Structural shape estimation device and structural shape estimation program
The structural shape estimation device and program improve feature point extraction and correction to ensure accurate movement routes for inspection equipment by dividing points into subdivided regions and identifying centroids, addressing inaccuracies in existing methods.
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- CHUBU ELECTRIC POWER CO INC
- Filing Date
- 2024-11-06
- Publication Date
- 2026-05-19
AI Technical Summary
Existing methods for determining the movement route of inspection equipment based on feature points from three-dimensional point clouds are inaccurate, leading to unsuitable movement courses for inspecting structures.
A structural shape estimation device and program that extracts feature points along the external contour of a structure by dividing points into subdivided regions and identifying the centroid of each region, ensuring accurate feature point extraction and correction for consistent shape estimation.
Ensures accurate feature points corresponding to the external contour, preventing unsuitable movement courses for inspection equipment and maintaining consistent shape estimation even for structures with varying point clouds representing identical shapes.
Smart Images
Figure 2026081813000001_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a structure shape estimation device and a structure shape estimation program.
Background Art
[0002] In order to inspect a structure, it is conceivable to move inspection equipment such as an inspection robot or a drone along the contour on the appearance of the structure. Specifically, when the inspection equipment moves along the movement route along the contour on the appearance of the structure, the inspection equipment checks whether there are any abnormalities such as cracks in the structure. In this case, in order to determine the movement route of the inspection equipment, it is necessary to grasp the shape of the structure.
[0003] For example, in Patent Document 1, the shape of a structure is estimated based on the three-dimensional coordinates of each point in a three-dimensional point cloud representing the structure. Specifically, an image including the three-dimensional point cloud representing the structure is divided into a number of细分 regions. Then, the center point of the细分 region including each point of the three-dimensional point cloud among the number of细分 regions is determined as a feature point representing the contour on the appearance of the structure. Since such feature points are determined for each of the number of细分 regions, it becomes possible to estimate the shape of the structure based on those feature points. The movement route of the inspection mechanism is determined based on the feature points determined for each细分 region.
Prior Art Documents
Patent Documents
[0004]
Patent Document 1
Summary of the Invention
Problems to be Solved by the Invention
[0005] However, in Patent Document 1, the center point of the sub-region where each point in the three-dimensional point cloud representing a structure exists among the numerous sub-regions is defined as a feature point corresponding to the external contour of the structure. Therefore, the defined feature point is not necessarily accurate in corresponding to the external contour of the structure. If the defined feature point is not accurate in corresponding to the external contour of the structure, when the movement course of the inspection equipment is determined based on the feature point for each of the numerous sub-regions, that movement course may be unsuitable for inspecting the structure with the inspection equipment. [Means for solving the problem]
[0006] The following describes the means and effects of solving the above problems. A structural shape estimation device that solves the above problems includes a control unit that extracts a number of feature points along the external contour of the structure based on the three-dimensional coordinates of each point in a three-dimensional point cloud representing the structure and its surroundings, and estimates the shape of the structure based on these feature points. The control unit is configured to perform a point cloud extraction process followed by a feature point extraction process. The point cloud extraction process extracts each point representing the external contour of the structure from a three-dimensional point cloud representing the structure and its surroundings. The feature point extraction process divides each point representing the external contour of the structure into a number of subdivided regions and finds the centroid of each point located within one of the subdivided regions. The feature point extraction process extracts the point closest to the centroid among the above points as a feature point corresponding to a location within the subdivided region of the external contour of the structure, and performs such feature point extraction for each subdivided region.
[0007] According to the above configuration, each point representing the external outline of the structure is divided into numerous sub-regions. Then, among the points present in each sub-region, the point closest to the centroid of those points is extracted as a feature point corresponding to the location within the sub-region of the external outline of the structure. Furthermore, such feature point extraction is performed for each of the numerous sub-regions. As a result, the feature points extracted for each of the numerous sub-regions become accurate as points corresponding to the external outline of the structure.
[0008] Examples of the above-mentioned structures include multiple blades of the same shape in a wind turbine. The control unit is configured to perform a feature point extraction process followed by a feature point correction process. The feature point extraction process extracts feature points for each of the multiple blades, corresponding to the outlines of their external appearance. The feature point correction process estimates the shape of each of the multiple blades based on the feature points extracted by the feature point extraction process, and corrects the feature points of the other blades so that their shapes are the same as the shape of the largest blade.
[0009] With the above configuration, depending on how the three-dimensional point cloud representing the blades of the wind turbine is acquired, even blades of the same shape may have three-dimensional point clouds representing different shapes. In this case, when feature points are extracted for each subdivided region corresponding to the external contour of each of the multiple blades, these feature points will differ for each blade. However, even in this case, the feature points of the other blades are corrected so that the shape of the other blades is the same as the shape of the largest blade among the multiple blades whose shapes are estimated from the feature points. This makes it possible to prevent the movement course of an aircraft, such as a drone used for inspection equipment, from becoming too close to the other blades when the movement course is determined based on the above feature points.
