CN116012399A - Point cloud plane identification and edge detection method - Google Patents

Abstract

The invention discloses a point cloud plane recognition and edge detection method. The method mainly includes a new multi-seed synchronous growth algorithm for identifying planes in complex backgrounds and a circular edge extraction algorithm considering mixed point clouds. Compared with the traditional region growing algorithm, the multi-seed synchronous growing algorithm proposed by the present invention has better robustness in initial point selection, and the calculation efficiency is nearly 5 times higher than that of the traditional region growing algorithm. Aiming at the problem of manual selection of many parameters in the point cloud processing algorithm, the present invention deduces the entropy threshold of the proposed algorithm through the thickness of the point cloud, reducing the influence of manual selection on the calculation results.

Description

Point Cloud Plane Recognition and Edge Detection Method

technical field

The invention belongs to the field of structural health monitoring and measurement, and more specifically relates to a point cloud plane recognition and an edge detection method considering mixed point clouds.

Background technique

It is a common construction operation to evaluate the quality of embedded parts before concrete pouring. If the positioning or size of the embedded components deviates, it will cause problems in the later equipment installation. Therefore, it is very necessary to carry out the positioning and size acceptance of all embedded components according to the drawings. At present, the quality assessment of pre-embedded components mainly relies on manual inspection. The method based on manual detection is time-consuming and labor-intensive, and the detection effect is not accurate enough. Therefore, it is necessary to provide solutions that can accurately and efficiently evaluate the quality of pre-embedded components.

As a non-contact measurement method, laser scanning technology has high measurement accuracy and high measurement efficiency. At present, laser scanning technology is gradually used in engineering quality inspection, 3D model reconstruction, structural health monitoring and construction accuracy tracking. Based on 3D laser scanning technology, visual information such as spatial coordinates and color information of the structure can be obtained faster and more accurately, so as to effectively monitor the service information of the structure, and reduce the heavy dependence on human interaction and lighting conditions. Based on the traditional region growing algorithm, it is prone to over-segmentation and low single-seed growth efficiency, and there are too many parameters that require manual intervention, such as the number of neighborhood points, curvature threshold, normal vector angle threshold, etc. In the calculation of geometric information, the problem of partial edge information is also caused by not considering the mixed pixels.

Contents of the invention

Based on the problems of over-segmentation and many artificial parameters in the traditional region growing algorithm, the present invention proposes a new point cloud plane and calculation geometric information method. The multi-seed joint growth algorithm (Simultaneous Growth of Multi-seeds, SGM) proposed by the present invention has the advantage of multi-plane growth at the same time, can detect multiple target planes at the same time, and greatly improves the plane detection efficiency. When calculating the object edge, a circular object edge detection algorithm considering the mixed point cloud is proposed, which is more accurate than the traditional algorithm. The calculation geometric information method proposed by the present invention can identify information such as structural geometric dimensions with high efficiency and high precision, is an economical and efficient structural dimension measurement method, and has the prospect of being widely used in actual bridge performance evaluation.

The present invention adopts the following technical solutions for solving the problems of the technologies described above:

The invention provides a point cloud plane recognition and an edge detection method considering a mixed point cloud, the method comprising the following steps:

S1. The standing laser scanner obtains the point cloud information of the embedded slab on the wall before pouring at the nuclear power construction site, and then transmits it to the computer in real time for data processing;

S2, obtaining the target plane through multiple iterations of the multi-seed synchronous growth (SGM) algorithm;

S3. The framework for calculating the geometric information of the target obtains all the information of the target board, including center, side length/radius and deflection;

S4. Register the construction coordinate system and the three-dimensional scanning coordinate system to the same coordinate system, calculate the deviation of all geometric information, and provide reference for actual engineering construction acceptance.

