Method of analyzing cylinder center axis of cylinder three-dimensional point-cloud data

Abstract

To provide a method of analyzing a cylinder center axis of cylinder three-dimensional point-cloud data allowing for obtaining a cylinder center axis with accuracy.SOLUTION: A method of analyzing a cylinder center axis of cylinder three-dimensional point-cloud data is for a processor 31 to execute: determining step 42 that determines data species 51-53 making up cylinder three-dimensional point-cloud data 20 representative of a cylinder surface Wb; selecting step 44 that selects one algorithm suited for the data species 51-53 determined at the determining step 42 from among a plurality of algorithms 61-65 for analyzing a cylinder center axis L in the cylinder three-dimensional point-cloud data 20; and analyzing step 45 that analyzes a cylinder center axis L in the cylinder three-dimensional point-cloud data 20 using the selected algorithm.SELECTED DRAWING: Figure 18

Description

[Technical Field]

[0001] The present invention relates to a method for analyzing a cylinder's central axis in three-dimensional cylindrical point cloud data. [Background technology]

[0002] Patent Document 1 describes a method of capturing an image of a pipe as a subject, generating three-dimensional cylindrical point cloud data from the captured image, and estimating the central axis of the cylinder from the three-dimensional cylindrical point cloud data. It also describes analyzing the central axis of the cylinder by using principal component analysis. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent Publication No. 2021-189822 Summary of the Invention [Problem to be solved by the invention]

[0004] It is known that there are various algorithms other than those using principal component analysis for analyzing the cylinder central axis of cylindrical 3D point cloud data. However, it has been found that depending on the algorithm, it may not be possible to obtain a highly accurate cylinder central axis through analysis.

[0005] The present invention has been made in view of the above-mentioned problems, and aims to provide a method for analyzing the central axis of a cylinder from three-dimensional point cloud data of a cylinder, which is capable of obtaining a highly accurate central axis of a cylinder. [Means for solving the problem]

[0006] One aspect of the present invention is a method for processing a signal by a processor, the method comprising: a determination step of determining a data type constituting cylindrical three-dimensional point cloud data representing a cylindrical surface; a selection step of selecting one algorithm corresponding to the data type determined in the determination step from among a plurality of algorithms for analyzing the cylinder central axis in the cylindrical 3D point cloud data; and an analysis step of analyzing the cylinder central axis in the cylinder three-dimensional point cloud data using the selected algorithm. [Effects of the Invention]

[0007] According to the above aspect, it is assumed that there are multiple algorithms for analyzing the cylinder center axis in cylindrical 3D point cloud data. Under this assumption, an algorithm is selected from the multiple algorithms according to the type of data constituting the cylindrical 3D point cloud data. The selected algorithm is then used to analyze the cylinder center axis of the cylindrical 3D point cloud data. In other words, the algorithm used to analyze the cylinder center axis is an appropriate algorithm according to the type of data constituting the cylindrical 3D point cloud data. Therefore, the obtained cylinder center axis is highly accurate.

[0008] As described above, according to the above aspect, it is possible to provide a method for analyzing a cylinder's central axis from three-dimensional cylinder point cloud data, which is capable of obtaining a highly accurate cylinder central axis. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a diagram showing a shape measurement system for measuring the shape of a target workpiece. [Figure 2] FIG. 1 is a diagram showing a cylinder center axis analysis system for analyzing the center axis of a cylindrical surface of a target workpiece. [Figure 3] FIG. 10 is a diagram showing an example of cylindrical three-dimensional point cloud data. [Figure 4] FIG. 2 is a hardware configuration diagram of a cylinder central axis analysis device that constitutes the cylinder central axis analysis system. [Figure 5] FIG. 2 is a functional block diagram of a cylinder center axis analysis device. [Figure 6] FIG. 10 is a diagram showing acceptable data, first deviant data, and second deviant data. [Figure 7] 10A and 10B are diagrams for explaining a geometric determination method for determining acceptable data, first deviant data, and second deviant data. [Figure 8] 10A and 10B are diagrams for explaining a statistical determination method for determining acceptable data, first deviant data, and second deviant data. [Figure 9] 10 is a flowchart showing a first determination process by a determination unit in the cylinder central axis analyzing device. [Figure 10] 10 is a flowchart showing a second determination process by a determination unit in the cylinder central axis analyzing device. [Figure 11] 10 is a flowchart showing a third determination process by a determination unit in the cylinder central axis analyzing device. [Figure 12] FIG. 2 is a diagram showing a plurality of algorithms stored in an algorithm storage unit in the cylinder central axis analysis device. [Figure 13] FIG. 10 is a diagram illustrating an algorithm using the least squares method. [Figure 14] 10A and 10B are diagrams illustrating a method for determining the direction vector of the central axis of a cylinder in an algorithm using principal component analysis. [Figure 15] 10A and 10B are diagrams illustrating a first method for determining the central axis of a cylinder from a direction vector. [Figure 16] FIG. 10 is a diagram illustrating a second method for determining the central axis of a cylinder from a direction vector. [Figure 17] 10A and 10B are diagrams illustrating a method for determining a direction vector of a cylinder's central axis in an algorithm that uses a normal vector. [Figure 18] 10 is a flowchart showing a selection process by a selection unit in the cylinder central axis analysis device. DETAILED DESCRIPTION OF THE INVENTION