[0010] The structural shape estimation program that solves the above problem extracts numerous feature points along the external contour of the structure based on the three-dimensional coordinates of each point in a three-dimensional point cloud representing the structure and its surroundings. This process involves having the computer sequentially perform point cloud extraction and feature point extraction. Furthermore, the computer is made to estimate the shape of the structure based on the numerous extracted feature points. The point cloud extraction process extracts each point representing the external contour of the structure from the three-dimensional point cloud representing the structure and its surroundings. The feature point extraction process divides each point representing the external contour of the structure into numerous subdivided regions and finds the centroid of each point within one of the subdivided regions. The feature point extraction process extracts the point closest to the centroid among the above points as a feature point corresponding to the location within the subdivided region of the external contour of the structure, and performs such feature point extraction for each subdivided region.
[0011] According to the above configuration, each point representing the external outline of the structure is divided into numerous sub-regions. Then, among the points present in each sub-region, the point closest to the centroid of those points is extracted as a feature point corresponding to the location within the sub-region of the external outline of the structure. Furthermore, such feature point extraction is performed for each of the numerous sub-regions. As a result, the feature points extracted for each of the numerous sub-regions become accurate as points corresponding to the external outline of the structure.
[0012] The above-mentioned structure includes multiple blades of the same shape in a wind turbine. The computer is said to perform a feature point extraction process followed by a feature point correction process. The feature point extraction process extracts feature points for each of the multiple blades, corresponding to the sub-regions of their external contours. The feature point correction process estimates the shape of each of the multiple blades based on the feature points extracted by the feature point extraction process, and corrects the feature points of the other blades so that their shapes are the same as the shape of the largest blade.
[0013] With the above configuration, depending on how the three-dimensional point cloud representing the blades of the wind turbine is acquired, even blades of the same shape may have three-dimensional point clouds representing different shapes. In this case, when feature points are extracted for each subdivided region corresponding to the external contour of each of the multiple blades, these feature points will differ for each blade. However, even in this case, the feature points of the other blades are corrected so that the shape of the other blades is the same as the shape of the largest blade among the multiple blades whose shapes are estimated from the feature points. This prevents the movement course of an aircraft, such as a drone used for inspection equipment, from becoming too close to the other blades when the movement course is determined based on the above feature points. [Brief explanation of the drawing]
[0014] [Figure 1] This is a schematic diagram showing the configuration of a computer. [Figure 2] This is a side view of a wind turbine. [Figure 3] Figure 2 is a front view showing multiple blades in a wind turbine. [Figure 4] Figure 1 is a flowchart showing the execution procedure of a computer-based program for estimating the shape of a structure. [Figure 5] Figure 2 is a perspective view showing a three-dimensional point cloud representing a wind turbine. [Figure 6] Figure 5 is a perspective view showing each point representing multiple blades cut from a three-dimensional point cloud. [Figure 7] Figure 6 shows perspective views representing the external contours of multiple blades extracted from each of the points shown. [Figure 8] This is a perspective view showing each point in Figure 7 divided into numerous subdivided regions. [Figure 9] Figure 8 is an enlarged view showing the feature points extracted for each subdivided region. [Modes for carrying out the invention]
[0015] Hereinafter, an embodiment of a structure shape estimation device and a structure shape estimation program will be described with reference to FIGS. 1 to 9. The computer 11 shown in FIG. 1 functions as a structure shape estimation device. The computer 11 includes a central processing unit 12, a storage unit 13, and a communication unit 14. The central processing unit 12 controls various devices such as the storage unit 13 and the communication unit 14 in the computer 11, and performs arithmetic processing and the like based on various data. The storage unit 13 is for storing various data. The communication unit 14 is for performing data exchange with the outside of the computer 11.
[0016] The central processing unit 12 plays the role of a control unit for estimating the shape of the structure. Specifically, the central processing unit 12 extracts a number of feature points along the outer contour of the structure based on the three-dimensional coordinates of each point in the three-dimensional point cloud representing the structure and its surroundings. The central processing unit 12 estimates the shape of the structure based on those feature points. Examples of the structure whose shape is estimated by the central processing unit 12 include the blade of a windmill used for power generation.