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

Based on the standing scanner to collect the point cloud data of the pre-embedded panels on the wall of the construction site, the collected data is down-sampled to 20%, and the down-sampling adopts the high-pass image filtering method. The down-sampling method has the advantages of fast speed and high sampling rate ;

A new strategy of M-estimator SAmple Consensus (MSAC) is proposed to select the initial point. The basic idea is to use iterative weighted least squares to estimate the regression coefficient, and determine the weight of each point according to the size of the regression residual, so as to achieve the purpose of robustness. Assuming that the existing point cloud collection is {P}, first select a point p _i ∈ P in the point cloud P, use the K nearest neighbor algorithm to find K points around point p _i , and perform plane fitting. Since the RANSAC method is easily affected by parameters and the least square method is difficult to effectively eliminate the influence of abnormal points, the present invention introduces an M estimation algorithm to solve the above problems. Robust M estimation eliminates the influence of outliers on model parameter estimation results by adaptively assigning different weights to samples.

Accurate segmentation of point clouds is achieved through an iterative growing strategy. First calculate all the K points around the initial point, then use the normal vector corresponding to the K points around the initial point and the normal vector to calculate the entropy value and the angle between the K points and the Ineligible to filter and delete, complete the first iteration, repeat The above process until the Ineligible angle is deleted after the iteration, the newly added points are 0, and the growth is completed.

Preferably, the multi-seed synchronous growth (SGM) algorithm is compared with the traditional point cloud region growing algorithm, and has the following characteristics: in terms of parameters, the algorithm proposed by the present invention only needs to set the number of neighbor points K of all points. , while the region growing algorithm not only needs to calculate K, but also needs angle threshold and curvature threshold between normal vectors, and the selection of these two thresholds depends on experience. In terms of algorithm robustness, the algorithm proposed by the present invention has strong robustness, but there is over-segmentation if the parameters of the region growing algorithm are improperly selected.

The plane identification method of multi-seed synchronous growth in S2, specifically:

S21. Based on the standing scanner, collect the point cloud data of the pre-embedded board on the wall of the construction site, and down-sample the collected data to 20%;

S22. Find K points around all points, use MSAC to divide K points into inner points and outer points, and then resume downsampling, and use the points whose inner points corresponding to all points are 0 as the initial stage of multi-seed synchronous growth algorithm point, so far the selection of the initial point is completed;

S23. Calculate all K points around the initial point, and then use the normal vector corresponding to the K points around the initial point and the normal vector to calculate the entropy value and the angle between the K points and the Ineligible to filter and delete, and complete the first iteration. Repeat the above process until the number of newly added points after the Ineligible angle is deleted after iteration is 0, and the growth is completed.

Use the point cloud thickness caused by the comprehensive error to derive the entropy value and included angle, specifically:

S31. Calculate the mean and variance of the point cloud, approximately distribute the point cloud into a Gaussian distribution, use 6σ as the thickness of the point cloud, and use e=3σ as the single-point precision error of p _i =(xi _, y _i , z _i ) ;

S32. Assuming that all the adjacent points are on the tangent plane without error, then the fitted tangent plane T(X), the normal vector of this plane is V ₀ . Since each point has a point accuracy, take the maximum point error e, p _i is within the range of the adjacent point p _i ∈ (0, e), and the tangent plane with the largest error is T(X′). The normal vector is V'; in the tangent plane T(X), assuming that the farthest point from the adjacent point to the center of the adjacent point is P _i ', then the tangent plane T(X') with the largest error and the tangent plane T(X') without error are The angle between X) is θ; the angle between the tangent plane with the largest calculation error and the tangent plane without error

S33. Set the positioning accuracy e of the adjacent point p _i = (xi _, y _i , zi ₎ , and calculate the local entropy in this field in

S34. If the information entropy satisfies: H _C (θ _k )=log(K), then the adjacent area is a plane; affected by the scanning accuracy, if H _C (θ _k )≠log(K), then 2 double error As its limit value, and the local entropy log(m′) as the initial reference, if the local entropy of point p _i With respect to log(m′) satisfies Then delete the point, otherwise, keep p _i , and then take H _pi closest to the local entropy log(m′) as the benchmark.

In S3, all information of the target board is obtained based on the calculation target geometric information framework, specifically:

S41. Clustering all target plane point clouds based on the Gaussian mixture model;

S42. Using an edge detection method considering mixed points, respectively acquire the edges of straight lines and circular objects, and calculate geometric information of all objects;

S43. Finally, register the three-dimensional scanning coordinate system to the construction coordinate system, and compare the difference between the construction information and the design information.

