Point cloud plane identification and edge detection method

Abstract

The application discloses a point cloud plane identification and edge detection method. The method mainly comprises a new multi-seed synchronous growing algorithm for identifying a plane in a complex background and a circular edge extraction algorithm considering a mixed point cloud. Compared with a traditional region growing algorithm, the multi-seed synchronous growing algorithm has better robustness in initial point selection, and the calculation efficiency is improved by nearly 5 times compared with the traditional region growing algorithm. In view of the problem that many parameters are manually selected in a point cloud processing algorithm, the entropy threshold of the algorithm is derived through the thickness of the point cloud, so that the influence of manual selection on the calculation result is reduced.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of structural health monitoring and measurement, and more particularly relates to a point cloud plane identification and edge detection method considering mixed point clouds. BACKGROUND

[0002] It is a common construction operation to evaluate the quality of embedded parts before pouring concrete. If the positioning or size of the embedded parts deviates, it will cause problems in the later equipment installation, so it is necessary to position and check the size of all embedded parts according to the drawings. At present, the quality evaluation of embedded parts mainly relies on manual detection. The manual detection method is time-consuming and laborious, and the detection effect is not accurate enough. Therefore, it is necessary to provide a solution that can accurately and efficiently evaluate the quality of embedded parts.

[0003] As a non-contact measurement method, laser scanning technology has high measurement accuracy and efficiency. At present, laser scanning technology is gradually applied to engineering quality detection, three-dimensional model reconstruction, structural health monitoring and construction precision tracking. Based on three-dimensional laser scanning technology, the spatial coordinates, color information and other visual information of the structure can be obtained more quickly and accurately, thereby effectively monitoring the service information of the structure and reducing the serious dependence on human interaction and lighting conditions. Based on the traditional region growing algorithm, over-segmentation phenomenon and low single-seed growth efficiency are prone to occur, and too many parameters need to be manually intervened, such as the number of neighborhood points, the curvature threshold, the normal vector angle threshold, etc. In the calculation of geometric information, some edge information is indeed not considered due to mixed pixels. SUMMARY

[0004] In view of the problems of over-segmentation and too many manual parameters existing in the traditional region growing algorithm, the present application proposes a new point cloud plane and geometric information calculation method. The simultaneous growth of multi-seeds (SGM) algorithm proposed in the present application has the advantage of simultaneous growth of multiple planes, which can detect multiple target planes at the same time and greatly improve the plane detection efficiency. When calculating the target edge, a circular target edge detection algorithm considering mixed point clouds is proposed, which is more accurate than the traditional algorithm. The geometric information calculation method proposed in the present application can efficiently and accurately identify the geometric dimensions of the structure, and is an economical and efficient structure dimension measurement method, which has a wide application prospect in actual bridge performance evaluation.

[0005] The present application adopts the following technical solutions to solve the above technical problems:

[0006] The present application provides a point cloud plane identification and edge detection method considering mixed point clouds, which comprises the following steps:

[0007] S1. A standing laser scanner acquires point cloud information of the embedded plate on the wall surface before pouring at the nuclear power construction site, and then transmits it to a computer in real time for data processing.

[0008] S2. The target plane is obtained through multiple iterations using the Multi-Subsynchronous Growth (SGM) algorithm.

[0009] S3. The framework for calculating target geometry information acquires all information about the target plate, including center, side length / radius, and deflection;

[0010] S4. Register the construction coordinate system and the 3D scanning coordinate system to the same coordinate system, calculate the deviation of all geometric information, and provide a reference for the actual construction and acceptance of the project.

[0011] As a further technical solution of the present invention, the standing laser scanner in S1 is the Austrian RIEGLVZ-400i remote three-dimensional laser scanner.

[0012] Point cloud data of embedded plates on the wall surface at the construction site were collected using a standing scanner. The collected data was downsampled to 20%. The downsampling method adopted a high-pass graph filtering method, which has the advantages of fast speed and high sampling rate.

