Apparatus, method and program for generating point cloud data

Abstract

The purpose of this disclosure is to reduce work time for creating correct answer data and make it possible to create correct answer data even for an object that does not exist in a real environment.　Disclosed are a device and method in which: point cloud data for correction, which has been obtained by using a point cloud measurement device to measure an object for acquiring correction data that has a known shape, and the placement positions of the point cloud measurement device and the object for acquiring correction data at the time of the measurement of the point cloud data for correction are used to calculate correction values corresponding to the relative positions of the object for acquiring correction data and the point cloud measurement device; simulated object-under-measurement data is used to calculate ideal point cloud coordinates that would be obtained when measuring an object under measurement using the point cloud measurement device; and correct answer point cloud data simulating the object under measurement is generated through the correction of the ideal point cloud coordinates using correction values corresponding to the relative positions of the object under measurement and point cloud measurement device.

Description

[Technical field]

[0001] The present disclosure relates to a technique for generating point cloud data having coordinate information. [Background technology]

[0002] Point cloud data acquired using a Mobile Mapping System (MMS) or the like is utilized for high-precision maps and structure measurement. Point cloud data only has coordinate information, and it is necessary to extract objects to be measured, such as utility poles and steel towers, from the acquired point cloud data. Hereinafter, this act of extraction is called modeling. Various methods have been proposed for modeling technology, one of which is a method using machine learning. One method of modeling using machine learning is a method in which feature points are learned using point cloud data acquired from a measurement object as ground truth data (see, for example, Patent Document 1).

[0003] Conventionally, in order to create this correct answer data, the point cloud data must be visually checked and information indicating that the point cloud obtained from the object is an object must be manually specified, which is a problem that takes a very long time to complete.In addition, since the learning data must be point cloud data obtained from the actual environment, there is a problem that it is not possible to create learning data for objects that have not yet been used in the actual environment. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2019-3527 A Summary of the Invention [Problem to be solved by the invention]

[0005] The present disclosure aims to reduce the work time required to create correct answer data, and to make it possible to create correct answer data even for objects that do not exist in the real environment. [Means for solving the problem]

[0006] The apparatus and method of the present disclosure include: calculating a correction value according to a relative position between the object for correction data acquisition and the point cloud measuring instrument, using correction point cloud data obtained by measuring a correction data acquisition object having a known shape with the point cloud measuring instrument, and the installation positions of the point cloud measuring instrument and the object for correction data acquisition when the correction point cloud data was measured; Using simulated data of the measurement object, ideal point cloud coordinates that would be obtained when the measurement object is measured by the point cloud measuring device are calculated; The ideal point cloud coordinates are corrected with a correction value according to the relative positions of the measurement object and the point cloud measuring instrument, thereby generating correct point cloud data that simulates the measurement object.

[0007] Specifically, the program of the present disclosure is a program for realizing each functional unit of an apparatus related to the present disclosure, and is a program for causing a computer to execute each step of a communication method performed by an apparatus related to the present disclosure. Effect of the Invention

[0008] According to the present disclosure, it is possible to reduce the work time required to manually create correct answer data, and to make it possible to create correct answer data even for objects that do not exist in the real environment. [Brief description of the drawings]

[0009] [Figure 1] 1 illustrates an example system configuration of the present disclosure. [Diagram 2] 11 is a flow diagram showing an example of a method for generating correction point cloud data in a correction data generation function unit. FIG. [Diagram 3] An example of acquiring correction point cloud data using a point cloud measuring instrument is shown. [Figure 4] 13 shows an example of correction data. [Diagram 5] An example of acquiring correction point cloud data using a flat plate is shown. [Figure 6]11 is a flow chart showing an example of a method for generating correction data in a correction data generating function unit. FIG. [Figure 7] An example of the probability distribution of Δx is shown below. [Figure 8] 11 is a flow chart showing an example of a method for generating correction data in a correction data generating function unit. FIG. [Figure 9] 1 shows an example of an approximated surface representing a coordinate error. [Figure 10] FIG. 11 is a flow diagram showing an example of a method for generating corrected point cloud coordinates D in the correct data point cloud generating function unit. [Figure 11] FIG. 1 is an explanatory diagram for generating ideal point cloud coordinates C from 3D CAD data. [Figure 12] 13 is an explanatory diagram for generating corrected point cloud coordinates D from point cloud coordinates C. FIG. [Figure 13] An example of acquiring correction point cloud data using a mobile point cloud measuring device is shown. [Figure 14] An example of acquiring correction point cloud data using a mobile point cloud measuring device is shown. [Figure 15] FIG. 1 is an explanatory diagram for generating ideal point cloud coordinates C from 3D CAD data. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] Hereinafter, the embodiments of the present disclosure will be described in detail with reference to the drawings. Note that the present disclosure is not limited to the embodiments shown below. These implementation examples are merely illustrative, and the present disclosure can be implemented in various forms with various modifications and improvements based on the knowledge of those skilled in the art. Note that components with the same reference numerals in this specification and drawings are assumed to indicate the same components.