In S42, the simplified method of directly determining the boundary points on the surface through the distribution of adjacent points is specifically:

S421. Considering that the surface of the pc is mostly flat, the PCA algorithm may be first used for dimensionality reduction;

S422. Divide the adjacent points of each detection point in the input data into 8 areas, and the points with at least one blank area are defined as boundary points;

S423. Using four straight lines around each center point to divide it into eight areas, and using at least two center points without NP in the area as boundary points;

S424. On the basis of the inner boundary points obtained in S412 and S413, select and increase the points close to the circular hole in the PC data, and estimate the size of the circular hole based on an iterative correction method of the extracted circle center and radius.

In S43, the scanning coordinate system is converted to the construction coordinate system through the ICP algorithm, specifically:

S431. Calculate the center coordinates of the design value and the identified center coordinates to obtain the three-dimensional position information deviations of all target boards;

S432. Calculate the normal vectors of all identification target boards facing the positive direction of the x-axis;

S433. Project the normal vector onto the yoz plane, and calculate an angle between the projected vector and a vector perpendicular to the xoy plane.

According to another aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the point cloud plane recognition and the edge detection method considering the mixed point cloud of the present invention are realized A step of.

According to yet another aspect of the present invention, a computer device is provided, including a memory, a processor, and a computer program stored on the memory and operable on the processor, and the point of the present invention is realized when the processor executes the program Cloud plane identification and steps in an edge detection method that considers mixed point clouds.

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

(1) Compared with the traditional region growing algorithm, the multi-seed joint growth algorithm proposed by the present invention is more robust in initial point selection, and all initial points can be iterated at the same time, and the calculation efficiency is increased by nearly 5 times, Effectively avoid problems such as over-segmentation and low single-seed growth efficiency;

(2) The entropy threshold of the proposed algorithm is deduced through the point cloud thickness, which effectively reduces the influence of manual selection of parameters on the calculation results;

(3) The present invention proposes a circular edge extraction algorithm that considers mixed point clouds, the measurement results are closer to the theoretical value, the accuracy is higher, and it has a wide range of application prospects for actual engineering structure monitoring.

Description of drawings

In order to illustrate the technical solutions of the embodiments of the present invention more clearly, the accompanying drawings of the embodiments will be briefly introduced below. Obviously, the accompanying drawings in the following description only relate to some embodiments of the present invention, rather than limiting the present invention .

Fig. 1 is method flowchart of the present invention;

Fig. 2 is the iterative process schematic diagram of multi-seed joint growth algorithm of the present invention;

Fig. 3 is that the present invention considers the boundary separation principle of mixed point cloud;

Fig. 4 is the boundary point judgment principle figure of the present invention: Fig. 4 (a) is center point; Fig. 4 (b) is mixing point; Fig. 4 (c) is boundary point;

Fig. 5 is a result diagram of geometric information measurement error in the embodiment of the present invention

Detailed ways

As shown in Figure 1-5:

Example 1:

The present invention provides a point cloud plane recognition and edge detection method based on the joint growth algorithm, as shown in Figure 1,

Step 1. The standing laser scanner obtains the point cloud information of the embedded slab on the wall before pouring at the nuclear power construction site, and then transmits it to the computer in real time for data processing;

It should be further explained that the model of the standing laser scanner in step 1 is Austrian RIEGL VZ-400i remote 3D laser scanner.

Step 2, multiple iterations based on the multi-seed synchronous growth algorithm proposed by the present invention to obtain the target plane;

As shown in Figure 1, the multi-seed synchronous growth algorithm proposed by the present invention mainly includes:

1) Based on the standing scanner to collect the point cloud data of the pre-embedded panels on the wall of the construction site, the collected data is down-sampled to 20%, and the down-sampling adopts the high-pass image filtering method. This down-sampling method has the advantages of fast speed and high sampling rate advantage;

2) A new strategy of M-estimator SAmple Consensus (MSAC for short) is proposed to select the initial point. The basic idea is to use iterative weighted least squares to estimate the regression coefficient, and determine the weight of each point according to the size of the regression residual, so as to achieve the purpose of robustness. Assuming that the existing point cloud collection is {P}, first select a point p _i ∈ P in the point cloud P, use the K nearest neighbor algorithm to find K points around point p _i , and perform plane fitting. Since the RANSAC method is easily affected by parameters and the least square method is difficult to effectively eliminate the influence of abnormal points, the present invention introduces an M estimation algorithm to solve the above problems. Robust M estimation eliminates the influence of outliers on model parameter estimation results by adaptively assigning different weights to samples.