[0010] (Embodiment) 1. Target work W The target workpiece W for shape measurement will be described with reference to Fig. 1. The target workpiece W is, for example, a rotating body such as a gear or a bearing. The measurement target portion of the target workpiece W is the outer or inner surface of the rotating body, such as the tooth forming portion of a gear or the rolling surface of a bearing.

[0011] In FIG. 1, a gear is taken as an example of the target workpiece W. The target workpiece W has a gear tooth forming portion Wa and a cylindrical surface Wb. However, the target workpiece W may have the cylindrical surface Wb formed around the entire circumference, or may have a portion of the cylindrical surface Wb formed only on a portion of the circumference. The gear tooth forming portion Wa and the cylindrical surface Wb have a coaxial central axis L. Note that if the target workpiece W is a bearing, the target workpiece W has a rolling surface and a cylindrical surface.

[0012] 2. Shape measurement system 1 The shape measurement system 1 will be described with reference to Fig. 1. The shape measurement system 1 measures the shape of a measurement target portion of a target workpiece W. The shape measurement system 1 analyzes the shape of the measurement target portion based on three-dimensional point cloud data of the measurement target portion and the central axis L of the target workpiece W. In other words, the shape measurement system 1 analyzes the shape of the measurement target portion by identifying the position coordinates of the three-dimensional point cloud data using the central axis L of the target workpiece W as a reference. Therefore, it is important to know the position of the central axis L of the target workpiece W.

[0013] As shown in FIG. 1, the profile measurement system 1 acquires three-dimensional point cloud data of the tooth forming portion Wa, which is the measurement target portion of the target workpiece W placed on the stage 2, for example. The profile measurement system 1 acquires the three-dimensional point cloud data of the gear tooth forming portion Wa, for example, by a grid projection method. The profile measurement system 1 may acquire the three-dimensional point cloud data of the gear tooth forming portion Wa using a contact probe or a non-contact laser displacement meter. The profile measurement system 1 may also acquire the three-dimensional point cloud data of the gear tooth forming portion Wa based on one or more image data of the tooth forming portion Wa, which is the measurement target portion of the target workpiece W. Note that the profile measurement system 1 may acquire the three-dimensional point cloud data of the gear tooth forming portion Wa using any method other than the above method.

[0014] When the grating projection method is applied, the shape measurement system 1 includes a projection device 3, an imaging device 4, and a measurement processing device 5. The projection device 3 projects a grating pattern onto the tooth formation area Wa, which is the measurement target area. The imaging device 4 captures the diffuse reflection light of the grating pattern projected onto the tooth formation area Wa, which is the measurement target area, to obtain image data.

[0015] The measurement processing device 5 generates three-dimensional point cloud data of the tooth formation portion Wa, which is the measurement target portion, based on the image data acquired by the imaging device 4. Furthermore, the measurement processing device 5 stores position information of the central axis L of the target workpiece W. Then, the measurement processing device 5 determines the shape of the tooth formation portion Wa based on the three-dimensional point cloud data of the tooth formation portion Wa and the position information of the central axis L. In particular, the measurement processing device 5 determines the shape of the gear tooth flank.

[0016] In addition, when the target workpiece W is a bearing, the same applies as when it is a gear. In this case, too, the shape of the rolling surface of the bearing can be determined.

[0017] 3. Cylinder central axis analysis system 10 The configuration of the cylinder central axis analysis system 10 and a method for analyzing the cylinder central axis using the cylinder central axis analysis system 10 will be described with reference to Fig. 2. In Fig. 2, the same components as those in Fig. 1 are denoted by the same reference numerals, and description thereof will be omitted.

[0018] The cylinder central axis analysis system 10 analyzes the central axis L of the cylindrical surface Wb (hereinafter referred to as the "cylinder central axis L") using cylindrical three-dimensional point cloud data of the cylindrical surface Wb of the target workpiece W. In other words, the cylinder central axis analysis system 10 executes a method for analyzing the cylinder central axis L of the cylindrical three-dimensional point cloud data. Here, as described above, the cylinder central axis L is coaxial with the central axis L of the gear tooth forming portion Wa. In other words, the cylinder central axis analysis system 10 aims to obtain the central axis L of the tooth forming portion Wa by analyzing the position of the cylinder central axis L.