[0017] FIG. 2 shows a windmill 16 including a blade 15 as a structure. The windmill 16 includes a support column 18 built on the ground 17 and a nacelle 19 disposed at the upper end of the support column 18. The blade 15 of the windmill 16 is fixed to a rotating shaft 20 protruding from the nacelle 19. The rotating shaft 20 and the blade 15 can rotate around the center line of the rotating shaft 20. Further, the blade 15 of the windmill 16 is a plurality of blades 15 having the same shape, and is arranged at equal intervals around the center line of the rotating shaft 20 as shown in FIG. 3.
[0018] The central processing unit 12 of the computer 11 shown in FIG. 1 estimates the shapes of the plurality of blades 15 based on the three-dimensional coordinates of each point in the three-dimensional point cloud representing the plurality of blades 15 and their peripheries. Specifically, the central processing unit 12 executes a shape estimation program for a structure for estimating the shapes of the plurality of blades 15. This shape estimation program causes the computer 11 to sequentially execute a point cloud extraction process and a feature point extraction process in order to estimate the shapes of the plurality of blades 15. Further, the shape estimation program for the structure causes the computer 11 to execute a feature point correction process for correcting the feature points so that the sizes of the plurality of blades 15 whose shapes have been estimated become the same.
[0019] The flowchart of FIG. 4 shows the execution procedure of the shape estimation program for the structure. FIG. 5 shows a three-dimensional point cloud representing a windmill 16 including a plurality of blades 15. As the process of step 101 (S101) in the above flowchart, the central processing unit 12 takes in the three-dimensional coordinates of each point in the three-dimensional point cloud as the data of the three-dimensional point cloud representing the windmill 16, in other words, the data of the three-dimensional point cloud representing the plurality of blades 15 and their peripheries. The data of the three-dimensional point cloud can be obtained by laser surveying the windmill 16 using a flying object such as a drone. In addition, the data of the three-dimensional point cloud may be obtained by other methods.
[0020] Such data of the three-dimensional point cloud may be stored in advance in the storage unit 13 of the computer 11. In this case, the central processing unit 12 takes in the data of the three-dimensional point cloud from the storage unit 13. The central processing unit 12 can also directly take in the data of the three-dimensional point cloud from the flying object via the communication unit 14. There may be a case where the data of the three-dimensional point cloud is stored in a server or the like. In this case, it is conceivable that the central processing unit 12 takes in the data of the three-dimensional point cloud stored in the server via the network by network connection by the communication unit 14.
[0021] The processes S102 to S104 in the flowchart above correspond to the point cloud extraction process, feature point extraction process, and feature point correction process described above, respectively. These processes will be explained individually in detail below.
[0022] <Point cloud extraction process (S102)> As part of the S102 process, the central processing unit 12 extracts points representing the external contours of the multiple blades 15 from the three-dimensional point cloud shown in Figure 5. More specifically, the central processing unit 12 classifies the three-dimensional point cloud shown in Figure 5 into points representing the multiple blades 15, points representing the support columns 18, and points representing the nacelles 19 using semantic segmentation or the like. Subsequently, the central processing unit 12 cuts out points representing the multiple blades 15 from the classified points, as shown in Figure 6. Furthermore, the central processing unit 12 extracts points representing the multiple blades 15 from the cut-out points representing the multiple blades 15, as shown in Figure 7.
[0023] <Feature point extraction process (S103)> As part of the process in S103, the central processing unit 12 divides each point representing the external contour of the multiple blades 15 into numerous subdivision regions 21, as shown in Figure 8. These subdivision regions 21 are, for example, cubic regions set in a three-dimensional coordinate system. The length of one side of a subdivision region 21 can be, for example, 30 mm. The central processing unit 12 finds the centroid of each point within a single subdivision region 21 and extracts the point closest to the centroid among these points as a feature point 22 corresponding to the location within the subdivision region 21 in the external contour of the blade 15. Furthermore, the central processing unit 12 extracts such feature points 22 for each of the numerous subdivision regions 21. The feature points 22 extracted in this way are shown in Figure 9. The central processing unit 12 extracts feature points 22 for each of the multiple blades 15, corresponding to each subdivision region 21 in their external contour.
[0024] <Feature point correction processing (S104)> As part of the process in S104, the central processing unit 12 estimates the shape of each of the multiple blades 15 based on the extracted feature points 22. The central processing unit 12 corrects the feature points 22 of the other blades so that the shapes of the other blades 15 are the same as the shape of the largest blade 15 among the multiple blades 15. Therefore, by estimating the shapes of the multiple blades 15 based on the corrected feature points 22, the sizes of the multiple blades 15 whose shapes are estimated become the same.
[0025] According to the embodiment described in detail above, the following effects and advantages can be obtained. (1) Each point representing the external contour of the multiple blades 15 is divided into numerous subdivision regions 21. Then, among the points present in one subdivision region 21, the point closest to the centroid of those points is extracted as a feature point 22 corresponding to the location of the external contour of the blade 15 within the subdivision region 21. Furthermore, such feature point 22 extraction is performed for each of the numerous subdivision regions 21. As a result, the feature points 22 extracted for each of the numerous subdivision regions 21 become accurate as points corresponding to the external contour of the blade 15.