The M estimate is generally defined as:

where n is the total number of points in the plane, ρ is the objective function and the residual of the loss function; is the residual item; is a robust scaling estimate of the residuals, depending on the unknown regression coefficient β.

The fitted plane equation is:

ax+by+cz+d=0(2)

MSAC will make a compensation for the influence of parameter selection, and can divide K into interior points and O _pi . When calculating the fitting plane of all points, the point corresponding to the outer point is 0 as the initial point set {I}, and the rest of the point set is {N}, that is

Among them, the normal vector of p _i is

3) Accurate segmentation of point clouds is achieved through an iterative growing strategy. First calculate all the K points around the initial point, then use the normal vector corresponding to the K points around the initial point and the normal vector to calculate the entropy value and the angle between the K points and the Ineligible to filter and delete, complete the first iteration, repeat The above process until the Ineligible angle is deleted after the iteration, the newly added points are 0, and the growth is completed. The iterative process is shown in Figure 2:

First calculate all the initial seed points I _i = (xi _, y _i , z _i ) and their corresponding normal vectors Then add all seed points to the target set {T} and select M B _m ,m∈{1,2,...,M} points around all I _i points and not in the T set, and finally use MSAC to calculate all B _m points normal vector of

The plane normal vector passing through the K points around point P is used as the normal vector of point P. When calculating the normal vector, assuming that the number of interior points I is k and the number of exterior points O is K-k, then the normal vector of point P is considered to be the real value of the geometric calculation composed of interior points, and the normal vector of P is calculated through the interior points Finally, use the same method to calculate the normal vectors of all interior points, and then calculate the angle between P and all the normal vectors of the surrounding k interior points and the standard plane

in:

In the above 3) in the process of realizing accurate point cloud segmentation through iterative growth strategy, the present invention uses point cloud thickness caused by comprehensive error to derive entropy value and included angle, and reduces the influence of artificially selected parameters on calculation results. Specifically:

1) Calculate the mean and variance of the point cloud, approximately distribute the point cloud as a Gaussian distribution, use 6σ as the thickness of the point cloud, and use e=3σ as the single-point precision error of p _i =(xi _, y _i ,zi ₎ .

2) Assuming that all the adjacent points are on the tangent plane without error, then the fitted tangent plane T(X), the normal vector of this plane is V ₀ . Since each point has a point accuracy, take the maximum point error e, p _i is within the range of the adjacent point p _i ∈ (0, e), and the tangent plane with the largest error is T(X′). The normal vector is V'. In the tangent plane T(X), assuming that the farthest point from the adjacent point to the center of the adjacent point is P _i ′, then the gap between the tangent plane T(X′) with the largest error and the tangent plane T(X) without error The angle is θ. Calculate the angle between the tangent plane with the largest error and the tangent plane without error

3) Assuming the positioning accuracy e of the neighboring point p _i = (xi _, y _, zi ₎ , calculate the local entropy in this field

in

4) If the information entropy satisfies: H _C (θ _k )=log(K), then the adjacent area is a plane. Affected by the scanning accuracy, if H _C (θ _k )≠log(K), the error can be doubled As its limit value, and the local entropy log(m′) as the initial reference, if the local entropy of point p _i With respect to log(m′) satisfies Then delete the point, otherwise, keep p _i , and then take H _pi closest to the local entropy log(m′) as the benchmark.

Preferably, the multi-seed synchronous growth (SGM) algorithm is compared with the traditional point cloud region growing algorithm, and has the following characteristics: in terms of parameters, the algorithm proposed by the present invention only needs to set the number of neighbor points K of all points. , while the region growing algorithm not only needs to calculate K, but also needs angle threshold and curvature threshold between normal vectors, and the selection of these two thresholds depends on experience. In terms of algorithm robustness, the algorithm proposed by the present invention has strong robustness, but there is over-segmentation if the parameters of the region growing algorithm are improperly selected.