[0019] The cylinder central axis analysis system 10 acquires cylindrical three-dimensional point cloud data of the cylindrical surface Wb and determines the cylinder central axis L using the acquired cylindrical three-dimensional point cloud data. As with the shape measurement system 1, the cylinder central axis analysis system 10 can acquire the cylindrical three-dimensional point cloud data using any method, such as a grid projection method, a method using a contact probe, a method using a non-contact laser displacement meter, or a method using image data of the target workpiece W. However, FIG. 2 illustrates a case where the grid projection method is used as an example.

[0020] The cylinder central axis analysis system 10 acquires cylindrical three-dimensional point cloud data of the cylindrical surface Wb of the target workpiece W, for example, by a grid projection method. In this case, the cylinder central axis analysis system 10 includes a projection device 3, an imaging device 4, and a cylinder central axis analysis device 11. The projection device 3 projects a grid pattern onto the cylindrical surface Wb, which is the portion to be measured. The imaging device 4 captures the diffuse reflection light of the grid pattern projected onto the cylindrical surface Wb, which is the portion to be measured, to acquire image data.

[0021] The cylinder central axis analyzing device 11 generates cylindrical three-dimensional point cloud data of the cylindrical surface Wb, which is the measurement target portion, based on the image data acquired by the imaging device 4. Then, the cylinder central axis analyzing device 11 determines the cylinder central axis L of the cylindrical surface Wb based on the cylindrical three-dimensional point cloud data of the cylindrical surface Wb.

[0022] 4. Cylindrical 3D point cloud data 20 The cylindrical three-dimensional point cloud data 20 will be described with reference to Fig. 3. In Fig. 3, the cylindrical surface Wb is indicated by a two-dot chain line, and a plurality of point data are located on the cylindrical surface Wb. This plurality of point data constitutes the cylindrical three-dimensional point cloud data 20. In other words, the cylindrical three-dimensional point cloud data 20 represents the cylindrical surface Wb. However, in Fig. 3, the cylindrical three-dimensional point cloud data 20 indicates data relating to a portion of the circumferential direction of the cylindrical surface Wb.

[0023] 5. Hardware configuration of the cylinder central axis analysis device 11 The hardware configuration of the cylinder central axis analyzing device 11 will be described with reference to Fig. 4. The cylinder central axis analyzing device 11 is configured with a known computer. For example, the cylinder central axis analyzing device 11 is configured with a processor 31, a memory 32, a storage device 33, an input / output device 34, a communication interface 35, etc.

[0024] 6. Functional configuration of the cylinder central axis analysis device 11 The functional configuration of the cylinder central axis analyzing device 11 and the processing method by the cylinder central axis analyzing device 11 will be described with reference to Figs. 5 to 18. As shown in Fig. 5, the cylinder central axis analyzing device 11 is configured to include a point cloud data acquiring unit 41, a determining unit 42, an algorithm storing unit 43, a selecting unit 44, and an analyzing unit 45. The elements 41 to 43, 45 are configured by the processor 31 in Fig. 4. The element 43 is configured by the storage device 33 in Fig. 4. In Fig. 5, the steps shown in parentheses in each block indicate the steps executed by the processor 31 and the elements 41 to 42, 44 to 45.

[0025] 6-1. Point cloud data acquisition unit 41 The point cloud data acquisition unit 41 shown in FIG. 5 is configured to acquire cylindrical three-dimensional point cloud data 20 of the cylindrical surface Wb of the target workpiece W shown in FIG. 3. The point cloud data acquisition unit 41 may generate the cylindrical three-dimensional point cloud data 20 shown in FIG. 3 based on image data acquired by the imaging device 4 shown in FIG. 2. Alternatively, the point cloud data acquisition unit 41 may simply receive separately generated cylindrical three-dimensional point cloud data 20. As described above, the cylindrical three-dimensional point cloud data 20 is point cloud data representing the cylindrical surface Wb of the target workpiece W. The cylindrical three-dimensional point cloud data 20 may be data covering the entire circumferential direction of the cylindrical surface Wb, or may be data covering only a portion of the circumferential direction of the cylindrical surface Wb.

[0026] 6-2. Judgment section 42 The determination unit 42 shown in Fig. 5 will be described with reference to Fig. 6 to Fig. 11. The determination unit 42 executes a determination step (determination process) by the processor 31 to determine the data types constituting the cylindrical three-dimensional point cloud data 20 acquired by the point cloud data acquisition unit 41. As shown in Table 1, the data types include an allowable data type 51, a first deviant data type 52, and a second deviant data type 53.

[0027] [Table 1]

[0028] The allowable data type 51 is a data type that mainly includes allowable data D1. The first outlier data type 52 is a data type that mainly includes first outlier data D2 among data (outlier data) other than allowable data D1. The second outlier data type 53 is a data type that mainly includes second outlier data D3 among data (outlier data) other than allowable data D1.

[0029] Here, if high-precision measurements are possible, the cylindrical 3D point cloud data 20 may be composed only of data located on the ideal cylindrical surface Wb, but this is not always the case. In other words, the cylindrical 3D point cloud data 20 may include data located at positions away from the ideal cylindrical surface Wb. Therefore, data close to the ideal cylindrical surface Wb is defined as acceptable data D1, and data that is far from the ideal cylindrical surface Wb is defined as first deviant data D2 and second deviant data D3 depending on the degree of deviation.

