Measuring system

Abstract

The present disclosure is a point cloud processing device that sequentially acquires point cloud data indicating the shape of a structure, extracts from the point cloud data a candidate point cloud of a columnar object by performing circle detection, groups circles present in a prescribed range from a set direction vector by extracting said circles from the candidate point cloud, and extracts the columnar object by using the grouped circles.

Description

Measurement System

[0001] The present disclosure relates to a technique for extracting cylindrical objects from point cloud data.

[0002] A technology has been developed that uses an on-board 3D laser scanner (Mobile Mapping System: MMS) to create a 3D model of an outdoor structure (see, for example, Patent Document 1). This technology creates a point cloud and a scan line in a space where no point cloud exists, and then creates a 3D model, thereby achieving good results in recall and precision even when the point cloud is sparse (when the vehicle speed is high).

[0003] We would like to realize a system that monitors the behavior of objects during construction, detects proximity and contact between objects, and notifies workers.However, conventional technology requires time to create scan lines and create 3D models, so it is not real-time capable, and there is an issue that it cannot be directly applied to point cloud data acquired by fixed 3D laser scanners that can measure and process in real time.

[0004] JP 2017-156179 A

[0005] The conventional techniques require time to complement the point cloud data, which means that they lack real-time capabilities. The present disclosure aims to achieve real-time and highly accurate cylindrical object extraction.

[0006] A measurement system according to the present disclosure includes a three-dimensional measuring device that measures the shape of a structure, and a point cloud processing device that acquires point cloud data indicating the shape of the structure from the three-dimensional measuring device.

[0007] The point cloud processing device of the present disclosure executes the point cloud processing method of the present disclosure, in which the point cloud processing device sequentially acquires point cloud data indicating the shape of a structure, performs circle detection from the point cloud data to extract a candidate point cloud of a cylindrical object, extracts circles that exist within a predetermined range from a set direction vector from the candidate point cloud to group them, and extracts the cylindrical object using the grouped circles.

[0008] In the extraction, a circle including a group of points having a predetermined reflection intensity may be extracted from the group of candidate points.

[0009] The above disclosures can be combined as much as possible.

[0010] According to the present disclosure, it is possible to realize real-time and highly accurate extraction of cylindrical objects.

[0011] 1 shows an example of the configuration of a measurement system according to the present disclosure; FIG. 2 shows an example of a scan line; FIG. 3 shows an example of a three-dimensional model of a utility pole and a power line; FIG. 4 shows an example of a point cloud processing method according to the present disclosure; FIG. 5 shows an example of a cluster; FIG. 6 shows an example of a circle detection; FIG. 7 shows an example of a grouping procedure executed by a point cloud processing device; and FIG. 8 is an explanatory diagram of the grouping procedure.

[0012] Hereinafter, 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 indicate the same components.

[0013] 1 shows an example of the configuration of a measurement system according to the present disclosure. The measurement system according to the present disclosure includes a point cloud processing device 91 and a three-dimensional measuring device 92. The three-dimensional measuring device 92 measures the shape of a structure. The point cloud processing device 91 is a computer that acquires point cloud data indicating the shape of the structure from the three-dimensional measuring device 92 and performs any processing using the point cloud data.

[0014] The three-dimensional measuring device 92 is any device that can acquire point cloud data representing the three-dimensional coordinates of the surface shape of any structure 93, such as a utility pole 81 or a power line 82, by scanning the surface of the structure, and may be, for example, a three-dimensional laser scanner. The three-dimensional laser scanner may be, for example, a three-dimensional LiDAR (Light Detection and Ranging), or may be an MMS.

[0015] The point cloud processing device 91 sequentially acquires and updates point cloud data in real time from the three-dimensional measuring device 92. The point cloud data may be generated from a stereo camera or images, instead of using the three-dimensional measuring device 92 such as a three-dimensional laser scanner. The point cloud processing device 91 may overlay several frames of point cloud data.

[0016] The point cloud processing device 91 may create a 3D model of a structure using the point cloud data. For example, the point cloud processing device 91 extracts an object, creates a 3D model of the object using the extracted point cloud data, and saves the created model information. To create the 3D model, for example, the point cloud processing device 91 connects point clouds 71 ​​to create a scan line 72 as shown in FIG. 2, and can create a 3D model 73 of a utility pole and a 3D model 74 of an electric wire as shown in FIG. 3 based on the scan line 72.

[0017] In the past, when creating the 3D model 73 of a utility pole, candidate circles were extracted by searching for points where the detected circles were connected in the height direction in all directions in any direction. As a result, circles that were not utility poles were also included as utility pole circles, which resulted in the creation of a 3D model with a shape that was different from that of a cylindrical object.

