processing apparatus 100 of the first embodiment. As described above, in this embodiment, a local coordinate system is defined for each cluster obtained by dividing the point cloud data. Then, the coordinate value in the local coordinate system of each coordinate point included in the cluster is corrected based on the regression surface. Therefore, according to the present embodiment, it is possible to remove the noise and error of the point cloud data without depending on the arrangement of the input point cloud data, that is, without being restricted by coordinates. In addition, since there is no restriction on coordinates, a polynomial surface of an arbitrary degree can be corrected.

  Further, in this embodiment, the point cloud data is collectively corrected by the equation (5). Therefore, according to the present embodiment, noise and errors can be efficiently removed from the point cloud data. In this embodiment, the matrix A ′ is uniquely determined for each cluster by the flow shown in FIG. 3. Therefore, according to the present embodiment, since the arbitrary point group data does not remain after correction, noise and errors can be appropriately removed.

  Also, in this embodiment, when defining a local coordinate system for a cluster, a direction (normal direction) perpendicular to the plane that minimizes the distance to each point included in the cluster is defined as the z-axis. . Then, each point is moved in the z-axis direction (normal direction) to correct the coordinate value of each point. Therefore, according to the present embodiment, the change in the density of the point group before and after the correction can be minimized. Therefore, for example, when the point cloud data is converted into a polygon model, it is possible to remove noise and errors in the point cloud data in advance using the method shown in this embodiment.

  In order to explain other effects of the present embodiment, the results of an experiment in which the processing shown in FIG. In the first experiment, the accuracy of correction was measured. At that time, a true point group was given by a function of Gz (Gx, Gy) = 0.2 Gx + sin (πGx / 3). Note that (Gx, Gy, Gz) represents the three-dimensional coordinates of an arbitrary point in the global coordinate system. The (Gx, Gy) coordinates of the point group consisting of 1681 points to be corrected are coordinate values randomly generated in the section [−10, 10]. Further, a Gz coordinate was generated by adding a Gaussian random number with a standard deviation of 0.2 as noise to the output value of the function Gz (Gx, Gy). In addition, the number of points of the cluster was fixed at 31 points, overlapping between clusters was allowed, and the regression surface was a quadratic polynomial. FIG. 5A is a scatter diagram obtained by projecting the point group to be corrected onto the Gx-Gz plane. In FIG. 5A, ◯ represents a true point group, and x represents a point group to which noise is applied. FIG. 5B shows the result of correction performed according to the flow shown in FIG. In FIG. 5B, ◯ represents a true point group, and Δ represents a corrected point group. The mean square error (Mean Square Error) of the point group of FIG. 5A and FIG. 5B with respect to the true point group was calculated to be 0.2475 and 0.0407, respectively. Thus, by correcting the point group by the method shown in the present embodiment, it is possible to accurately remove the noise component included in the point group. In addition, in order to remove a noise component more accurately, when the processing is completed for all the points included in the point cloud data (YES in step S307), the processing device 100 converts the corrected point cloud data again. You may make it input.

  In the second experiment, the time required for the correction process was measured under the same conditions as in the first experiment. However, Gz coordinates were generated by adding a Gaussian random number with a standard deviation of 0.4 as noise to the output value of the function Gz (Gx, Gy). When the point cloud data was collectively corrected by the method shown in this example, the time required for moving each coordinate point by curved surface regression was 105 seconds. Including the time required for the search for neighboring points of 375 seconds, the time required for the entire correction process was 480 seconds. As described above, according to the present embodiment in which the point cloud data is collectively corrected without the need for iterative processing, the time required for the correction processing can be shortened. On the other hand, when the point cloud data is corrected by a method that requires iterative processing to increase the accuracy of correction, for example, the method described in Reference Document 2 below, the time required for the search for neighboring points is also 375 seconds. However, the time required for the entire correction process was 3649 seconds.

[Reference 2]
Paola Valdivia, Douglas Cedrim, Fabiano Petronetto, Afonso Paiva, Luis Gustavo Nonato, "Normal Correction towards Smoothing Point-Based Surfaces" Conference on Graphics, Patterns and Images (SIBGRAPI), 2013

[Example 2]
In the second embodiment, an example will be described in which only the point cloud data in a specific range is corrected by a user's interactive instruction. Unless otherwise noted, the processing apparatus in the second embodiment shall perform the same processing as in the first embodiment. FIG. 6 is a flowchart showing the flow of processing in the processing apparatus 100 of the second embodiment.

  In step S501, the data acquisition unit 201 acquires point cloud data in the same manner as in step S301.