[0030] The concepts of the allowable data D1, the first outlier data D2, and the second outlier data D3 will be explained with reference to Fig. 6. In Fig. 6, the allowable data D1 is indicated by a black circle, the first outlier data D2 is indicated by a white circle, and the second outlier data D3 is indicated by a white square.

[0031] The allowable data D1 is data included in the allowable definition region P that defines the cylindrical surface Wb. The allowable definition region P is a region sandwiched between thresholds Th1 and Th2 and centered on the ideal cylindrical surface Wb. The distance from the ideal cylindrical surface Wb to the thresholds Th1 and Th2 is a small value. In other words, the point data located within the allowable definition region P represents an almost ideal cylindrical surface Wb.

[0032] The first deviant data D2 is data that is outside the allowable definition region P and is located within a predetermined distance from the allowable definition region P. As shown in Fig. 6, the first deviant data D2 is data that is located outside the allowable definition region P and in a region from the thresholds Th1 and Th2 to the thresholds Th3 and Th4 that are a predetermined distance away. In other words, the first deviant data D2 is data that is located in a region between the thresholds Th1 and Th3 and a region between the thresholds Th2 and Th4.

[0033] The second deviant data D3 is data that is outside the allowable definition region P and is located further away than the predetermined distance. That is, the second deviant data D3 is data that is located further away than the thresholds Th3 and Th4.

[0034] Here, a specific example of the determination process for the allowable data D1, the first outlier data D2, and the second outlier data D3 will be described with reference to Figures 7 and 8. As the determination process for the allowable data D1, the first outlier data D2, and the second outlier data D3, the geometric determination method shown in Figure 7 or the statistical determination method shown in Figure 8 can be applied. However, the determination methods shown in Figures 7 and 8 are examples, and other methods can also be applied.

[0035] The geometric determination method shown in FIG. 7 determines the number of points inside the sphere centered on the point of interest by a threshold value. As shown in (a) of FIG. 7, when the radius of the sphere is r1 and the threshold value is n1, if the number of points inside the sphere is greater than or equal to the threshold value n1, it is regarded as acceptable data D1. On the other hand, as shown in (b) of FIG. 7, when the radius of the sphere is r1 and the threshold value is n1, if the number of points inside the sphere is less than the threshold value n1, it is not acceptable data D1, but is regarded as first outlier data D2 or second outlier data D2.

[0036] For the data recognized as outlier data in (b) of FIG. 7, the first outlier data D2 and the second outlier data D3 are determined by the following method. As shown in (c) of FIG. 7, when the radius of the sphere is r2 (> r1) and the threshold value is n2 (< n1), if the number of points inside the sphere is greater than or equal to the threshold value n2, it is regarded as first outlier data D2. This is because there may be acceptable data D1 in the vicinity of the first outlier data D2, and thus a large number of points are located in the vicinity.

[0037] On the other hand, as shown in (d) of FIG. 7, when the radius of the sphere is r2 and the threshold value is n2, if the number of points inside the sphere is less than the threshold value n2, it is not first outlier data D2, but is regarded as second outlier data D2.

[0038] The statistical determination method shown in FIG. 8 can apply, for example, a method using neighboring points. For example, the arithmetic mean value μi of the distances from k neighboring points is calculated for each point xi, and statistical determination is made using the calculated arithmetic mean value μi. First, as shown in (a) of FIG. 8, the distances d1 to dk between the point of interest xi and k neighboring points are calculated, and the arithmetic mean value μi of the distances d1 to dk is calculated. The arithmetic mean values μ1 to μn for all points of interest x1 to xn are calculated. Then, the overall arithmetic mean value μ ave and the standard deviation σ are calculated.

[0039] As shown in (b) of FIG. 8, constants m1 and m2 are set, and when the arithmetic mean value μi of the point of interest xi is such that "the overall arithmetic mean value μ ave+m1×σ”, it is set as the allowable data D1. ave +m1×σ" or more, and "overall arithmetic mean μ ave +m2×σ”, it is determined to be the first outlier data D2. ave +m2×σ”, the data is set as the second outlier data D3.

[0040] Next, the data type determination process by the determination unit 42 will be described with reference to Fig. 9 to Fig. 11. The determination process by the determination unit 42 is executed by the processor 31. Any one of the following three determination processes can be applied to the data type determination process.

[0041] 9, in the first determination process, first, it is determined (S1) whether or not first outlier data D2 exists in the cylindrical three-dimensional point cloud data 20. Then, it is determined (S2) whether or not second outlier data D3 exists in the cylindrical three-dimensional point cloud data 20. These processes apply the methods described with reference to FIGS. 6 to 8, etc.