[0011] (Overall composition) 1 shows an example of a system configuration of the present disclosure. The system of the present disclosure includes a point cloud data generating device 91, a point cloud measuring instrument 92, a measurement target 93, and a correction data acquisition object 95. The point cloud data generating device 91 includes a corrective data point cloud generating function unit 11, a correction data generating function unit 12, and a storage unit 13. The device of the present disclosure can also be realized by a computer and a program, and the program can be recorded on a recording medium or provided via a network.

[0012] The present disclosure relates to A method for creating correct answer data to be used in machine learning that performs modeling based on measured point cloud data, comprising: Using correction point cloud data 22 obtained by measuring a correction data acquisition object 95 having a known shape with a point cloud measuring instrument 92, and the installation positions of the point cloud measuring instrument 92 and the correction data acquisition object 95 when the correction point cloud data 22 was measured, correction data 23 corresponding to the relative positions of the correction data acquisition object 95 and the point cloud measuring instrument 92 is calculated; Using 3D CAD data 21 simulating the measurement object 93, ideal point cloud coordinates C obtained when the measurement object 93 is measured by a point cloud measuring device 92 are calculated; A correction value corresponding to the relative position of the measurement object 93 and the point cloud measuring instrument 92 used when calculating the point cloud coordinate C is extracted from the correction data 23; The extracted correction values ​​are used to correct the ideal point cloud coordinates C of the measurement object 93, thereby calculating corrected point cloud coordinates D, and generating corrective point cloud data simulating the measurement object 93.

[0013] The storage unit 13 stores a point cloud measurement method 24 and a point cloud measuring instrument installation position 25 . The point cloud measurement method 24 is data for setting the point cloud measurement method (such as a laser emission method) 24 of the point cloud measuring instrument 92. The point cloud measurement method is, for example, a method by which the point cloud measuring instrument 92 acquired the point cloud coordinates A, and an example of the point cloud measurement method is the laser emission method. The point cloud measuring instrument installation position 25 is data for setting the installation position of the point cloud measuring instrument 92. The installation position includes the installation position coordinates where the point cloud measuring instrument 92 was installed when measuring the point cloud of the correction data acquisition object 95. The 3D CAD data 21 is simulated data that simulates a measurement object 93. The measurement object 93 is any object to be modeled, and examples of the object include a utility pole and a steel tower.

[0014] The correction data generation function unit 12 acquires a plurality of patterns of point clouds of a correction data acquisition object 95 using a point cloud measuring instrument 92, and acquires correction point cloud data 22. Then, the correction data generation function unit 12 creates correction data 23 from the correction point cloud data 22.

[0015] 3D CAD data 21 simulating a measurement object 93 is input to the supervised data point cloud generating function unit 11. Using the 3D CAD data 21, the supervised data point cloud generating function unit 11 calculates ideal point cloud coordinates C that are obtained when the measurement object 93 is measured by a point cloud measuring instrument 92. Then, the supervised data point cloud generating function unit 11 creates corrected point cloud coordinates D by adding correction data 23 to the obtained ideal point cloud coordinates C. Then, the supervised data point cloud generating function unit 11 outputs the corrected point cloud coordinates D as supervised point cloud data.