[0013] A novel strategy, M-estimator Sampling Consensus (MSAC), is proposed for selecting initial points. The basic idea is to use iterative weighted least squares estimation to estimate regression coefficients, and then determine the weights of each point based on the magnitude of the regression residuals, thereby achieving robustness. Assuming the existing point cloud set is {P}, first, a point p is selected from point cloud P. i ∈P, use the K-nearest neighbor algorithm to find p i The algorithm iterates through K points surrounding a given point and performs a plane fitting. Since the RANSAC method is easily affected by parameters and the least squares method struggles to effectively eliminate outliers, this invention introduces the M-estimation algorithm to address these issues. Robust M-estimation eliminates the influence of outliers on the model parameter estimation results by adaptively assigning different weights to the samples.

[0014] Accurate point cloud segmentation is achieved through an iterative growth strategy. First, calculate the K points surrounding each initial point. Then, use the normal vectors of the K points surrounding the initial point and the normal vector of the point to calculate the entropy value and the angle between the K points and the point to filter out ineligible points and delete them, completing the first iteration. Repeat the above process until the number of newly added points after deleting ineligible angles is 0, completing the growth.

[0015] Preferably, the Multi-Subsynchronous Growth (SGM) algorithm, compared with traditional point cloud region growing algorithms, has the following characteristics: Regarding parameters, the algorithm proposed in this invention only needs to set the number of nearest neighbors K for all points, while the region growing algorithm, in addition to calculating K, also requires a threshold for the angle between normal vectors and a curvature threshold, and the selection of these two thresholds is highly dependent on experience. Regarding algorithm robustness, the algorithm proposed in this invention has strong robustness, while the region growing algorithm suffers from oversegmentation due to improper parameter selection.

[0016] The planar identification method for multiple sub-synchronous growth in S2 is as follows:

[0017] S21. Collect point cloud data of embedded plates on the wall surface at the construction site using a standing scanner, and downsample the collected data to 20%.

[0018] S22. Find the K points around all points, use MSAC to divide the K points into inner points and outer points, then restore downsampling, and use the points with 0 inner points corresponding to all points as the initial points of multiple sub-synchronous growth algorithms. This completes the selection of the initial points.

[0019] S23. Calculate the K points around the initial point, then use the normal vectors of the K points around the initial point and the normal vector of the point to calculate the entropy value and the angle between the K points and the point to filter out the Ineligible points and delete them. This completes the first iteration. Repeat the above process until the number of newly added points after deleting the Ineligible angles is 0, thus completing the growth.

[0020] The entropy value and included angle are derived from the point cloud thickness caused by the overall error, specifically as follows:

[0021] S31. Calculate the mean and variance of the point cloud, approximate the point cloud distribution as a Gaussian distribution, take 6σ as the thickness of the point cloud, and take e = 3σ as p. i =(x i ,y i ,z i The single-point accuracy error;

[0022] S32. Assuming all neighboring points lie on an error-free tangent plane, the fitted tangent plane T(X) has a normal vector of V0. Since each point has a positional accuracy, the maximum positional error e is taken as p. i At the nearest point p i Within the interval (0,e), the tangent plane with the largest error is T(X′), and the normal vector of this plane is V′; in the tangent plane T(X), assume that the point farthest from the center of the nearest neighboring point is P. i If X'' is the tangent plane with the largest error, then the angle between the tangent plane T(X) with the largest error and the tangent plane T(X) without error is θ; calculate the angle between the tangent plane with the largest error and the tangent plane without error.

[0023] S33, Let the nearest point be p. i =(x i ,y i ,z i Given the positioning accuracy e, calculate the local entropy in that area. in

[0024] S34. If the information entropy satisfies: H C (θ k If H = log(K), then the neighboring region is a plane; due to the influence of scanning accuracy, if H C (θ k If ) ≠ log(K), then the error can be multiplied by 2. As its limit, and taking the local entropy log(m′) as the initial reference, if point p i Local entropy Relative to log(m′) satisfies Delete the point if the value is not found, otherwise keep the point. i Then, the H closest to the local entropy log(m′) pi As a benchmark.