[0042] Next, the determination unit 42 determines whether the number of first outlier data D2 is equal to or greater than a predetermined number Na (S3). The predetermined number Na can be set arbitrarily. If the number of first outlier data D2 is equal to or greater than the predetermined number Na (S3: Yes), the data type constituting the cylindrical 3D point cloud data 20 is determined to be the first outlier data type 52.

[0043] If the first outlier data D2 is less than the predetermined number Na (S3: No), the judgment unit 42 judges whether the second outlier data D3 is equal to or greater than the predetermined number Nb (S4). The predetermined number Nb can be set arbitrarily. If the second outlier data D3 is equal to or greater than the predetermined number Nb (S4: Yes), the data type constituting the cylindrical 3D point cloud data 20 is determined to be the second outlier data type 53. If the second outlier data D3 is less than the predetermined number Nb (S4: No), the data type constituting the cylindrical 3D point cloud data 20 is determined to be the allowable data type 51.

[0044] 10, in the second determination process, first, it is determined whether or not first outlier data D2 exists in the cylindrical three-dimensional point cloud data 20 (S11). Then, it is determined whether or not second outlier data D3 exists in the cylindrical three-dimensional point cloud data 20 (S12). These processes apply the methods described with reference to FIGS. 6 to 8, etc.

[0045] Next, the determination unit 42 determines whether the number of second outlier data D3 is equal to or greater than a predetermined number Nc (S13). The predetermined number Nc can be set arbitrarily. If the number of second outlier data D3 is equal to or greater than the predetermined number Nc (S13: Yes), the data type constituting the cylindrical 3D point cloud data 20 is determined to be the second outlier data type 53.

[0046] If the second outlier data D3 is less than the predetermined number Nc (S13: No), the determination unit 42 determines whether the first outlier data D2 is equal to or greater than the predetermined number Nd (S14). The predetermined number Nb can be set arbitrarily. If the first outlier data D2 is equal to or greater than the predetermined number Nd (S14: Yes), the data type constituting the cylindrical 3D point cloud data 20 is determined to be the first outlier data type 52. If the first outlier data D2 is less than the predetermined number Nd (S14: No), the data type constituting the cylindrical 3D point cloud data 20 is determined to be the allowable data type 51.

[0047] In the third determination process, as shown in Table 2, the data types include an allowable data type 51, a first deviant data type 52, a second deviant data type 53, and an abnormal data type 54.

[0048] [Table 2]

[0049] 11, in the third determination process, first, it is determined whether or not first outlier data D2 exists in the cylindrical three-dimensional point cloud data 20 (S21). Then, it is determined whether or not second outlier data D3 exists in the cylindrical three-dimensional point cloud data 20 (S22). These processes apply the methods described with reference to FIGS. 6 to 8, etc.

[0050] Next, the determination unit 42 determines whether the first outlier data D2 is equal to or greater than a predetermined number Ne and whether the second outlier data D3 is equal to or greater than a predetermined number Nf (S23). The predetermined numbers Ne and Nf can be set arbitrarily. If the first outlier data D2 is equal to or greater than the predetermined number Ne and the second outlier data D3 is equal to or greater than the predetermined number Nf (S23: Yes), the data type constituting the cylindrical 3D point cloud data 20 is determined to be the abnormal data type 54.

[0051] Next, the determination unit 42 determines whether the number of first outlier data D2 is equal to or greater than a predetermined number Ng (S24). The predetermined number Ng can be set arbitrarily. If the number of first outlier data D2 is equal to or greater than the predetermined number Ng (S24: Yes), the data type constituting the cylindrical 3D point cloud data 20 is determined to be the first outlier data type 52.

[0052] If the first outlier data D2 is less than the predetermined number Ng (S24: No), the judgment unit 42 judges whether the second outlier data D3 is equal to or greater than the predetermined number Nh (S25). The predetermined number Nh can be set arbitrarily. If the second outlier data D3 is equal to or greater than the predetermined number Nh (S25: Yes), the data type constituting the cylindrical 3D point cloud data 20 is determined to be the second outlier data type 53. If the second outlier data D3 is less than the predetermined number Nh (S25: No), the data type constituting the cylindrical 3D point cloud data 20 is determined to be the allowable data type 51.

[0053] In the third determination process, in steps S24 and S25, a determination is first made on the first missing data D2, followed by a determination on the second missing data D3, as in the first determination process. Alternatively, a determination may be first made on the second missing data D3, followed by a determination on the first missing data D2, as in the second determination process.

[0054] 6-3. Algorithm storage unit 43 The algorithm storage unit 43 shown in FIG. 5 stores a plurality of algorithms 61 to 65 for analyzing the cylinder central axis L in the cylinder three-dimensional point cloud data 20, as shown in FIG.

[0055] The algorithm memory unit 43 stores, for example, an algorithm 61 using the least squares method, a first algorithm 62 using principal component analysis, a second algorithm 63 using principal component analysis, a first algorithm 64 using normal vectors, and a second algorithm 65 using normal vectors.