Step 3: Obtain all information of the target board based on the framework for calculating target geometric information proposed by the present invention, including center, side length (radius), and deflection.

1) Cluster all object plane point clouds based on a Gaussian mixture model. Each Gaussian mixture model is composed of four Gaussian distributions, and each Gaussian distribution is called a cluster. These Gaussian distributions are combined to form the probability density function of the Gaussian mixture model:

In the formula, N is the number of models; π _n represents the weight coefficient, which means the probability of each type of cluster being selected, and N(x|μ _n ,Σ _n ) is the Gaussian distribution density Indicates the square of the standard deviation of the nth class.

The nth sub-model can be expressed as:

Assuming that there are K sample data collected, it can be considered that these data points are generated by a certain Gaussian distribution, then the likelihood function of GMM can be expressed as:

Since the maximum value cannot be obtained directly, the EM algorithm is used to obtain the result iteratively.

2) Use the circular edge extraction algorithm considering the mixed point cloud to obtain the edges of straight lines and circular objects respectively, and calculate the geometric information such as the side length (radius), center, and corner of all objects. A simplified method to directly determine the boundary points on the surface through the distribution of neighboring points, specifically:

A) Considering that the surface of pc is mostly flat, PCA algorithm can be used for dimensionality reduction first.

B) As shown in Figure 3, the original point cloud model is divided into two parts by using the boundary extraction algorithm considering the mixed point cloud, and the real edge can be more accurately described by the boundary with the mixed point cloud in the outer circle.

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

C) Use four straight lines around each center point to divide into eight areas, and use at least two center points without point clouds in the area as boundary points.

D) On the basis of obtaining the inner boundary points above, select and increase the points close to the circular hole in the PC data, and estimate the size of the circular hole based on the iterative correction method of the extracted circle center and radius.

Step 4, register the construction coordinate system and the 3D scanning coordinate system in the same coordinate system, calculate the deviation of all geometric information, and provide reference for the actual project construction acceptance.

It needs to be further explained that this method mainly uses the ICP algorithm to convert the scanning coordinate system to the construction coordinate system, specifically:

A) Calculate the central coordinates of the design value and the identified central coordinates to obtain the three-dimensional position information deviation of all target boards;

B) Calculate the normal vectors of all recognition target boards facing the positive direction of the x-axis. Because the deviation of the rotation angle generally does not exceed 15°, it is only necessary to calculate the acute angle of the included angle between the two normal vectors.

C) Project the normal vector onto the yoz plane, calculate the angle between the projection vector and the vector perpendicular to the xoy plane, only need to consider the acute angle.

Example

Taking a certain wall in the construction stage as an example to illustrate the specific implementation process of the present invention, use a vertical scanner to scan the target wall, build a three-dimensional model, and further verify the feasibility of the multi-seed synchronous growth algorithm and geometric information calculation framework proposed by the present invention sex.

Step 1. The standing laser scanner obtains the point cloud information of the embedded slab on the wall before pouring at the nuclear power construction site, and then transmits it to the computer in real time for data processing;

In step 2, the target plane is obtained through multiple iterations based on the multi-seed simultaneous growth (SGM) algorithm proposed by the present invention. The target area only needs 4 iterations to cover the entire plane, and a total of 11 targets in the figure are all identified.

Step 3: Obtain all information of the target board based on the framework for calculating target geometric information proposed by the present invention, including center, side length (radius), and deflection. Firstly, the Gaussian mixture model is used to classify all target areas, and after all the targets are obtained, MSAC is used for plane fitting. For a circular target, if the scan does not produce a mixed point, then directly use the outermost point cloud of the circular area to perform MSAC circle fitting to obtain the center and radius of the circle; if the scan produces a mixed point, first use MSAC to calculate the approximate circle Fitting the plane will divide all points into inner points and outer points. The inner points are used for plane fitting, and the outer points can be understood as noise, that is, mixed point clouds. Project all the outer points generated by the MSAC algorithm to the plane, exclude the points that the projection falls in the circular area, and then calculate the innermost point of all projected points, and use MSAC circle fitting to get the center and radius of the circle. The same is true for the square target, which will not be repeated here. The error between the mixed point cloud and the design value without considering the mixed point cloud is shown in Fig. 4. It can be seen from FIG. 5 that the algorithm proposed by the present invention has higher calculation accuracy in comparison with the algorithm not considering the mixing point.