[0018] Therefore, in the present disclosure, when the target object is a cylindrical object such as a utility pole, the circular shape of the point cloud data reflected by the cylindrical object is distinguished from the circular shapes of the point cloud data reflected by other structures. Specifically, the circular shapes detected from the point cloud data are grouped. The grouping can be performed based on, for example, the following information: - The position of the center point of the circle - The size of the radius of the circle - The reflection intensity of the point cloud data

[0019] When dividing into groups, the center points of the circles are searched for and conditions are classified based on reflection intensity to improve accuracy. Whether the objects that make up the circle group are cylindrical (utility poles) is determined based on information such as the height and inclination of the objects. The identified cylindrical objects are then modeled.

[0020] Specifically, the measurement system of this embodiment includes the following steps: Step S1: generating a candidate point group made up of circles detected as candidates for a cylindrical object from input point cloud data; Step S2: calculating an angle α between a vertical line passing through the center point of the Nth circle and a line passing through the center points of the Nth circle and the N+1th circle, where the bottom circle included in the candidate point group is the Nth circle; and Step S3: calculating an angle α between the vertical line passing through the center points of the Nth circle and the N+1th circle, where the angle α is equal to or smaller than the threshold angle α. Ｔ If α is greater than the threshold angle α, the N+1th circle is excluded from the candidate point group and the next circle is renumbered to N+1. Ｔ Step S2: Repeating this until a circle satisfying the following condition is found; setting the initial value of i to 1, calculating the angle β formed by the line passing through the center point of the N+i-1th circle and the center point of the N+ith circle and the line passing through the center point of the N+ith circle and the center point of the N+i+1th circle, and determining whether β is a threshold angle β Ｔ If β is greater than the threshold angle β, the (N+i+1)th circle is excluded from the candidate point group and the next circle is renumbered to (N+i+1)th. Ｔ The method executes the following steps: Step S3, which repeats the process until a circle satisfying the following criteria is found; Step S4, which repeats Step S3 while incrementing i by 1 until the (N+i+1)th circle becomes the topmost circle included in the candidate point cloud; and Step S5, which generates a cylindrical object model based on the multiple circles remaining in the candidate point cloud. This will be described in detail below.

[0021] 4 shows an example of the point cloud processing method of the present disclosure. A point cloud processing device 91 of the present disclosure executes the point cloud processing method of the present disclosure. The point cloud processing method of this embodiment includes steps S11 to S17. Specifically, the point cloud processing device 91 executes steps S11 to S17. S11: Point density clustering process S12: Circle detection process S13: Grouping process S14: Cylindrical object determination process S15: Cylindrical object extraction process S16: Create a 3D model of the cylindrical object S17: Save 3D model information

[0022] In step S12, the above-mentioned step S1 is executed, and in step S13, the above-mentioned steps S2 to S4 are executed. Steps S14 to S17 correspond to the above-mentioned step S5.

[0023] (Step S11) The point cloud processing device 91 sequentially acquires point cloud data and performs clustering processing based on the point density. Through clustering, the point cloud processing device 91 extracts a point cloud G0 that may include cylindrical objects as shown in FIG. 5. The parameters of the extraction algorithm can be set to any value. For example, it is possible to identify the utility pole point cloud by manually extracting and labeling a utility pole point cloud, a cable point cloud, a wall surface point cloud, etc.

[0024] Here, the point cloud processing device 91 applies filtering based on reflection intensity to the input point cloud, thereby reducing the number of points used in the calculation and increasing the speed.

[0025] (Step S12) The point cloud processing device 91 divides the point cloud G0 into multiple parts and performs circle detection from the point cloud data in each divided area to extract candidate point clouds for cylindrical objects. The division is performed, for example, at a width set in the longitudinal direction of the point cloud G0. In this case, the point cloud processing device 91 may set the longitudinal direction of the point cloud G0 to the z-axis direction and divide the point cloud G0 along a plane parallel to the horizontal plane.

[0026] FIG. 6 shows an example of the detection of a circle 75. For circle detection, known RANSAC (Random Sample Consensus) processing or the like can be used. Here, before performing circle detection, a process may be performed in which the three-dimensionally scattered point clouds contained in each longitudinally divided region are projected onto a plane perpendicular to the longitudinal direction for each region. Performing circle detection on a plane in this manner improves accuracy and computation speed compared to performing circle detection in a three-dimensionally scattered state. The point cloud processing device 91 may also extract multiple circles with different radii. Furthermore, clustering may be performed based on point density in step S11, and circle detection may be performed for clusters with point densities equal to or greater than a threshold in step S12, thereby increasing the processing speed.