[0056] The algorithm 61 using the least squares method will be described with reference to Fig. 13. The algorithm 61 using the least squares method generates an approximate cylindrical shape by three-dimensional approximation using a cylindrical shape for the cylindrical three-dimensional point cloud data 20. Alternatively, the algorithm 61 using the least squares method generates an approximate arc shape by two-dimensional approximation using an arc for each Z-axis coordinate (shown in Figs. 2 and 3) for the cylindrical three-dimensional point cloud data 20. Next, the algorithm 61 using the least squares method determines the central axis of the generated approximate cylindrical shape or approximate arc shape as the cylindrical central axis L of the cylindrical three-dimensional point cloud data 20.

[0057] The first algorithm 62 using principal component analysis will be described with reference to Fig. 14 and Fig. 15. As shown in Fig. 14, the first algorithm 62 using principal component analysis performs principal component analysis on the cylindrical 3D point cloud data 20 to determine one of the first principal component L1, the second principal component L2, and the third principal component L3 as the direction vector VL of the cylinder central axis L. In Fig. 14, the first principal component L1 is illustrated as the direction vector VL of the cylinder central axis L.

[0058] 15, a first algorithm 62 using principal component analysis determines, among the lines having a direction vector VL, a line whose distance from each point of the cylindrical three-dimensional point cloud data 20 is close to a certain value as the cylinder's central axis L. For example, an optimization calculation can be used. For example, the point (position) through which the cylinder's central axis L passes is used as a variable, and the standard deviation of the distance between each point and the cylinder's central axis L is used as an objective function, and the position of the cylinder's central axis L when the standard deviation, which is the objective function, is minimized is determined.

[0059] The second algorithm 63 using principal component analysis will be described with reference to Fig. 14 and Fig. 16. As shown in Fig. 14, the second algorithm 63 using principal component analysis determines the direction vector VL of the cylinder central axis L, similar to the first algorithm 62 using principal component analysis.

[0060] 16, the central axis of the approximate arc shape of the point group 21 located on a plane 22 obtained by projecting each point of the cylindrical three-dimensional point group data 20 in the direction of the direction vector VL is determined as the cylinder central axis L. In other words, the axis that passes through the center point of the approximate arc shape and has the direction vector VL is determined as the cylinder central axis L.

[0061] The first algorithm 64 using normal vectors will be described with reference to FIGS. 15 and 17. As shown in FIG. 17, the first algorithm 64 using normal vectors generates, for each point in the cylindrical 3D point cloud data 20, a mesh region formed by multiple neighboring points in the cylindrical 3D point cloud data 20 and a normal vector 23 at the point included in the mesh region. Here, the mesh region can be, for example, a region formed by three or more points positioned so as to surround a point of interest. The mesh region is assumed to be a plane. The normal vector 23 is assumed to be a vector that has a normal direction to the mesh region and passes through the point of interest.

[0062] Next, a first algorithm 64 using normal vectors determines a direction vector orthogonal to the normal vector 23 at each point of the cylindrical three-dimensional point cloud data 20 as the direction vector VL of the cylinder central axis.

[0063] 15, a first algorithm 64 using a normal vector determines, among the lines having a direction vector VL, a line whose distance from each point of the cylindrical 3D point cloud data 20 is close to a certain value as the cylinder central axis L. This is the same processing as the first algorithm 62 using the principal component analysis described above.

[0064] The second algorithm 65 using normal vectors will be described with reference to Fig. 16 and Fig. 17. As shown in Fig. 17, the second algorithm 65 using normal vectors determines the direction vector VL of the cylinder central axis L, similar to the first algorithm 64 using normal vectors. Subsequently, as shown in Fig. 16, the second algorithm 65 using normal vectors determines the cylinder central axis L, similar to the second algorithm 63 using principal component analysis.

[0065] 6-4.Selection section 44 The selection process (selection step) by the selection unit 44 shown in Fig. 5 will be described with reference to Fig. 18. The selection unit 44 executes the selection process (selection processing) by the processor 31 to select one algorithm from among a plurality of algorithms 61 to 65 for analyzing the cylinder central axis L in the cylindrical three-dimensional point cloud data 20. In particular, the selection unit 44 selects one algorithm from among the plurality of algorithms 61 to 65 according to the data types 51 to 53 determined in the determination process (determination step) by the determination unit 42.

[0066] Table 3 shows a number of performance indices for the data types 51 to 53 and the algorithms 61 to 65.

[0067] [Table 3]

[0068] As shown in Table 3, algorithm 61 using the least squares method is extremely good in terms of processing time, but has low tolerance for small and large outliers. First algorithm 62 and second algorithm 63 using principal component analysis are good in terms of processing time, have extremely high tolerance for small outliers, and slightly low tolerance for large outliers. First algorithm 64 and second algorithm 65 using normal vectors are not so good in terms of processing time, have slightly low tolerance for small outliers, and are extremely good tolerance for large outliers. Taking into account the relationships in Table 3, the selection unit 44 selects algorithms 61 to 65 according to the data types 51 to 53.