[0016] (First embodiment) In this embodiment, a case will be described in which the point cloud measuring instrument 92 is of a fixed type. Step S11: Point group coordinates A(x A ,y A ,z A ) to get the Step S12: The coordinates of the installation position of the correction data acquisition object 95 are acquired. Step S13: The diameter of the correction data acquisition object 95 is acquired. Step S14: The installation position coordinates (0,0) of the point cloud measuring instrument 92 are acquired from the point cloud measuring instrument installation position 25. Step S15: Using the installation position coordinates and diameter of the correction data acquisition object 95, the ideal point cloud coordinates B(x B ,y B ,z B ) is calculated. Step S16: Calculate the distance L from the point cloud coordinate B to the point cloud measuring instrument 92, and the angle θ at which the light from the point cloud measuring instrument 92 is reflected at the point cloud coordinate B. In this embodiment, an example of the angle θ will be described as an incident angle at which the light from the point cloud measuring instrument 92 is incident on the point cloud coordinate B. Step S17: Calculate the coordinate error (Δx, Δy, Δz) between the point cloud coordinates A and the point cloud coordinates B. Here, Δx=x B -x A ,Δy=y B -y A ,Δz=z B -z A It is. Step S18: The coordinate error Δ(Δx, Δy, Δz) between the point cloud coordinates A and B is associated with the angle θ and distance L of the point cloud coordinate B to generate correction data.

[0017] FIG. 3 shows an example of acquiring correction point cloud data using a point cloud measuring instrument 92. The object 95 for acquiring correction data is a perfect sphere. of Point Cloud Measuring Instrument 92 The correction data acquisition object 95 is measured using the above. At this time, the correction data acquisition object 95 and the fixed point of Group Instruments 92 Change the horizontal and vertical distance of a fixed of Point Cloud Measuring Instrument 92 Measures the correction-data-acquisition object 95 at distances L1, L2, and L3. This makes it possible to acquire point cloud coordinates A of a plurality of patterns with different distances L and angles θ to the correction-data-acquisition object 95.

[0018] Next, the installation position coordinates and the diameter of the sphere of the correction data acquisition object 95 are acquired in advance. The correction data generation function unit 12 calculates the ideally obtained point cloud coordinates B from the obtained point cloud coordinates A and the diameter of the sphere. Then, the correction data generation function unit 12 calculates the point cloud measurement instrument coordinates B from the point cloud coordinates B. 92 Calculate the distance L to the target point, and the angle θ.

[0019] Next, the difference between point cloud coordinate A and point cloud coordinate B is calculated as a coordinate error Δ for each (x, y, z) axis. This gives the coordinate error Δ(Δx, Δy, Δz). Finally, the distance and angle (L, θ) of the point cloud coordinates B are linked to the coordinate error Δ to generate correction data 23. An example of the correction data is shown in Fig. 4. Δx included in the correction data 23 is calculated for each combination of the distance L and the angle θ. The same applies to Δy and Δz.

[0020] Here, it is preferable that the material of the correction data acquisition object 95 is the same as that of the measurement object 93. The correction data acquisition object 95 may have the same shape as the measurement object 93, or may have a different shape. For example, the correction data acquisition object 95 may be a flat plate 95P. In the case of the flat plate 95P, it is preferable that not only the horizontal direction shown in FIG. 5(a) and the vertical direction shown in FIG. 5(b) are changed, but also the installation angles α and β with respect to the point cloud measuring instrument 92 are changed. This makes it possible to more accurately obtain the coordinate error Δ according to the angle θ.

[0021] (Correction data generation function unit 12) FIG. 6 is a flow diagram showing a first method for generating the correction data 23 in the correction data generating function unit 12. As shown in FIG. Step S21: Δx, Δy, and Δz for θ and L are obtained N times. This allows the coordinate error Δ for the angle θ and distance L to be measured N times (Δx 1 ~Δx N ,Δy 1 ~Δy N ,Δz 1 ~Δz N ) to get the Step S22: (Δx 1 ~Δx N ,Δy 1 ~Δy N ,Δz 1A probability distribution (frequency distribution) of the coordinate error Δ is created from Δx acquired N times (Δx = 1, Δy = 1, Δz = 1, ΔzN). For example, as shown in FIG. 7, a probability distribution of Δx is created for Δx acquired N times. Probability distributions are also created for Δy and Δz in the same way as for Δx. In this way, the probability distribution of the coordinate error Δ is generated. Then, Δx, Δy, and Δz are determined based on the created probability distribution. For example, the most probable values ​​in the probability distribution are determined as Δx, Δy, and Δz. Step S23: The angle θ and the distance L are associated with the coordinate error Δ and stored as correction data.