[0026] (2) When inspection equipment such as an inspection robot or drone is moved along the external contour of the blade 15 in order to inspect the blade 15 of the wind turbine 16, the movement route of the inspection equipment for inspection is determined based on the feature points 22, which are points corresponding to the external contour of the blade 15. As described above, the feature points 22 are accurate as points corresponding to the external contour of the blade 15. Therefore, when the movement course of the inspection equipment is determined based on the feature points 22 for each of the many subdivided regions 21, it is suppressed that the movement course becomes unsuitable for the inspection of the blade 15 by the inspection equipment.
[0027] (3) Depending on how the three-dimensional point cloud representing the blades 15 of the wind turbine 16 is acquired, even if the blades 15 have the same shape, the three-dimensional point clouds representing them may represent different shapes. In this case, when feature points 22 are extracted for each subdivided region 21 corresponding to the external contour of each of the multiple blades 15, these feature points 22 will be different for each blade 15. However, even in this case, the feature points 22 of the other blades 15 are corrected so that the shape of the other blades 15 is the same as the shape of the largest blade 15 among the multiple blades 15 whose shapes are estimated from the feature points 22. This makes it possible to prevent the movement course of an aircraft, such as a drone used as inspection equipment, from becoming too close to the other blades 15 when the movement course of the aircraft used to inspect the blades 15 is determined based on the feature points 22.
[0028] The above embodiment can also be modified as follows, for example. The above embodiment and the following modifications can be combined and implemented to the extent that they do not contradict each other technically. Feature point correction processing is not always necessary.
[0029] While the blades 15 of a wind turbine 16 were used as an example of a structure whose shape is to be estimated, the shapes of other structures, such as radio towers, bridge piers, and chimneys, may also be estimated. [Explanation of Symbols]
[0030] 11… Computer 12…Central Processing Unit 13...Storage section 14… Communications Department 15…Blade 16...Windmill 17...ground 18…post 19… Nasser 20…Rotation axis 21…Subdivision area 22…Features
Claims
1. The system includes a control unit that extracts numerous feature points along the external contour of a structure based on the three-dimensional coordinates of each point in a three-dimensional point cloud representing the structure and its surroundings, and estimates the shape of the structure based on these feature points. The control unit performs a point cloud extraction process, and then performs a feature point extraction process. The point cloud extraction process extracts points representing the external contour of the structure from the three-dimensional point cloud representing the structure and its surroundings. The feature point extraction process involves dividing each point representing the external outline of the structure into a number of subdivided regions, determining the centroid of each point within one of the subdivided regions, extracting the point closest to the centroid among the points as the feature point corresponding to a location within the subdivided region in the external outline of the structure, and performing such feature point extraction for each of the subdivided regions.
2. The aforementioned structure consists of multiple blades of the same shape in a wind turbine, The control unit performs the feature point correction process after performing the feature point extraction process. The feature point extraction process extracts feature points for each of the subdivided regions corresponding to the external contours of each of the multiple blades, The structural shape estimation device according to claim 1, wherein the feature point correction process estimates the shape of each of the multiple blades based on the feature points extracted by the feature point extraction process, and corrects the feature points of the other blades so that the shapes of the other blades are the same as the shape of the largest blade among the multiple blades.
3. As a process to extract a large number of feature points along the external contour of a structure based on a three-dimensional point cloud representing the structure and its surroundings, the computer is made to sequentially perform point cloud extraction and feature point extraction processes, and then the computer is made to estimate the shape of the structure based on the large number of extracted feature points. The point cloud extraction process extracts points representing the external contour of the structure from the three-dimensional point cloud representing the structure and its surroundings. The feature point extraction process is a structural shape estimation program that divides each point representing the external outline of the structure into a number of subdivided regions, finds the centroid of each point located within one of the subdivided regions, extracts the point closest to the centroid among the points as a feature point corresponding to a location within the subdivided region in the external outline of the structure, and performs such feature point extraction for each of the subdivided regions.
4. The aforementioned structure consists of multiple blades of the same shape in a wind turbine, The aforementioned computer performs the feature point extraction process, and then performs the feature point correction process. The feature point extraction process extracts feature points for each of the subdivided regions corresponding to the external contours of each of the multiple blades, The feature point correction process estimates the shape of each of the multiple blades based on the feature points extracted by the feature point extraction process, and corrects the feature points of the other blades so that the shapes of the other blades are the same as the shape of the largest blade among the multiple blades, according to claim 3, for the structural shape estimation program.