[0025] In S3, all information about the target plate is obtained based on the computational target geometric information framework, specifically:

[0026] S41. Cluster all target planar point clouds based on Gaussian mixture model;

[0027] S42. Using an edge detection method that considers mixed points, the edges of straight lines and circular targets are obtained respectively, and the geometric information of all targets is calculated.

[0028] S43. Finally, register the 3D scanning coordinate system to the construction coordinate system and compare the difference between the construction information and the design information.

[0029] The simplified method in S42 for directly determining the boundary points on the surface through the distribution of neighboring points is as follows:

[0030] S421. Considering that the surface of a PC is mostly flat, we can first use the PCA algorithm for dimensionality reduction.

[0031] S422. Divide the adjacent points of each detection point in the input data into 8 regions, and define the points with at least one blank region as boundary points;

[0032] S423. Divide each center point into eight regions using four straight lines, and use the center points of at least two regions that do not contain NP as boundary points;

[0033] S424. Based on the inner boundary points obtained in S412 and S413, select and add points in the PC data that are close to the circular hole, and estimate the size of the circular hole based on the iterative correction method of the extracted center and radius.

[0034] In S43, the scanning coordinate system is transformed to the construction coordinate system using the ICP algorithm, specifically as follows:

[0035] S431. Calculating the center coordinates of the design values ​​and the center coordinates of the identification can obtain the three-dimensional position information deviation of all target plates;

[0036] S432. Calculate the normal vector of all target boards in the positive x-axis direction;

[0037] S433. Project the normal vector onto the yoz plane and calculate the angle between the projected vector and the vector perpendicular to the xoy plane.

[0038] According to another aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the point cloud plane recognition and edge detection method considering mixed point clouds of the present invention.

[0039] According to another aspect of the present invention, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the point cloud plane recognition and edge detection method considering mixed point clouds of the present invention.

[0040] Compared with the prior art, the present invention has at least the following beneficial effects:

[0041] (1) Compared with traditional region growing algorithms, the multi-subjoint growing algorithm proposed in this invention has better robustness in initial point selection, and all initial points can be iterated simultaneously, improving computational efficiency by nearly 5 times and effectively avoiding problems such as over-segmentation and low efficiency of single seed growth.

[0042] (2) The entropy threshold of the proposed algorithm was derived by the point cloud thickness, which effectively reduced the impact of manually selected parameters on the calculation results;

[0043] (3) The present invention proposes a circular edge extraction algorithm that considers mixed point clouds. The measurement results are closer to the theoretical values ​​and the accuracy is higher. It has broad application prospects in practical engineering structure monitoring. Attached Figure Description

[0044] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments will be briefly described below. Obviously, the drawings described below only relate to some embodiments of the present invention and are not intended to limit the present invention.

[0045] Figure 1 This is a flowchart of the method of the present invention;

[0046] Figure 2 This is a schematic diagram illustrating the iterative process of the multi-sub-joint growth algorithm of this invention;

[0047] Figure 3 This invention considers the principle of boundary separation of mixed point clouds;

[0048] Figure 4 This is a diagram illustrating the boundary point determination principle of the present invention: Figure 4 (a) is the center point; Figure 4 (b) is the mixing point; Figure 4 (c) is the boundary point;

[0049] Figure 5 This is a diagram showing the geometric information measurement error results in an embodiment of the present invention. Detailed Implementation

[0050] like Figures 1-5 As shown:

[0051] Example 1:

[0052] This invention provides a point cloud plane recognition and edge detection method based on a joint growth algorithm, such as... Figure 1 As shown,

[0053] Step 1: A standing laser scanner acquires point cloud information of the embedded plate on the wall surface before pouring at the nuclear power plant construction site, and then transmits it to a computer in real time for data processing.

[0054] It should be further noted that the standing laser scanner in step 1 is the Austrian RIEGL VZ-400i remote 3D laser scanner.

[0055] Step 2: The target plane is obtained through multiple iterations based on the various sub-synchronous growth algorithms proposed in this invention;

[0056] like Figure 1 As shown, the various sub-synchronous growth algorithms proposed in this invention mainly include:

[0057] 1) The point cloud data of the wall embedded plate at the construction site is collected by a standing scanner. The collected data is downsampled to 20%. The downsampling adopts the high-pass graph filtering method, which has the advantages of fast speed and high sampling rate.

