Data processing device, data processing method, and program

Abstract

Efficiently detect targets. [Solution] A data processing device 300 for extracting a point cloud relating to a target of a specific shape from a point cloud obtained by laser scanning, comprising a calculation unit 302 that acquires a point cloud obtained by laser scanning with the target as the scan range, extracts a point cloud from the point cloud that fits the visual shape of the specific shape, and extracts a point cloud that fits the specific shape from the extracted point cloud.

Description

Technical Field

[0001] This disclosure relates to laser scanning technology.

Background Art

[0002] Patent Document 1 describes a technique for detecting a target for surveying by laser scanning.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] Efficiently detect the target.

Means for Solving the Problems

[0005] This disclosure is a data processing device that extracts a point group related to a target of a specific shape from a point group obtained by laser scanning, and acquires a point group obtained by laser scanning with the target as a scanning range, extracts a point group that fits the visual shape of the specific shape from the point group, and is provided with an arithmetic unit that extracts a point group that fits the specific shape for the extracted point group. This disclosure can also be grasped as an invention of a method and an invention of a program.

Effects of the Invention

[0006] The target can be efficiently detected.

Brief Description of the Drawings

[0007] [Figure 1] It is a conceptual diagram of an embodiment. [Figure 2]This is a functional block diagram of a data processing device. [Figure 3] This is a flowchart showing an example of the processing procedure. [Figure 4] This is a flowchart showing an example of the processing procedure. [Figure 5] This is a flowchart showing an example of the processing procedure. [Modes for carrying out the invention]

[0008] (1) First Embodiment Figure 1 shows a laser scanner 100, spherical target 200, and spherical target 201. Spherical targets 200 and 201 are surveying targets with a spherical shape. By detecting spherical targets 200 and 201 with a laser scan by the laser scanner 100 and calculating the center position of their spherical shape, the precise location of spherical targets 200 and 201 is detected. Although two spherical targets are shown in Figure 1, the number is not limited. The number is selected depending on the application and the required accuracy.

[0009] (Laser scanner) The laser scanner 100 comprises a tripod 11, which is a support means for supporting the laser scanner 100 at the installation site; a base portion 12 fixed to the upper part of the tripod 11; a horizontal rotating portion 13 held on the base portion 12 in a manner that allows for horizontal rotation; and a vertical rotating portion 14 held on the horizontal rotating portion 13 in a manner that allows for vertical rotation. Horizontal rotation is rotation around an axis extending in the vertical direction. Vertical rotation is rotation around an axis extending in the horizontal direction.

[0010] The vertical rotating unit 14 is equipped with an optical system that emits and receives distance measuring light. Laser distance measuring light is emitted in pulses from this optical system. The horizontal rotating unit 13 and the vertical rotating unit 14 rotate electrically. The horizontal rotation angle of the horizontal rotating unit 13 and the vertical rotation angle of the vertical rotating unit 14 are measured by encoders. Laser scanning is performed by rotating the horizontal rotating unit 13 horizontally and then rotating the vertical rotating unit 14 vertically while emitting distance measuring light in pulses from the optical system.

[0011] (Data processing device) The processing related to the calculation of the target position is performed by the data processing device 300. The data processing device 300 is configured using a computer. The computer that constitutes the data processing device 300 includes a CPU, RAM, ROM, a hard disk drive, and an interface device.

[0012] Figure 2 is a functional block diagram of the data processing device 300. The functional parts in Figure 2 are realized by the execution of a computer program by the CPU of the computer that constitutes the data processing device 300. The computer program is stored in the memory of the computer. It is also possible to realize at least one of the functional parts in Figure 2 with dedicated hardware. It is also possible to integrate the data processing device 300 into the laser scanner 100.

[0013] The data processing device 300 comprises a point cloud data receiving unit 301, a calculation unit 302, and a communication device 303. The point cloud data receiving unit 301 receives point cloud data obtained by laser scanning using the laser scanner 100.