[0022] 8 is a flow diagram showing a second creation method of the correction data 23 in the correction data generation function unit 12. This creation method differs from the first creation method in that an approximation is performed using a value from one measurement for each (L, θ), and the correction becomes one value for a certain (L, θ). Step S31: The coordinate error Δ for the angle θ and the distance L is obtained. Step S32: Calculate an approximated surface that represents the coordinate error Δ for the angle θ and the distance L. An example of the approximated surface is shown in FIG. Step S33: The approximate surface equation for the angle θ and the distance L is stored as correction data.

[0023] Here, in step S32, approximation may be performed for Δx, Δy, and Δz according to θ and L to uniquely determine the values. For example, the closer θ is to 0, the smaller Δx, Δy, and Δz are, and the larger θ is, the larger Δx, Δy, and Δz are. Also, the closer L is to 0, the larger Δx, Δy, and Δz are, and when it reaches a certain value, Δx, Δy, and Δz become smaller, and when it exceeds that value, Δx, Δy, and Δz become larger. For this reason, it is desirable to decompose on the θ axis and approximate L with a quadratic or cubic function.

[0024] (Correct data point cloud generation function part 11) Next, the operation of the supervised data point cloud generating function unit 11 will be described with reference to Fig. 10 to Fig. 12. Fig. 10 is a flow diagram showing an example of a method for generating corrected point cloud coordinates D in the supervised data point cloud generating function unit 11. Step S41: The 3D CAD data 21 of the measurement object 93 and the installation position coordinates of the measurement object 93 are obtained. Step S42: The point cloud measuring instrument installation position 25 is acquired. Step S43: As the point cloud measurement method 24 of the point cloud measuring instrument 92, the laser emission method (laser emission angle) is acquired.

[0025] Step S44: Using the 3D CAD data 21 of the measurement object 93, calculate the point cloud coordinates C of the measurement object 93 that are ideally acquired when the measurement object 93 is measured with the point cloud measuring instrument 92. At this time, using the point cloud measuring instrument installation position 25 and the point cloud measurement method 24, the relative positions of the measurement object 93 and the point cloud measuring instrument 92 when the correction point cloud data 22 is acquired are included. Step S45: Using the installation position coordinates of the measurement target object 93, the installation position coordinates 25 of the point cloud measuring instrument 92, and the point cloud measurement method 24, the angle θ and the distance L at each point cloud coordinate C are calculated.

[0026] Step S46: The correction data 23 is obtained. Step S47: A coordinate error Δ(Δx, Δy, Δz) at the angle θ and the distance L is calculated from the correction data 23. For example, the coordinate error Δ(Δx, Δy, Δz) corresponding to (La, θa) is extracted from the correction data 23. Step S48: As shown in FIG. 12, a coordinate error Δ(Δx, Δy, Δz) is added to the point cloud coordinates C to calculate corrected point cloud coordinates D. Step S49: The corrected point cloud coordinates D are output to the outside as correct point cloud data.

[0027] In step S45, for example, if the point cloud measurement method is by emitting laser light, the distance and angle θ in the point cloud coordinate Ca is (La, θa) as shown in Fig. 11. In this case, the coordinate error Δ corresponding to (La, θa) is extracted from the correction data 23 (S47), and the extracted coordinate error Δ is added to the point cloud coordinate Ca. This makes it possible to calculate the corrected point cloud coordinate D in which the point cloud coordinate Ca is corrected.

[0028] Second embodiment In this embodiment, a case will be described in which the point cloud measuring instrument 92 is a mobile point cloud measuring instrument 92M such as an MMS. Figures 13 and 14 show an example of acquiring the correction point cloud data 22 using the mobile point cloud measuring instrument 92M. The correction data acquisition object 95 is a perfect sphere.

[0029] In this embodiment, in step S11 shown in Fig. 2, the movable point cloud measuring instrument 92M is used to measure the correction data acquisition object 95 in order to acquire the point cloud coordinates A. At this time, as shown in Fig. 13, the measurement (travel) direction of the movable point cloud measuring instrument 92M with respect to the correction data acquisition object 95 is set as a straight line D s , and curve Dc are obtained in two patterns.