[0058] 2) A novel strategy for selecting initial points is proposed: M-estimator Sampling Consensus (MSAC). The basic idea is to use iterative weighted least squares estimation to estimate regression coefficients, and determine the weights of each point based on the magnitude of the regression residuals, thereby achieving robustness. Assuming the existing point cloud set is {P}, first select a point p from the point cloud P. i ∈P, use the K-nearest neighbor algorithm to find p i The algorithm iterates through K points surrounding a given point and performs a plane fitting. Since the RANSAC method is easily affected by parameters and the least squares method struggles to effectively eliminate outliers, this invention introduces the M-estimation algorithm to address these issues. Robust M-estimation eliminates the influence of outliers on the model parameter estimation results by adaptively assigning different weights to the samples.

[0059] The M-estimation is generally defined as:

[0060]

[0061] In the formula, n is the total number of points in the plane, and ρ is the objective function, which is also the residual of the loss function; It is the residual term; It is a robust scaling estimate of the residuals, which depends on the unknown regression coefficient β.

[0062] The fitted plane equation is:

[0063] ax + by + cz + d = 0(2)

[0064] MSAC compensates for the impact of parameter selection and can divide the K points into interior points and O points. pi When calculating the fitted plane for all points, the points corresponding to the outlier of 0 are used as the initial point set {I}, and the remaining point set is {N}, i.e.

[0065]

[0066] Where, p i The normal vector is

[0067] 3) Achieve accurate point cloud segmentation through an iterative growth strategy. First, calculate the K points surrounding each initial point. Then, use the normal vectors of the K points surrounding the initial point and the normal vector of the point to calculate the entropy value. Use the angle between the K points and the normal vector to filter out ineligible points and delete them, completing the first iteration. Repeat this process until the number of newly added points after deleting ineligible angles is 0, completing the growth. The iterative process is as follows: Figure 2 As shown:

[0068] First, calculate all the initial seed points I. i =(xi ,y i ,z i ) and its corresponding normal vector Then add all seed points to the target set {T} and select all I... i M B's surrounding the point and not in set T m For points m ∈ {1, 2, ..., M}, finally use MSAC to calculate all B. m Normal vector of a point

[0069]

[0070]

[0071] The normal vector of point P is obtained by taking the plane normal vectors of the K points surrounding point P. When calculating the normal vector, assuming there are k interior points I and Kk exterior points O, the normal vector of point P is calculated as the true value based on the geometry formed by the interior points. After calculating the normal vector of P using the interior points, the normal vectors of all interior points are calculated using the same method. Then, the angles between the normal vectors of P and all the normal vectors of the k surrounding interior points and the standard plane are calculated.

[0072]

[0073] in:

[0074]

[0075]

[0076] In the process of achieving accurate point cloud segmentation through iterative growth strategy described in 3) above, this invention utilizes the point cloud thickness caused by comprehensive errors to derive the entropy value and the included angle, reducing the influence of manually selected parameters on the calculation results. Specifically:

[0077] 1) Calculate the mean and variance of the point cloud, approximate the point cloud distribution as a Gaussian distribution, take 6σ as the thickness of the point cloud, and take e = 3σ as p. i =(x i ,y i ,z i The single-point accuracy error.

[0078] 2) Assuming all neighboring points lie on the error-free tangent plane, the fitted tangent plane T(X) has a normal vector of V0. Since each point has a positional accuracy, the maximum positional error e is taken. i At the nearest point p i Within the interval (0, e), the tangent plane with the largest error is T(X′), and the normal vector of this plane is V′. In the tangent plane T(X), assume that the point farthest from the center of a neighboring point is P.i If X'' is the tangent plane with the largest error, then the angle between the tangent plane T(X) with the smallest error and the tangent plane T(X) without error is θ. Calculate the angle between the tangent plane with the largest error and the tangent plane without error.