[0069] 18, the selection unit 44 determines whether the target cylindrical 3D point cloud data 20 is configured by the allowable data type 51 (S31). If it is determined that the target cylindrical 3D point cloud data 20 is configured by the allowable data type 51 (S31: Yes), the selection unit 44 selects an algorithm for the allowable data type associated with the allowable data type 51 from among a plurality of algorithms 61 to 65 (S32). For example, the selection unit 44 selects the algorithm 61 that uses the least squares method as the algorithm for the allowable data type (S32).

[0070] Next, if the data type is not an allowable data type 51 (S31: No), the selection unit 44 determines whether the target cylindrical 3D point cloud data 20 is composed of the first outlier data type 52 (S33). If it is determined that the data is composed of the first outlier data type 52 (S33: Yes), the selection unit 44 selects an algorithm for the first outlier data type associated with the first outlier data type 52 from among the multiple algorithms 61 to 65 (S34). For example, the selection unit 44 selects a first algorithm 62 or a second algorithm 63 that uses principal component analysis as the algorithm for the first outlier data type (S34). Either the first algorithm 62 or the second algorithm 63 may be selected.

[0071] Next, if the target cylindrical 3D point cloud data 20 is not the first outlier data type 52 (S33: No), the selection unit 44 determines whether the target cylindrical 3D point cloud data 20 is composed of the second outlier data type 53 (S35). If it is determined that the target cylindrical 3D point cloud data 20 is composed of the second outlier data type 53 (S35: Yes), the selection unit 44 selects an algorithm for the second outlier data type associated with the second outlier data type 53 from among the multiple algorithms 61 to 65 (S36). For example, the selection unit 44 selects the first algorithm 64 or the second algorithm 65 that uses a normal vector as the algorithm for the second outlier data type (S36). Either the first algorithm 64 or the second algorithm 65 may be selected.

[0072] If the data type is not the second outlier data type 53 (S35: No), the selection unit 44 does not select an algorithm. This case corresponds to the case where the determination unit 42 determines that the data type is an abnormal data type in the second determination process shown in FIG.

[0073] 6-5.Analysis Department 45 5 executes an analysis step (analysis process) in which the processor 31 uses the algorithm selected by the selection unit 44 to analyze the cylinder central axis L in the cylinder three-dimensional point cloud data 20. Here, the analysis unit 45 is used to determine the central axis of the three-dimensional point cloud data of the gear tooth forming portion Wa or the rolling surface of the bearing of the target workpiece W. In other words, the analysis unit 45 determines the cylinder central axis L determined using the cylinder three-dimensional point cloud data 20 as the central axis of the gear or the rolling surface.

[0074] For example, when the cylindrical 3D point cloud data 20 is composed of the allowable data type 51, the analysis unit 45 analyzes the cylinder center axis L in the cylindrical 3D point cloud data 20 using an algorithm 61 that uses the least squares method. When the cylindrical 3D point cloud data 20 is composed of the first outlier data type 52, the analysis unit 45 analyzes the cylinder center axis L using a first algorithm 62 or a second algorithm 63 that uses principal component analysis. When the cylindrical 3D point cloud data 20 is composed of the second outlier data type 53, the analysis unit 45 analyzes the cylinder center axis L using a first algorithm 64 or a second algorithm 65 that uses a normal vector.

[0075] 7.Effects In this embodiment, it is assumed that there are a plurality of algorithms 61 to 65 for analyzing the cylinder central axis L in the cylindrical 3D point cloud data 20. On this assumption, an algorithm is selected from the plurality of algorithms 61 to 65 according to the data types 51 to 53 that constitute the cylindrical 3D point cloud data 20. Then, the selected algorithm is used to analyze the cylinder central axis L of the cylindrical 3D point cloud data 20. In other words, the algorithm used to analyze the cylinder central axis L is an appropriate algorithm according to the data type that constitutes the cylindrical 3D point cloud data 20. Therefore, the obtained cylinder central axis L is highly accurate.

[0076] In particular, an appropriate algorithm is selected taking into consideration the performance shown in Table 3. Therefore, depending on the data type, it is possible to obtain the cylinder central axis L with high accuracy, and among these, an algorithm with a short processing time can be applied.

[0077] Furthermore, as can be seen from Table 3, if there is almost no difference in accuracy between the algorithms 62 and 63 using principal component analysis and the algorithms 64 and 65 using normal vectors, it is advisable to set a threshold value that gives priority to the algorithms 62 and 63 using principal component analysis, thereby shortening the processing time. [Explanation of symbols]

[0078] 10 Cylinder central axis analysis system 31 processors 42 Judgment section (judgment process) 43 Algorithm memory section 44 Selection section (selection process) 45 Analysis Department (Analysis process) 51~54 Data type 61~65 Algorithm Wb Cylindrical surface L Cylinder center axis P Allowable definition area

Claims

1. The processor a determination step of determining a data type constituting cylindrical three-dimensional point cloud data representing a cylindrical surface; a selection step of selecting one algorithm corresponding to the data type determined in the determination step from among a plurality of algorithms for analyzing the cylinder central axis in the cylindrical 3D point cloud data; an analysis step of analyzing the cylinder central axis in the cylinder three-dimensional point cloud data using the selected algorithm.