[0027] (Step S13) The point cloud processing device 91 groups the candidate point cloud by extracting circles that exist within a predetermined range from the set direction vector. The predetermined range to search can be arbitrarily determined using parameters. Figure 7 shows an example of the grouping procedure executed by the point cloud processing device 91. For example, the point cloud processing device 91 performs steps S21 to S27. The position of the center point of the N+2th circle is searched for based on the angle between the Nth and N+1th circles. This is explained in detail below.

[0028] S21: The point cloud processing device 91 calculates the angle formed by the line passing through the center point of the Nth divided circle and the center point of the N+1th divided circle, as shown in Fig. 8. The line connecting the center point of the Nth divided circle to the center point of the N+1th divided circle corresponds to the set direction vector.

[0029] In this embodiment, the bottom row where the circle extraction has been successful is the Nth division. In this case, since the N-1th division circle has not been extracted, a vertical line L0 in the z-axis direction can be used instead of the line passing through the center point of the N-1th division circle and the center point of the Nth division circle. At this time, the point cloud processing device 91 extracts a line L0 passing through the center point C1 of the Nth division circle and the center point C2 of the N+1th division circle. １ The angle α between the line L0 and the line L1 is calculated. The line L0 corresponds to the set direction vector.

[0030] S22: The point cloud processing device 91 determines whether the angle α is equal to or smaller than the threshold angle α Ｔ It is determined whether the angle is within the threshold angle α Ｔ can be set to a value according to the expected inclination of the cylindrical object, such as 50°. Ｔ If the angle α is within the threshold angle α, steps S23 to S25 are executed. Ｔ If the number of circles exceeds the number of divisions, step S21 is performed for each candidate circle in the N+1th division in order (S26). If there is no candidate circle in the N+1th division, the process ends.

[0031] S23 and S24: The point cloud processing device 91 calculates the angle β between a line L1 passing through the center point C1 of the Nth divided circle and the center point C2 of the (N+1)th divided circle, and a line L2 passing through the center point C2 of the (N+1)th divided circle and the center point C3 of the (N+2)th divided circle. The line L1 corresponds to the set direction vector. In this way, the angle β between the lines passing through the center points of the adjacent circles one level above is calculated.

[0032] S25: The angle β is the threshold angle β Ｔ It is determined whether the angle is within the threshold angle β. Ｔ Regarding (1), since the straight line L1 approximately represents the central axis of the cylindrical object, the angle can be set to within 20°, for example, about 15°.

[0033] The angle β is the threshold angle β Ｔ If the angle β is within the threshold angle β, the circle is moved up by one step and steps S23 and S24 are executed again. This process is executed until no circle is detected in the height direction, and the angle β is equal to or smaller than the threshold angle β. Ｔ Group circles within a certain range.

[0034] On the other hand, the angle β is the threshold angle β Ｔ If the number of circles exceeds 1, step S24 is performed for each candidate circle in the row above the bottom row (for example, the N+1th division) in order (S27). If there are no candidate circles, the process ends.

[0035] In this manner, in this embodiment, an appropriate circle is extracted from among the candidate circles based on the direction vector defined by the center point of circle 75. Both angles α and β are angles centered on the direction vector, and may be considered in a two-dimensional plane (not limited to the x-z plane) as shown in Figure 8, or in three-dimensional space.

[0036] 7 is based on the assumption that no circles are missing from each divided region, but circles may be missing along the way. For example, even if only the Nth and N+3th divided circles exist, the process may proceed to step S14. This allows the cylindrical object to be properly extracted and the entire length of the cylindrical object to be properly measured, even if the point cloud cannot be measured due to an obstacle.

[0037] (Step S14) The point cloud processing device 91 determines cylindrical objects using the circles grouped in step S13. For example, among the circles in the same group, circles whose center points are arranged on the same straight line and whose radii are within an error range are determined to be the same cylindrical object.

[0038] (Step S15) The point cloud processing device 91 extracts cylindrical objects using the center points of the circles that have been determined to be the same cylindrical object in step S13 from among the grouped circles.

[0039] (Steps S16 and S17) The point cloud processing device 91 models the cylindrical object extracted in step S14 and stores the three-dimensional model information in the storage medium 91M. For example, the longitudinal length of the cylindrical object is calculated by connecting the center points of the extracted circles, and the radius of the cylindrical object model is calculated using the radius of the extracted circles, thereby creating the cylindrical object model.

[0040] As described above, the measurement system of this embodiment can extract cylindrical objects with high accuracy from point cloud data acquired by a fixed three-dimensional measuring device 92 capable of real-time processing, without using a dictionary.