[0030] Here, the measurement (travel) direction of the mobile point cloud measuring instrument 92M may be any number of patterns equal to or greater than two. For example, the curvature of the curve Dc may be further subdivided and acquired. Also, as shown in FIG. 13, the horizontal and vertical positions of the correction data acquisition object 95 may be changed to acquire two or more patterns of installation position coordinates of the correction data acquisition object 95, such as the correction data acquisition object 95-1 and 95-2. Also, the installation position coordinates of the correction data acquisition object 95 may be subdivided and acquired. Instead of changing the horizontal installation position coordinates of the correction data acquisition object 95, the traveling position of the mobile point cloud measuring instrument 92M may be changed.

[0031] In this embodiment as well, the installation position coordinates of the object 95 for acquiring correction data and the diameter of the sphere are acquired in advance (S12 to S14). In the subsequent step S15, the installation position coordinates of the object 95 for acquiring correction data and the diameter of the sphere are used to calculate the point cloud coordinates B that are ideally obtained in each measurement direction. For example, the point cloud coordinates B (x B ,y B ,z B ) to calculate

[0032] Furthermore, in step S16, the distance L and angle θ from point cloud coordinate B to the point cloud measuring instrument 92 are calculated for each measurement direction. Then, in step S17, the coordinate error Δ(Δx, Δy, Δz) between point cloud coordinate A and point cloud coordinate B is calculated for each measurement direction. Then, in step S18, the coordinate error Δ(Δx, Δy, Δz) is linked to the distance L and angle θ of point cloud coordinate B to create correction data 23. At this time, correction data 23 is created for each measurement direction.

[0033] Here, it is desirable that the correction-data-obtaining object 95 is made of the same material as the measurement target object 93. The correction-data-obtaining object 95 may be a flat plate 95P as shown in Figures 5(a) and 5(b).

[0034] (Correction data generation function unit 12) A first generation method of the correction data 23 in the correction data generation function unit 12 will be described. In this embodiment, steps S21 to S23 are executed as in the first embodiment. At this time, each step is executed for each measurement direction executed in step S11. As a result, correction data in which the probability distribution of the coordinate error Δ is linked to the distance L and the angle θ is generated for each measurement direction.

[0035] In this embodiment, the correction data generation function unit 12 may also generate the correction data using the second generation method shown in Fig. 8. This allows approximation using a value from one measurement for each (L, θ), and allows correction to be set to one value for a certain (L, θ).

[0036] (Correct data point cloud generation function part 11) Next, the operation of the correct data point cloud generation function unit 11 will be described with reference to Fig. 15. 3D CAD data 21 of the measurement target 93 is acquired (S41), a point cloud measuring instrument installation position 25 is acquired (S42), a point cloud measurement method 24 is acquired (S43), and a point cloud coordinate C ideally acquired from the 3D CAD data 21 is calculated (S44).

[0037] Furthermore, the distance L and angle θ in the point cloud coordinate C are calculated (S45) using the point cloud measuring instrument installation position 25 and the point cloud measurement method 24. For example, as shown in Fig. 15, the distance and angle (La, θa) in the point cloud coordinate Ca and the distance and angle (Lb, θb) in the point cloud coordinate Cb are calculated.

[0038] Next, using the correction data 23 (S46), the coordinate error Δ(Δx, Δy, Δz) at the point cloud coordinate Ca is calculated from (La, θa) (S47). Then, as shown in Fig. 12, the coordinate error Δ(Δx, Δy, Δz) corresponding to (La, θa) is added to the point cloud coordinate Ca to calculate the corrected point cloud coordinate Da at the point cloud coordinate Ca (S48). This process is performed for all of the point cloud coordinates C.

[0039] Here, in this embodiment, correction data in which the angle θ and distance L are linked to the probability distribution of the coordinate error Δ is generated for each measurement direction. Therefore, in this embodiment, the corrected point cloud coordinates D are calculated for each measurement direction. For example, if the measurement direction is a straight line Ds, the corrected coordinates of the straight line are used, and if the measurement direction is a curve Dc, the corrected coordinates of the curve are used. Finally, the corrected point cloud coordinates D are output to the outside for each measurement direction and stored as a corrected data point cloud (S49).