Step 4, register the construction coordinate system and the 3D scanning coordinate system in the same coordinate system, calculate the deviation of all geometric information, and provide reference for the actual project construction acceptance.

Example 2:

The computer-readable storage medium of this embodiment stores a computer program thereon, and when the program is executed by a processor, the steps in the point cloud plane recognition and edge detection method of Embodiment 1 are implemented.

The computer-readable storage medium in this embodiment may be an internal storage unit of the terminal, such as a hard disk or memory of the terminal; the computer-readable storage medium in this embodiment may also be an external storage device of the terminal, such as a plug-in Connectable hard disk, smart memory card, secure digital card, flash memory card, etc.; further, the computer-readable storage medium may also include both an internal storage unit of the terminal and an external storage device.

The computer-readable storage medium in this embodiment is used to store computer programs and other programs and data required by the terminal, and the computer-readable storage medium can also be used to temporarily store outputted or to-be-outputted data.

Example 3:

The computer equipment of this embodiment includes a memory, a processor, and a computer program stored on the memory and operable on the processor. When the processor executes the program, it realizes the point cloud plane recognition and edge detection method of Embodiment 1. in the steps.

In this embodiment, the processor may 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 the processor can also be any conventional processor, etc.; the memory can include read-only memory and random access memory, and provide instructions and data to the processor, part of the memory Non-volatile random access memory may also be included, for example, memory may also store device type information.

Those skilled in the art should understand that the content disclosed in the embodiments may be provided as methods, systems, or computer program products. Accordingly, the present solution can take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present aspect may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) having computer-usable program code embodied therein.

The solution is described with reference to the method according to the embodiment of the solution, and the flowchart and/or block diagram of the computer program product, it should be understood that each process and/or block in the flowchart and/or block diagram can be realized by computer program instructions , and flow charts and/or combinations of processes and/or blocks in block diagrams; these computer program instructions can be provided to processors of general purpose computers, special purpose computers, embedded processors or other programmable data processing devices to produce a machine , causing instructions executed by a processor of a computer or other programmable data processing equipment to generate means for realizing the functions specified in one or more procedures of the flow chart and/or one or more blocks of the block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to operate in a specific manner, such that the instructions stored in the computer-readable memory produce an article of manufacture comprising instruction means, the instructions The device realizes the function specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions can also be loaded onto a computer or other programmable data processing device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby The instructions provide steps for implementing the functions specified in the flow chart or blocks of the flowchart and/or the block or blocks of the block diagrams.

Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be implemented through computer programs to instruct related hardware, and the programs can be stored in a computer-readable storage medium. During execution, it may include the processes of the embodiments of the above-mentioned methods. Wherein, the storage medium may be a magnetic disk, an optical disk, a read-only memory (Read-Only Memory, ROM) or a random access memory (Random Access Memory, RAM) and the like.

The examples described in the present invention are only to describe the preferred implementation of the present invention, and are not intended to limit the concept and scope of the present invention. Variations and improvements should fall within the protection scope of the present invention.

Claims (8)

1. A method for point cloud plane identification and edge detection, the method comprising the steps of:

s1, acquiring point cloud information of a wall surface embedded plate before pouring in a nuclear power construction site by a standing type laser scanner, and then transmitting the point cloud information to a computer in real time for data processing;

s2, obtaining a target plane through multiple iterations of multiple sub-synchronous growth algorithms;

s3, calculating all information of the target plate, including center, side length/radius and deflection, of the frame of the target geometric information;

and S4, registering the construction coordinate system and the three-dimensional scanning coordinate system to the same coordinate system, calculating the deviation of all geometric information, and providing reference for practical engineering construction acceptance.