[0014] The arithmetic unit 302 performs the processing shown in Figures 3 to 5. The communication device 303 communicates with the laser scanner 100 and with other devices. Communication is performed using a wireless LAN line or an internet line.

[0015] (An example of processing) FIG. 3 is a flowchart showing an example of the procedure of the processing performed in the data processing apparatus 300 of FIG. 2. It is assumed that the dimensions (diameter) of the spherical targets are known. Prior to the processing of FIG. 3, a laser scan is performed using the laser scanner 100 to include the spherical targets 201 and 202 in the scan range.

[0016] (Step S101) Obtain the laser scan data obtained by the laser scanner 100.

[0017] (Step S102) Create a depth map based on the laser scan data obtained in Step S101. The depth map is a 2D map with depth direction information (distance information) added to the point cloud. As a depth map, a form in which approximately similar distance ranges are grouped and divided into areas is also possible.

[0018] (Step S103) The diameters of the spherical targets 200 and 201 are known. Therefore, if the distance from the laser scanner 100 is determined, the approximate apparent diameter as seen from the laser scanner 100 is determined. In the depth map created in Step 102, estimate the circular shape (i.e., the apparent shape) of the apparent diameter of the spherical targets 200 and 201 corresponding to the distance information, and search for the point cloud that fits the circle.

[0019] In the above search, move the circle estimated above on the depth map. At this time, change the diameter of the circle according to the distance information on the depth map. That is, the diameter of the above circle is relatively small at a far distance and relatively large at a near distance. Search for the point cloud distributed in a circular shape that fits this circle.

[0020] (Step S104) In step S103, the location of the searched point cloud is used as the specified coordinate, and the surrounding point cloud is clipped. For example, the coordinates of the circular point cloud searched in step S102 are used as the specified coordinate, and the point cloud is clipped from a range 1.2 times its distribution radius.

[0021] (Step S105) Outliers (abnormal values) are removed from the point cloud obtained in step S104.

[0022] (Step S106) In step S105, the center position (center coordinates) of the point cloud from which outliers have been removed is calculated.

[0023] (Step S107) Using the point cloud obtained in step S101, a spherical fitting is performed with the center position calculated in step S106 as the initial value and the radius as the variable.

[0024] (Step S108) In step S107, the center coordinates of the point cloud fitted to the sphere are calculated.

[0025] (Step S109) The point cloud obtained in step S101 is subjected to a second fitting using the center coordinates obtained in step S108 as initial values ​​and a constant radius. The radius is calculated based on the distance between the laser scanner 100 and the center coordinates obtained in step S108, and based on the radii of known spherical targets 200 and 201.

[0026] (Step S110) In step S109, outliers are removed from the resulting point cloud. Specifically, the standard deviation of the deviation from the sphere is calculated, and points exceeding a coefficient multiple of the standard deviation (×2) are removed as outliers (noise points).

[0027] (Step S111) The point cloud processed in step S110 is subjected to another fitting with a constant radius to calculate the final center position (the center position of the spherical shapes of spherical targets 200 and 201).

[0028] (Superiority) The effects of noise in detecting spherical targets using a laser scanner are effectively eliminated, enabling highly accurate estimation of the center position. The process in step S103 involves searching for a partial point cloud that is fitted two-dimensionally to the target point cloud. This process is computationally intensive and highly efficient.

[0029] Performing a three-dimensional fitting from the outset is computationally intensive and prone to errors due to noise. The process in step S103 suppresses the occurrence of this problem. In this embodiment, a point cloud is extracted to fit the target shape in stages. That is, first a first fitting is performed using the 2D distribution of the point cloud, and then a 3D distribution fitting is performed. This method is highly efficient and allows for high accuracy in calculating the final target center position.

[0030] In the process of extracting point clouds from the point cloud obtained by the laser scanner 100 to fit the apparent shape of a specific shape (step S103), the size of the apparent shape (in this case, a circle) is adjusted according to the distance from the laser scanner 100 to the point cloud to be fitted. This enables efficient and highly accurate detection of spherical targets.