2. The data types constituting the cylindrical three-dimensional point cloud data are: an allowable data type including allowable data included in an allowable definition region that defines the cylindrical surface, and having less than a predetermined number of outliers outside the allowable definition region; a deviant data type including a predetermined number or more of the deviant data; The selection step includes: If it is determined that the cylindrical 3D point cloud data is configured by the permissible data type, an algorithm for the permissible data type associated with the permissible data type is selected; 2. A method for analyzing a cylindrical center axis of cylindrical three-dimensional point cloud data according to claim 1, wherein, when it is determined that the cylindrical three-dimensional point cloud data is composed of the outlier data type, an algorithm for the outlier data type corresponding to the outlier data type is selected.

3. the algorithm for the allowable data type is an algorithm using the least squares method, The method for analyzing a cylinder's central axis of cylindrical three-dimensional point cloud data according to claim 2 , wherein the algorithm for the outlier data type is an algorithm using principal component analysis or an algorithm using normal vectors.

4. The outlier data type is a first deviant data type including a predetermined number or more of first deviant data, the first deviant data being the deviant data and located within a predetermined distance from the allowable definition region; a second deviant data type including a predetermined number or more of second deviant data, the second deviant data being the deviant data and located away from the allowable definition region by more than the predetermined distance; The selection step includes: If it is determined that the cylindrical three-dimensional point cloud data is composed of the first outlier data type, selecting an algorithm for the first outlier data type associated with the first outlier data type; 3. The method for analyzing a cylindrical center axis of cylindrical three-dimensional point cloud data according to claim 2, wherein, when it is determined that the cylindrical three-dimensional point cloud data is composed of the second outlier data type, an algorithm for the second outlier data type associated with the second outlier data type is selected.

5. the algorithm for the allowable data type is an algorithm using the least squares method, the first outlier data type algorithm is an algorithm using principal component analysis; The method for analyzing a cylinder central axis of three-dimensional cylindrical point cloud data according to claim 4 , wherein the algorithm for the second outlier data type is an algorithm that uses a normal vector.

6. The algorithm using the least squares method is An approximate cylindrical shape is generated by three-dimensional approximation using a cylindrical shape, or an approximate arc shape is generated by two-dimensional approximation using a circular arc, The method for analyzing a cylinder central axis of cylindrical three-dimensional point cloud data according to claim 3 , wherein a central axis of the generated approximate cylindrical shape or the generated approximate circular arc shape is determined as the cylinder central axis.

7. The algorithm using the principal component analysis is performing a principal component analysis of the cylindrical three-dimensional point cloud data to determine one of a first principal component, a second principal component, and a third principal component as a direction vector of the cylinder central axis; 6. The method for analyzing the cylinder central axis of cylindrical 3D point cloud data according to claim 5, wherein, among the straight lines having the direction vector, a straight line whose distance from each point of the cylindrical 3D point cloud data is close to a constant value is determined as the cylinder central axis, or the central axis of an approximate arc shape of a point cloud on a plane obtained by projecting each point of the cylindrical 3D point cloud data in the direction of the direction vector is determined as the cylinder central axis.

8. The algorithm using the normal vector is generating a mesh region formed by a plurality of neighboring points in the cylindrical three-dimensional point cloud data and a normal vector at a point included in the mesh region for each point in the cylindrical three-dimensional point cloud data; determining a direction vector orthogonal to the normal vector at each point of the cylindrical three-dimensional point cloud data as a direction vector of the cylinder central axis; 6. The method for analyzing the cylinder central axis of cylindrical 3D point cloud data according to claim 5, wherein, among the straight lines having the direction vector, a straight line whose distance from each point of the cylindrical 3D point cloud data is close to a constant value is determined as the cylinder central axis, or the central axis of an approximate arc shape of a point cloud on a plane obtained by projecting each point of the cylindrical 3D point cloud data in the direction of the direction vector is determined as the cylinder central axis.

9. The analyzing step In a target workpiece having a gear tooth forming portion or a rolling surface of a bearing and the cylindrical surface, the target workpiece is used to determine a central axis of three-dimensional point cloud data of the gear tooth forming portion or the rolling surface, The method for analyzing a cylinder's central axis of a cylinder's three-dimensional point cloud data according to any one of claims 1 to 8, wherein the cylinder's central axis determined using the cylinder's three-dimensional point cloud data is determined as the central axis of the tooth forming portion or the central axis of the rolling surface.

Images