[0079] 3) Assume a neighboring point p i =(x i ,y i ,z i Given the positioning accuracy e, calculate the local entropy in that area.

[0080] in

[0081] 4) If the information entropy satisfies: H C (θ k If H = log(K), then the neighboring region is a plane. Due to the influence of scanning accuracy, if H... C (θ k If ) ≠ log(K), then the error can be multiplied by 2. As its limit, and taking the local entropy log(m′) as the initial reference, if point p i Local entropy Relative to log(m′) satisfies Delete the point if the value is not found, otherwise keep the point. i Then, the H closest to the local entropy log(m′) pi As a benchmark.

[0082] Preferably, the Multi-Subsynchronous Growth (SGM) algorithm, compared with traditional point cloud region growing algorithms, has the following characteristics: Regarding parameters, the algorithm proposed in this invention only needs to set the number of nearest neighbors K for all points, while the region growing algorithm, in addition to calculating K, also requires a threshold for the angle between normal vectors and a curvature threshold, and the selection of these two thresholds is highly dependent on experience. Regarding algorithm robustness, the algorithm proposed in this invention has strong robustness, while the region growing algorithm suffers from oversegmentation due to improper parameter selection.

[0083] Step 3: Based on the framework for calculating the geometric information of the target proposed in this invention, obtain all information of the target plate, including center, side length (radius), and deflection.

[0084] 1) Clustering of all target planar point clouds based on Gaussian mixture models. Each Gaussian mixture model consists of N Gaussian distributions, each Gaussian distribution is called a cluster, and these Gaussian distributions together constitute the probability density function of the Gaussian mixture model:

[0085]

[0086]

[0087] In the formula, N is the number of models; π n This represents the weighting coefficient, which signifies the probability of each cluster being selected. N(x|μ n ,Σ n ) is the Gaussian distribution density This represents the square of the nth standard deviation.

[0088] The nth sub-model can be represented as:

[0089]

[0090] Assuming there are K collected sample data points, which can be assumed to be generated by a Gaussian distribution, the likelihood function of the GMM can be expressed as:

[0091]

[0092] Since the maximum value cannot be obtained directly, the EM algorithm is used to find the result iteratively.

[0093] 2) Using a circular edge extraction algorithm that considers mixed point clouds, the edges of straight lines and circular targets are obtained separately, and the side lengths (radii), centers, and corners of all targets are calculated. A simplified method for directly determining the boundary points on the surface through the distribution of neighboring points is as follows:

[0094] A) Considering that the surface of a PC is mostly flat, we can first use the PCA algorithm for dimensionality reduction.

[0095] B) such as Figure 3 As shown, the original point cloud model is divided into two parts by using a boundary extraction algorithm that considers the mixed point cloud. The true edge can be described more accurately by the boundary with the mixed point cloud on the outer ring.

[0096] B) such as Figure 4 As shown, the neighboring points of each detection point in the input data are divided into 8 regions, and points with at least one blank region are defined as boundary points.

[0097] C) Divide each center point into eight regions using four straight lines, and use the center points of at least two regions that do not have point clouds as boundary points.

[0098] D) Based on the inner boundary points obtained above, select and add points in the PC data that are close to the circular hole, and estimate the size of the circular hole based on the iterative correction method of the extracted center and radius.

[0099] Step 4: Register the construction coordinate system and the 3D scanning coordinate system to the same coordinate system, calculate the deviation of all geometric information, and provide a reference for the actual construction and acceptance of the project.

[0100] It should be further explained that this method mainly uses the ICP algorithm to transform the scanning coordinate system to the construction coordinate system, specifically:

[0101] A) Calculating the center coordinates of the design values ​​and the center coordinates of the identification can yield the three-dimensional positional information deviations of all target plates;

[0102] B) Calculate the normal vector of all target boards in the positive x-axis direction. Since the deviation of the rotation angle is generally no more than 15°, it is only necessary to calculate the acute angle between the two normal vectors.

[0103] C) Project the normal vector onto the yoz plane and calculate the angle between the projected vector and the vector perpendicular to the xoy plane. Only acute angles need to be considered.