  In step S <b> 502, the cluster determination unit 202 designates a range of point cloud data to be processed based on a user instruction input to the processing apparatus 100 via the input interface 105. Prior to this processing, the cluster determination unit 202 may present the acquired point cloud data to the user via the display device 109. Then, the user may perform an operation for designating a range to be processed from the point cloud data displayed on the display device 109 on the operation unit 110. In addition, when the process of step S504 described later is performed, the process of step S502 is not necessarily required.

  In step S503, the cluster determination unit 202 divides the point cloud data into a plurality of clusters as in the process of step S302. At this time, the cluster determination unit 202 divides the cluster for the point cloud data to be processed, which is designated in the process of step S502. If the process in step S502 is not executed, the cluster is divided for the entire point cloud data acquired by the data acquisition unit 201.

  In step S504, the cluster determination unit 202 limits the clusters to be processed. Prior to this process, the cluster determination unit 202 may present the cluster information acquired by the process of step S503 to the user via the display device 109. Then, the user may perform an operation for designating a cluster to be processed from the clusters displayed on the display device 109 on the operation unit 110. Note that the cluster determination unit 202 does not necessarily need to execute the process of step S504 when the process of step S502 is performed.

  The processing in steps S505 to S509 is the same as the processing in steps S303 to S307 except that the processing is limited to the point cloud data or cluster range determined based on the user's instruction.

  The above is the process performed by the processing apparatus 100 of the second embodiment. According to the above processing, the processing target can be limited to a necessary portion based on the user's judgment, so that the point cloud data can be corrected more efficiently.

[Example 3]
In the third embodiment, an example will be described in which a point cloud data correction method is determined based on information for identifying acquired point cloud data. Unless otherwise noted, the processing apparatus in the third embodiment performs the same processing as in the first embodiment. FIG. 7 is a flowchart showing the flow of processing in the processing apparatus 100 of the third embodiment.

  In step S601, the data acquisition unit 201 acquires point cloud data in the same manner as in step S301. Furthermore, the data acquisition unit 201 acquires identification data associated with the point cloud from metadata included in the point cloud data. Alternatively, the data acquisition unit 201 generates identification data from the point cloud data. The identification data may be one value for the entire point group, or may be a different value for each subset of the point group. As an example, when the data acquisition unit 201 is point cloud data obtained from a human face, the data acquisition unit 201 selects the head, nose, mouth, chin from the point cloud indicated by the obtained point cloud data. -Identify parts such as ears and cheeks, and associate different identification data for each point group corresponding to each part. As another example, the data acquisition unit 201 associates different identification data for each point cloud data based on the metadata included in the point cloud data. Or the data acquisition part 201 identifies the to-be-photographed object shown with the said point cloud data from the acquired point cloud data, and matches the identification data produced | generated based on the identification result with the said point cloud data. Thereby, for example, different identification data is associated with the point cloud data depending on whether the acquired point cloud data is human point cloud data or vehicle point cloud data. The identification data may be in any form as long as it is information that can identify the shape or type of the whole or a part of the subject.

  In step S602, the cluster determination unit 202 determines a cluster in the same manner as in step S302.

In step S603, the cluster determination unit 202 assigns a parameter associated with the identification data to each cluster. Examples of parameters include, but are not limited to, the type of regression surface (for example, real constant a _ij ) and the degree of a polynomial. For example, the parameter may be information on the cluster such as the number of clusters or duplication, or information on the processing such as the number of iterations. The cluster determination unit 202 stores data indicating the correspondence between the identification data and the parameters in the secondary storage device 104 or the external storage device 108.

  The processes in steps S604 to S608 are the same as the processes in steps S303 to S307. In step S606, the regression surface calculation unit 204 reads parameters associated with each cluster from the secondary storage device 104 or the external storage device 108, and determines a regression surface based on the read parameters.

  The above is the processing performed by the processing apparatus 100 of the third embodiment. According to the above processing, correction can be performed using the optimum parameters for the point cloud data or each cluster, so that more accurate correction of the point cloud data can be performed.

  Needless to say, the embodiments described in the second and third examples may be combined.

[Example 4]
In the fourth embodiment, an example will be described in which the correction method of the point cloud data is adaptively determined based on the relationship between the regression surface and the coordinate values of the points included in the cluster. Unless otherwise noted, the processing apparatus 100 in the fourth embodiment performs the same processing as in the first embodiment.

  The processing apparatus 100 in the present embodiment performs the same processing as the processing apparatus in the first embodiment. However, in this embodiment, in the process of step S305, the regression surface calculation unit 204 calculates a regression surface and then compares the regression surface with the coordinate values of points included in the cluster. For example, the regression surface calculation unit 204 may evaluate the distance between the point and the regression surface for each point, or may evaluate the total distance between the point and the regression surface. Based on the evaluation result, the regression surface calculation unit 204 changes the parameters of the regression surface and calculates the regression surface again. In addition, the regression surface calculation unit 204 deletes points where the distance of the curved surface exceeds a predetermined threshold based on the evaluation result. The regression surface calculation unit 204 may perform other processing based on the evaluation result.