[0041] If the reflection intensity R of the object is known, the point cloud processing device 91 may further extract a circle including a point cloud of a predetermined reflection intensity from the candidate point cloud in the grouping of steps S21 to S27. For example, for reflection intensity R, which can take a value of 0≦R≦65535, by setting the reflection intensity threshold to a value between 5 and 20, the point cloud of a concrete pole can be extracted.

[0042] The point cloud processing device 91 can extract point cloud data of a cylindrical object more accurately by executing steps S21 to S27 for a circle including a point cloud that satisfies the reflection intensity threshold. For example, the percentage of point clouds that satisfy the reflection intensity threshold can be set arbitrarily, but can be set to, for example, 30% or more.

[0043] The point cloud processing device 91 may determine that points in a series of circles belong to the same group if the difference between the average values ​​of the point cloud reflection intensities R included in each circle is within 7, where the reflection intensities R can take values ​​of 0≦R≦65535. In addition, the number of point clouds included in a cluster, the difference in circle radius, etc. can also be used to improve accuracy and speed.

[0044] For example, in steps S21 to S27, the point cloud processing device 91 may determine that circles having a radius difference equal to or greater than a predetermined circle radius difference threshold belong to a different group. This allows the point cloud data of guy wires and support poles located near utility poles to be in a different group from the point cloud data of the utility poles, making it possible to extract the point cloud data of the utility poles more accurately.

[0045] (Second Embodiment) In this embodiment, the height direction with respect to the ground is defined as the z-axis. If a cylindrical object is extracted using the above procedure without tilting the z-axis in any direction, it is possible to extract cylindrical objects tilted at 45° with respect to the z-axis. If the target object is a utility pole under construction, the direction and angle it faces will change depending on the situation, such as when it is placed on the ground, when it is held or suspended, or when it is freestanding. Therefore, by inputting the longitudinal angle of the cylindrical object, which is the target object, in advance using the three-dimensional measuring device 92 as a reference, it is possible to accommodate various angles.

[0046] The following are examples of methods for inputting the longitudinal angle of a cylindrical object in advance: First method: The angle is judged visually before calculation and input by a person into the point cloud processing device 91. Second method: An object made of a highly reflective material or structure or an object with an unusual shape is attached to a target object such as a utility pole or heavy machinery and used as a marker for axis calculation.

[0047] By inputting the angle in advance, it is possible to reduce the calculation time for the longitudinal direction of the point cloud G0 in the first embodiment. It is also possible to extract utility pole point clouds from all directions without narrowing down the angle.

[0048] In the first method, the distance from the three-dimensional measuring device 92 to the bottom and top of the cylindrical object is input in advance to the point cloud processing device 91, making it possible to calculate the amount of tilt in the longitudinal direction of the cylindrical object. However, in this embodiment, the tilt direction of the cylindrical object cannot be identified, so the point cloud processing device 91 keeps the amount of tilt in the longitudinal direction of the cylindrical object constant and performs circle detection for each direction in the horizontal plane. Then, thresholds are set for the distance from the three-dimensional measuring device 92, the size of the extracted cluster, etc., and a utility pole point cloud is extracted.

[0049] In the second method, for example, two highly reflective objects are attached to the grip of a heavy machine. The point cloud processing device 91 extracts the highly reflective objects as markers in step S11 or S12, and can set the axis connecting the two markers as the z-axis shown in FIG. 5 in step S12.

[0050] Other Embodiments The device of the present invention 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.

[0051] 71: Point cloud 72: Scan line 73: 3D model of utility pole 74: 3D model of power line 81: Utility pole 82: Power line 91: Point cloud processing device 91M: Storage medium 92: 3D measuring device 93: Structure

Claims

1. A point cloud processing device that sequentially acquires point cloud data that indicates the shape of a structure, extracts a candidate point cloud of cylindrical objects by performing circle detection from the point cloud data, groups the candidate point cloud by extracting circles that exist within a predetermined range from a set direction vector, and extracts cylindrical objects using the grouped circles.

2. The point cloud processing device according to claim 1, wherein, in the grouping, a circle including a point cloud having a predetermined reflection intensity is further extracted from the candidate point cloud.

3. A measurement system comprising: a three-dimensional measuring device that measures the shape of a structure; and a point cloud processing device according to claim 1 or 2 that acquires point cloud data indicating the shape of the structure from the three-dimensional measuring device.

4. A point cloud processing method comprising: sequentially acquiring point cloud data indicating the shape of a structure; extracting a candidate point cloud for a cylindrical object by performing circle detection from the point cloud data; extracting circles that exist within a predetermined range from a set direction vector from the candidate point cloud to group them; and extracting cylindrical objects using the grouped circles.

Images