[0104] Example

[0105] Taking a wall during the construction phase as an example, the specific implementation process of this invention is illustrated. A vertical scanner is used to scan the target wall and establish a three-dimensional model, further verifying the feasibility of the multi-sub-synchronous growth algorithm and geometric information calculation framework proposed in this invention.

[0106] Step 1: A standing laser scanner acquires point cloud information of the embedded plate on the wall surface before pouring at the nuclear power plant construction site, and then transmits it to a computer in real time for data processing.

[0107] Step 2 involves iterating multiple times using the Multi-Subsynchronous Growth (SGM) algorithm proposed in this invention to obtain the target plane. The target region only requires 4 iterations to cover the entire plane, and all 11 targets in the figure are identified.

[0108] Step 3: Based on the framework for calculating target geometry proposed in this invention, obtain all information about the target plate, including center, side length (radius), and deflection. First, classify all target regions using a Gaussian mixture model. After obtaining all targets, perform plane fitting using MSAC. For circular targets, if no mixing points are generated during scanning, directly use the outermost point cloud of the circular region for MSAC circle fitting to obtain the center and radius. If mixing points are generated during scanning, first use MSAC to calculate the fitting plane for the circle, which will divide all points into inner and outer points. Inner points are used for plane fitting, and outer points can be understood as noise, i.e., the mixed point cloud. Project all outer points generated by the MSAC algorithm onto this plane, excluding points whose projections fall within the circular region. Then calculate the innermost point of all projected points and use MSAC circle fitting to obtain the center and radius. The same applies to square targets, and will not be elaborated further. The error between considering and not considering the mixed point cloud and the design value is as follows:Figure 4 As shown. From Figure 5 As can be seen from the comparison between the algorithm of this invention and the algorithm that does not consider the mixing points, the algorithm proposed in this invention has higher calculation accuracy.

[0109] Step 4: Register the construction coordinate system and the 3D scanning coordinate system to the same coordinate system, calculate the deviation of all geometric information, and provide a reference for the actual construction and acceptance of the project.

[0110] Example 2:

[0111] The computer-readable storage medium of this embodiment stores a computer program that, when executed by a processor, implements the steps in the point cloud plane recognition and edge detection method of Embodiment 1.

[0112] The computer-readable storage medium in this embodiment can be an internal storage unit of the terminal, such as the terminal's hard disk or memory; the computer-readable storage medium in this embodiment can also be an external storage device of the terminal, such as a plug-in hard disk, smart memory card, secure digital card, flash memory card, etc. equipped on the terminal; furthermore, the computer-readable storage medium can include both the terminal's internal storage unit and external storage devices.

[0113] The computer-readable storage medium of this embodiment is used to store computer programs and other programs and data required by the terminal. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0114] Example 3:

[0115] The computer device of this embodiment includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the point cloud plane recognition and edge detection method of Embodiment 1.

[0116] In this embodiment, the processor can be a central processing unit, or other general-purpose processors, digital signal processors, application-specific integrated circuits, off-the-shelf programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc. The memory can include read-only memory and random access memory, and provides instructions and data to the processor. A portion of the memory can also include non-volatile random access memory. For example, the memory can also store device type information.

[0117] Those skilled in the art will understand that the content disclosed in the embodiments can be provided as a method, system, or computer program product. Therefore, this solution can take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this solution can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage) containing computer-usable program code.

[0118] This solution is described with reference to flowchart illustrations and / or block diagrams of methods and computer program products according to embodiments of this solution. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0119] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0120] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0121] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0122] The examples described herein are merely preferred embodiments of the invention and are not intended to limit the concept and scope of the invention. Any modifications and improvements made by those skilled in the art to the technical solutions of the invention without departing from the design concept of the invention should fall within the protection scope of the invention.