[0040] As described above, the supervised data generating device 91 of this embodiment includes the supervised data point cloud generating function unit 11, and therefore can automatically generate supervised point cloud data using the correction data 23. As a result, according to this embodiment, it is no longer necessary to manually generate supervised data from a point cloud, and the work time required for manually generating supervised data can be reduced.

[0041] Furthermore, the supervised data generating device 91 of this embodiment corrects the point cloud coordinates C using the coordinate error Δ according to the angle θ and the distance L, making it possible to generate simulation data close to actual data that takes into account the measurement error of the point cloud measuring instrument 92. This makes it possible for this embodiment to create supervised data even if the measurement target 93 does not exist in the actual facility. This makes it possible for this embodiment to reproduce the actual facility using 3D CAD, and to generate supervised data for machine learning by point cloud simulation that takes into account the measurement error of the point cloud measuring instrument 92. [Industrial Applicability]

[0042] The present disclosure can be applied to the information and communications industry. [Explanation of symbols]

[0043] 11: Correct data point cloud generation function 12: Correction data generation function unit 21: 3D CAD data 22: Point cloud data for correction 23: Correction data 24: Point Cloud Measurement Method 25: Point cloud measuring instrument installation location 91: Point cloud data generator 92: Point cloud measuring instrument 92F: Fixed point cloud measuring instrument 92M: Mobile point cloud measuring instrument 93: Measurement object 95, 95-1, 95-2, 95-3, 95P: Objects for acquiring correction data

Claims

1. calculating a correction value according to a relative position between the object for correction data acquisition and the point cloud measuring instrument, using correction point cloud data obtained by measuring a correction data acquisition object having a known shape with the point cloud measuring instrument, and the installation positions of the point cloud measuring instrument and the object for correction data acquisition when the correction point cloud data was measured; Using simulated data of the measurement object, ideal point cloud coordinates that would be obtained when the measurement object is measured by the point cloud measuring device are calculated; correcting the ideal point cloud coordinates with a correction value according to the relative positions of the measurement object and the point cloud measuring device, thereby generating correct point cloud data simulating the measurement object; Device.

2. measuring point cloud coordinates A by measuring the correction point cloud data obtained by measuring the correction data acquisition object with point cloud measuring instruments installed at a plurality of installation positions at different distances; Calculating ideal point cloud coordinates B when the object for obtaining correction data is measured using the installation position and shape of the object for obtaining correction data; Calculating a distance L from the point cloud coordinate B to the point cloud measuring instrument using an installation position of the point cloud measuring instrument; Calculating an angle θ at which light from the point cloud measuring instrument is reflected at the point cloud coordinate B using an installation position of the point cloud measuring instrument and a shape of the correction data acquisition object; A coordinate error Δ of the point cloud coordinate B is calculated by calculating a difference between the point cloud coordinate A and the point cloud coordinate B; The coordinate error Δ in the point cloud coordinate B is linked to the distance L and the angle θ in the point cloud coordinate B, and output as the correction value.

2. The apparatus of claim 1.

3. an approximated surface representing a coordinate error Δ with respect to the angle θ and the distance L is calculated, and the calculated approximated surface is output as the correction value; 3. The apparatus of claim 2.

4. The coordinate error Δ for the angle θ and the distance L is resolved by the angle θ, and the approximated surface is obtained by approximating the distance L with a quadratic function or a cubic function.

4. The apparatus of claim 3.

5. The correction data acquisition object is made of the same material as the measurement target.

5. Apparatus according to any one of claims 1 to 4.

6. calculating a correction value according to a relative position between the object for correction data acquisition and the point cloud measuring instrument, using correction point cloud data obtained by measuring a correction data acquisition object having a known shape with the point cloud measuring instrument, and the installation positions of the point cloud measuring instrument and the object for correction data acquisition when the correction point cloud data was measured; Using simulated data of the measurement object, ideal point cloud coordinates that would be obtained when the measurement object is measured by the point cloud measuring device are calculated; correcting the ideal point cloud coordinates with a correction value according to the relative positions of the measurement object and the point cloud measuring device, thereby generating correct point cloud data simulating the measurement object; method.

7. A program for causing a computer to realize each of the functional units of the device according to any one of claims 1 to 5.

Images