  According to the above processing, it is possible to robustly correct point cloud data against local shape non-uniformity of point cloud data and the presence of abnormal points.

  Needless to say, the embodiments described in the second, third, and fourth examples may be combined.

(Other examples)
The present invention supplies a program that realizes one or more functions of the above-described embodiments to a system or apparatus via a network or a storage medium, and one or more processors in the computer of the system or apparatus read and execute the program This process can be realized. It can also be realized by a circuit (for example, ASIC) that realizes one or more functions.

202 Cluster determination unit 203 Coordinate conversion unit 204 Regression surface calculation unit 205 Coordinate correction unit

Claims (12)

Dividing means for dividing a set of three-dimensional coordinate points representing the shape of the subject into a plurality of clusters;
For each cluster, define a local coordinate system with the coordinate axis determined based on the variance of each coordinate point included in the cluster as one coordinate axis, and calculate the coordinate value of each coordinate point in the local coordinate system A calculation means;
Regression surface calculation means for calculating a regression surface corresponding to the cluster based on the coordinate values calculated by the coordinate value calculation means;
Based on the regression surface calculated by the regression surface calculation means, of the coordinate values in the local coordinate system of each coordinate point included in the cluster, the coordinate value calculation means is determined based on the variance of each coordinate point An information processing apparatus comprising: correction means for correcting a component corresponding to the coordinate axis. The correction means further includes:
Based on coordinate values in the local coordinate system of each coordinate point included in the cluster, wherein the coordinate value calculation means determines a component corresponding to the coordinate axis determined based on the variance of each coordinate point. The information processing apparatus according to claim 1, wherein the coordinate value in the global coordinate system of each coordinate point is corrected. A set of coordinate points representing the shape of the subject whose coordinate values in the global coordinate system have been corrected by the correcting means is input to the dividing means,
The information processing apparatus according to claim 2, wherein processing by the dividing unit, the coordinate value calculating unit, the regression surface calculating unit, and the correcting unit is repeatedly executed. The coordinate value calculation means includes
When defining local coordinates for each of the plurality of clusters, a variance value of a distance between one coordinate axis of the local coordinate system and a plane formed by each coordinate point included in the cluster and the other two axes The information processing apparatus according to any one of claims 1 to 3, wherein the information processing apparatus is determined so as to be minimized. The clustering unit forms each cluster by accumulating a predetermined number of coordinate points having a short distance from each other in a set of coordinate points representing the shape of the subject. The information processing apparatus according to item. The information processing method according to any one of claims 1 to 5, wherein the regression surface calculation unit determines a parameter of the regression surface based on identification data that can identify at least the shape or type of the subject. apparatus. The regression surface calculation means calculates a regression surface again by changing parameters of the regression surface based on the relationship between the position of each coordinate point included in the cluster and the calculated regression surface corresponding to the cluster. The information processing apparatus according to any one of claims 1 to 6. The regression surface is a P-th order (P = 1, 2, 3...) Polynomial,
The information processing apparatus according to claim 6, wherein a parameter of the regression surface includes at least one of a real constant of the polynomial or an order of the polynomial. Of the set of coordinate points representing the shape of the subject, with respect to a set of coordinate points included in a range specified by the user, the dividing means, the coordinate value calculating means, the regression surface calculating means, and the correcting means, The information processing apparatus according to any one of claims 1 to 8, wherein the process according to claim 1 is executed. Among the plurality of clusters obtained by dividing the set of coordinate points representing the shape of the subject, the coordinate value calculation means, the regression surface calculation means, and the correction means for the cluster designated by the user The information processing apparatus according to any one of claims 1 to 9, wherein the process according to claim 1 is executed. A division step of dividing a set of three-dimensional coordinate points representing the shape of the subject into a plurality of clusters;
For each cluster, define a local coordinate system with the coordinate axis determined based on the variance of each coordinate point included in the cluster as one coordinate axis, and calculate the coordinate value of each coordinate point in the local coordinate system A calculation step;
A regression surface calculation step for calculating a regression surface corresponding to the cluster based on the coordinate values calculated in the coordinate value calculation step;
Based on the regression surface calculated in the regression surface calculation step, of the coordinate values in the local coordinate system of each coordinate point included in the cluster, corresponding to the coordinate axis determined based on the variance of each coordinate point And a correction step of correcting the component.   The program which functions a computer as each means of the information processing apparatus as described in any one of Claims 1-10.

Images