Claims

1. A method for point cloud planar recognition and edge detection, characterized in that, The method includes the following steps: S1. A standing laser scanner acquires point cloud information of the embedded plate on the wall surface before pouring at the nuclear power construction site, and then transmits it to a computer in real time for data processing. S2. Obtain the target plane through multiple iterations using various sub-synchronous growth algorithms; S3. The framework for calculating target geometry information acquires all information about the target plate, including center, side length / radius, and deflection; S4. Register the construction coordinate system and the 3D scanning coordinate system to the same coordinate system, calculate the deviation of all geometric information, and provide a reference for the actual construction and acceptance of the project. The planar identification method for multiple sub-synchronous growth in S2 is as follows: S21. Collect point cloud data of embedded plates on the wall surface at the construction site using a standing scanner, and downsample the collected data to 20%. S22. Find the K points around all points, use MSAC to divide the K points into inner points and outer points, then restore downsampling, and use the points with 0 inner points corresponding to all points as the initial points of multiple sub-synchronous growth algorithms. This completes the selection of the initial points. S23. Calculate the entropy value of the normal vector of each of the K points around the initial point and the entropy value of the normal vector. Use the entropy value of the normal vector and the threshold of the angle between the normal vectors to filter out points that do not meet the rules and delete them. Complete the first iteration. Repeat the iteration process until the number of newly added points is 0, and the growth is completed. In S3, all information about the target plate is obtained based on the computational target geometric information framework, specifically: S41. Cluster all target planar point clouds based on Gaussian mixture model; S42. Using an edge detection method that considers external points, obtain the edges of straight lines and circular targets respectively, and calculate the geometric information of all targets; S43. Finally, register the 3D scanning coordinate system to the construction coordinate system and compare the difference between the construction information and the design information. The entropy value and included angle are derived from the point cloud thickness caused by the overall error, specifically as follows: S31. Calculate the mean and variance of the point cloud, and convert the point cloud distribution to a Gaussian distribution. As the thickness of the point cloud, As Single-point accuracy error; S32, All neighboring points are on the error-free tangent plane. Upper, tangent plane ; At nearby points Within the interval, the tangent plane with the largest error is obtained as follows: tangent plane The normal vector is ; in the tangent plane In this context, assume that the point farthest from the center of a neighboring point is... The tangent plane with the largest error With error-free tangent plane The included angle between them is ; Calculate the angle between the tangent plane with the largest error and the tangent plane without error. ; S33. Calculate local entropy ,in ; S34. If the information entropy satisfies: If the adjacent area is a plane, then the area is affected by the scanning accuracy. Then the error can be multiplied by 2. As its limit, and taking the local entropy as its limit value, As an initial reference, if point Local entropy Compared to satisfy Delete the point if it is not deleted, otherwise keep it. Then, the closest to the local entropy of As a benchmark; Point The number of normal vectors involved in entropy calculation within the neighborhood; express The corresponding local point cloud entropy value.

2. The method according to claim 1, characterized in that, The simplified method in S42 for directly determining the boundary points on the surface through the distribution of neighboring points is as follows: S421. Considering that the surface of PC is mostly flat, we first use the PCA algorithm for dimensionality reduction, where PC represents the point cloud; S422. Divide the adjacent points of each detection point in the input data into 8 regions, and define the points with at least one blank region as boundary points; S423. Divide each center point into eight regions using four straight lines. Use the center points of at least two regions that do not have NP as boundary points. NP represents a normal point. S424. Based on the obtained inner boundary points, select and add points in the PC data that are close to the circular hole, and estimate the size of the circular hole based on the iterative correction method of the extracted center and radius.

3. The method according to claim 2, characterized in that, In S43, the scanning coordinate system is transformed to the construction coordinate system using the ICP algorithm, specifically as follows: S431. Calculating the center coordinates of the design values ​​and the center coordinates of the identification can obtain the three-dimensional position information deviation of all target plates; S432. Calculate the normal vector of all target boards in the positive x-axis direction; S433, Project the normal vector onto... Plane, calculate the projection vector and the vector perpendicular to it. The angle between vectors in a plane.

4. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by the processor, it implements the steps in the point cloud plane recognition and edge detection method as described in any one of claims 1 to 3.

5. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the point cloud plane recognition and edge detection method as described in any one of claims 1 to 3.

Images