[0031] (2) Second embodiment An example of using circular targets instead of spherical targets 200 and 201 is described. Figure 4 is a flowchart of an example of the process. A circular target is a disc-shaped target. The dimensions of the circular target are assumed to be known.

[0032] (Step S201) The laser scan data is acquired. The details are the same as in step S101.

[0033] (Step S202) Create a depth map. The details are the same as in step S101.

[0034] (Step S203) The system searches for elliptical point clouds with dimensions corresponding to the distance, referencing distance information on the depth map. Elliptical shapes are targeted because the circular target may not always be facing directly towards the laser scanner 100. The above search uses the ellipticity of the ellipse as a variable. Therefore, circular shapes are also included in the search conditions (a circular target appears circular when viewed from the front).

[0035] (Step S204) In step S203, the point cloud location found is used as the specified coordinate, and the surrounding point cloud is clipped.

[0036] (Step S205) Outlines are removed from the point cloud obtained in step 204.

[0037] (Step S206) Step 204 processes the point cloud and extracts point clouds that exhibit similar reflection characteristics. Specifically, one of the following processes (1) to (4) is performed: (1) Extraction of point clouds exhibiting similar reflection intensity, (2) Extraction of point clouds exhibiting similar color (RGB intensity or brightness), (3) Extraction of point clouds with a predetermined reflection intensity, (4) Extraction of point clouds with a predetermined color. The point clouds are assumed to be distributed in a circular pattern.

[0038] (Step S207) In step S206, a point cloud is extracted that fits the outer periphery of the extracted point cloud into a circular shape. That is, from the point cloud extracted in step S206, a point cloud along the outer edge of the circle is extracted. The size of the circle is estimated from the average distance from the laser scanner 100 to the target point cloud.

[0039] (Step S208) In step S207, the center coordinates of the extracted point cloud (a point cloud distributed on the circumference) are calculated. These center coordinates will be the coordinates of the center of the circular target.

[0040] (3) Third Embodiment This section describes the case where a cone target is used instead of spherical targets 200 and 201. Figure 5 is a flowchart of an example of the process. A cone target is a target with a cone shape. The dimensions of the cone target are assumed to be known. A cone target is also called a cone target and is used by placing it on the ground or floor with its apex facing upwards.

[0041] (Step S301) The laser scan data is acquired. The details are the same as in step S101.

[0042] (Step S302) Create a depth map. The details are the same as in step S101.

[0043] (Step S303) The system references distance information on the depth map and searches for a cloud of triangular points with dimensions corresponding to the distance.

[0044] (Step S304) After processing in step S303, point clouds exhibiting similar reflection characteristics are extracted. Specifically, one of the following processes (1) to (4) is performed: (1) Extraction of point clouds exhibiting similar reflection intensity, (2) Extraction of point clouds exhibiting similar color (RGB intensity or brightness), (3) Extraction of point clouds with a predetermined reflection intensity, (4) Extraction of point clouds with a predetermined color.

[0045] (Step S305) In step S304, point clouds are extracted from the extracted point clouds to fit the shape of the cone of the cone target.

[0046] (Step S306) Based on the point cloud extracted in step S305, the coordinates of the cone's vertex are determined.

[0047] (Step S307) Specifying the radius of the base of the cone shape, a point cloud is extracted from the point cloud extracted in step S304 to fit the shape of the cone target. In this process, the coordinates of the vertices obtained in step S306 are used as constraints.

[0048] (Step S308) In step S307, the ground point cloud is removed from the extracted point cloud. In this process, the point cloud that fits to a plane extended perpendicular to the axis of the cone shape is removed as the ground point cloud. If the cone target is placed on the floor surface, the point cloud to be removed will be the floor surface point cloud.

[0049] (Step S309) Based on the point cloud processed in step S308, the center of the cone target is determined. The center of the cone target is predetermined as the position of its vertex or centroid.

[0050] (4) Fourth Embodiment It is also possible to perform the following operations instead of the operations in steps S201 to S203 and steps S301 to S303.