2. The method according to claim 1, wherein the plane recognition method for the synchronous growth of the plurality of sub-species in S2 comprises the following steps:

s21, collecting wall surface embedded plate point cloud data of a construction site based on a standing scanner, and downsampling the collected data to 20%;

s22, finding out K points around all points, dividing the K points into inner points and outer points by using MSAC, then restoring downsampling, taking the points with the number of 0 corresponding to all points as initial points of a plurality of sub-synchronous growth algorithms, and completing the selection of the initial points;

s23, calculating K points around all initial points, screening Ineligible by using normal vector quantity and normal vector quantity calculated by the normal vector quantity corresponding to the K points around the initial points and included angles between the K points and the normal vector quantity, deleting the Ineligible, completing first iteration, repeating the process until the number of newly added points is 0 after deleting the Ineligible angle after iteration, and completing growth.

3. The method according to claim 1, wherein the entropy and the angle are derived by using the thickness of the point cloud caused by the integrated error, in particular:

s31, calculating the mean value and variance of the point cloud, approximately distributing the point cloud into Gaussian distribution, taking 6σ as the thickness of the point cloud, and taking e=3σ as p _i ＝(x _i ,y _i ,z _i ) Single point precision error of (a);

s32, assuming that all adjacent points are on an error-free tangential plane, fitting the tangential plane T (X), wherein the normal vector of the plane is V ₀ . Because each point has the maximum value e, p of the point position error of point position accuracy _i At the adjacent point p _i In the range of the E (0, e) interval, obtaining a tangential plane with the largest error as T (X '), wherein the normal vector of the plane is V'; t (X) in the tangential plane, assuming that the point furthest from the adjacent point to the center of the adjacent point is P' _i The included angle between the tangential plane T (X') with the largest error and the tangential plane T (X) without error is theta; calculating the included angle between the tangential plane with the largest error and the tangential plane without error

S33, setting adjacent point p _i ＝(x _i ,y _i ,z _i ) Calculating the local entropy in the field

Wherein->

S34, if the information entropy satisfies: h _C (θ _k ) Log (K), then the neighborhood is planar; affected by scanning accuracy, if H _C (θ _k ) Not equal to log (K), 2 times the error

As its limit value and the local entropy log (m') as the initial reference, if point p _i Local entropy of->

Satisfy +.>

Delete the point, otherwise preserve p _i Then the nearest local entropy log (m') is added>

As a reference.

4. The method according to claim 1, wherein in S3 all information of the target plate is acquired based on the calculation target geometry information frame, specifically:

s41, clustering all target plane point clouds based on a Gaussian mixture model;

s42, respectively acquiring edges of the straight line and the round target by using an edge detection method considering the outer points, and calculating geometric information of all the targets;

and S43, finally registering the three-dimensional scanning coordinate system to a construction coordinate system, and comparing the difference value of the construction information and the design information.

5. The method according to claim 4, wherein the simplified method of directly determining boundary points on the surface by the distribution of neighboring points in S42 is as follows:

s421, considering that the surface of pc is mostly flat, the PCA algorithm can be used for dimension reduction firstly;

s422, dividing adjacent points of each detection point in the input data into 8 areas, wherein a point with at least one blank area is defined as a boundary point;

s423, dividing the periphery of each central point into eight areas by using four straight lines, and taking the central point without NP in at least two areas as a boundary point;

s424, on the basis of the inner boundary points obtained in S412 and S413, selecting and adding points close to the circular holes in the PC data, and estimating the sizes of the circular holes based on an iterative correction method of the extracted circle centers and the radii.

6. The method according to claim 4, wherein the scanning coordinate system is converted into the construction coordinate system by ICP algorithm in S43, in particular:

s431, calculating the center coordinates of the design values and the recognized center coordinates to obtain the three-dimensional position information deviation of all target plates;

s432, calculating normal vectors of all the identification target plates in the positive direction of the x axis;

s433, projecting the normal vector to a yoz plane, and calculating the vector included angle between the projection vector and the vector perpendicular to the xoy plane.

7. A computer-readable storage medium having stored thereon a computer program, characterized by: the program when executed by a processor implements the steps in the point cloud plane identification and edge detection method as claimed in any one of claims 1 to 5.

8. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps in the point cloud plane identification and edge detection method according to any of claims 1-5 when the program is executed.

Images