[0051] (If the target is circular) First, point cloud data is acquired. Next, point clouds with normals in the same direction are separated from the acquired point cloud data. Then, from the point clouds separated by normals, an elliptical shape set to the size of a circular target according to the distance is searched for. At this time, the ellipticity is adjusted based on the direction of the normals. After this, the processes from step S204 onwards are executed.

[0052] (If the target is a cone-shaped target) First, point cloud data is acquired. Next, point clouds with normals in the same direction are separated from the acquired point cloud data. Then, the ground region is estimated from the normal information. Next, point clouds distributed in the shape of an isosceles triangle with its base touching the estimated ground region are searched for. Alternatively, point clouds distributed in the shape of an isosceles triangle with its vertex in the direction of the normal of the estimated ground region are searched for. At this time, the size of the isosceles triangle is adjusted according to the distance to the point cloud to match the dimensions of the actual cone target. After this, the processing from step S304 onwards is executed. This configuration can also handle cases where the cone target is placed on sloping ground.

[0053] (5) Fifth embodiment It is also possible to perform the following operations instead of the operations in steps S101 to S103, steps S201 to S203, and steps S301 to S303.

[0054] (Using RGB maps) First, point cloud data is acquired. Next, candidate point clouds for the target are searched from the acquired point cloud data based on the target's color, pattern, and color scheme (or relative color difference). During this process, information on the distance and normal vectors of the point clouds is also referenced to find candidate point clouds for the target.

[0055] (Using a brightness map) The target is identified by its brightness. High brightness, low brightness, and a predetermined specific reflectance value are used to search for candidate point clouds representing the target. During this process, information on the distance and normal vector of the point clouds is also referenced to find candidate point clouds representing the target.

[0056] (6) Sixth Embodiment In step S103, step S203, or S303, the target point cloud may be rotated to search for a point cloud that fits the apparent shape of the target. [Explanation of Symbols]

[0057] 100... Laser scanner, 11... Tripod, 12... Base unit, 13... Horizontal rotation unit, 14... Vertical rotation unit, 200... Spherical target, 201... Spherical target, 300... Data processing unit.

Claims

1. A data processing device that extracts point clouds relating to a target of a specific shape from point clouds obtained by laser scanning, A point cloud is obtained by laser scanning with the aforementioned target as the scan range. From the aforementioned point cloud, a point cloud is extracted that fits the visual shape of the specific shape. A data processing device comprising a calculation unit that extracts a point cloud that fits the extracted point cloud to the specific shape.

2. The aforementioned specific shape is spherical, The data processing device according to claim 1, wherein the apparent shape is circular.

3. The aforementioned specific shape is circular, The data processing device according to claim 1, wherein the apparent shape includes an elliptical shape.

4. The aforementioned specific shape is a cone, The data processing device according to claim 1, wherein the apparent shape is triangular.

5. From the point cloud obtained by the aforementioned laser scan, the point cloud of the ground is extracted. The data processing device according to claim 1, wherein the apparent shape is an isosceles triangle whose base is tangent to the point cloud of the ground.

6. In the process of extracting a point cloud from the aforementioned point cloud that fits the visual shape of the specific shape, The data processing device according to any one of claims 1 to 5, wherein the apparent size of the shape is adjusted according to the distance from the laser scanner that performed the laser scan to the point cloud to be fitted.

7. A data processing method for extracting point clouds related to a target of a specific shape from point clouds obtained by laser scanning, A point cloud is obtained by laser scanning with the aforementioned target as the scan range. From the aforementioned point cloud, a point cloud is extracted that fits the visual shape of the specific shape. A data processing method for extracting a point cloud that fits the specific shape from the extracted point cloud.

8. A program that causes a computer to perform data processing to extract point clouds related to a target of a specific shape from a point cloud obtained by laser scanning, to the computer A point cloud is obtained by laser scanning with the aforementioned target as the scan range. From the aforementioned point cloud, extract a point cloud that fits the visual shape of the specific shape. A program that extracts a point cloud that fits the specific shape from the extracted point cloud.

Images