CN113902688A - Non-uniform point cloud surface area segmentation method based on improved region growing method - Google Patents

Non-uniform point cloud surface area segmentation method based on improved region growing method Download PDF

Info

Publication number
CN113902688A
CN113902688A CN202111114728.3A CN202111114728A CN113902688A CN 113902688 A CN113902688 A CN 113902688A CN 202111114728 A CN202111114728 A CN 202111114728A CN 113902688 A CN113902688 A CN 113902688A
Authority
CN
China
Prior art keywords
point
points
area
surface area
interior
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111114728.3A
Other languages
Chinese (zh)
Inventor
宋健
王新雅
徐林
陈磊
吴文清
周小燚
刘泓佚
杜永军
唐志强
李燕军
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Wuxi Traffic Construction Engineering Group Co ltd
Wuxi City Key Construction Project Management Center
Southeast University
Original Assignee
Wuxi Traffic Construction Engineering Group Co ltd
Wuxi City Key Construction Project Management Center
Southeast University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Wuxi Traffic Construction Engineering Group Co ltd, Wuxi City Key Construction Project Management Center, Southeast University filed Critical Wuxi Traffic Construction Engineering Group Co ltd
Priority to CN202111114728.3A priority Critical patent/CN113902688A/en
Publication of CN113902688A publication Critical patent/CN113902688A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
    • G06T17/20—Finite element generation, e.g. wire-frame surface description, tesselation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10028—Range image; Depth image; 3D point clouds
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Geometry (AREA)
  • Quality & Reliability (AREA)
  • Computer Graphics (AREA)
  • Software Systems (AREA)
  • Length Measuring Devices By Optical Means (AREA)

Abstract

The application relates to a non-uniform point cloud surface area segmentation method based on an improved region growing method. The method comprises the following steps: removing outlier data points of the three-dimensional laser point cloud to be segmented to obtain a point cloud main body; carrying out down-sampling on the point cloud main body, carrying out rough segmentation, and defining the space coordinate range of each area and area estimation; determining the grid type of each grid of each area based on the eigenvalue of the covariance matrix of k neighborhood points; forming a uniform point data set according to points in all uniform point grids of each area, and determining initial seed points corresponding to each area; formulating a growth rule based on curvature and normal vector information, and determining the growing inner point hair density of each surface area; stopping extracting the area interior points of the area which is greater than or equal to the density threshold value to obtain the area interior points of the area; and (4) carrying out consistency check on the surface area smaller than the density threshold, extracting non-growing interior points in the non-uniform point grid of the surface area to form the surface area interior points of the surface area, and realizing the extraction of the surface area interior points with poor uniformity and continuity.

Description

Non-uniform point cloud surface area segmentation method based on improved region growing method
Technical Field
The application relates to the technical field of three-dimensional laser point cloud plane segmentation, in particular to a non-uniform point cloud plane region segmentation method based on an improved region growing method.
Background
The laser three-dimensional scanning technology is a machine vision type non-contact intelligent sensing technology, and can measure the outline of a three-dimensional object with millimeter-scale precision and generate high-precision space coordinates of a large number of surface points (also called point clouds) of the object to be measured so as to describe the refined geometric characteristics of the object. By applying the technology in the building industry, the relevant data of the whole and local geometric forms and the spatial linear shapes of the components can be quickly acquired, the geometric shape verification and the virtual construction of the key components by workers are assisted, and the relevant digital modeling and filing are realized.
At present, many advances are made in the research of point cloud surface domain segmentation algorithm, wherein the region growing method based on surface features is widely applied due to high efficiency and small memory consumption of the method, and the defects of the method are as follows: the selection of initial seed points lacks a uniform optimal criterion, the accurate and effective area segmentation can be ensured only by correctly selecting the seed points, currently, the point with the minimum curvature is generally selected as the seed point, but the segmentation effect is poor under the conditions of more target planes and the like; the algorithm has high requirements on uniformity and continuity of point cloud density, point cloud is often uneven and discontinuous due to the fact that measured objects are complex in structure and numerous in accessories, mutual shielding is generated, arrangement of a survey station is limited due to field scanning environment conditions and the like, the applicability of a traditional region growing method is weakened, and extraction of points in a surface area with poor uniformity and continuity cannot be achieved.
Disclosure of Invention
In view of the above, it is necessary to provide a non-uniform point cloud surface region segmentation method based on an improved region growing method, which can extract the points in the surface region with poor uniformity and continuity.
A non-uniform point cloud surface region segmentation method based on an improved region growing method comprises the following steps:
removing outlier data points of the three-dimensional laser point cloud to be segmented by adopting a filtering algorithm to obtain a point cloud main body;
adopting a three-dimensional cubic grid filter to perform downsampling on the point cloud main body to obtain a downsampled point cloud main body;
roughly dividing the area of the point cloud main body after the down-sampling, roughly defining the space coordinate range of each area, estimating the area of each area, and numbering each area;
re-dividing the surface areas into corresponding three-dimensional cubic grids, judging the uniformity of point cloud distribution in each grid of each surface area based on the characteristic value of a k neighborhood point covariance matrix, and determining the grid types of the grids of each surface area, wherein the grid types comprise uniform point grids and non-uniform point grids;
forming a mean point data set corresponding to each area according to points in all the mean point grids of each area, calculating a center of mass point corresponding to each mean point data set, and determining the mean point data set of each area to a point with the minimum Euclidean distance to the center of mass point as an initial seed point corresponding to each area;
formulating a growth rule based on curvature and normal vector information, extracting growth interior points of each surface area according to initial seed points corresponding to each surface area, calculating the ratio of the growth interior points of each surface area to the area of the corresponding surface area, and determining the growth interior point hair density of each surface area;
judging whether the growing inner point hair density of each area is greater than or equal to a density threshold value, stopping carrying out area inner point extraction on the area greater than or equal to the density threshold value, wherein the growing inner point of the area is the area inner point of the area;
and taking the growing interior points corresponding to the surface area smaller than the density threshold as target surface area interior points, performing consistency check, extracting non-growing interior points in the non-uniform point grid of the surface area, and forming the surface area interior points of the surface area by the non-growing interior points and the growing interior points.
In one embodiment, the step of removing outlier data points of the three-dimensional laser point cloud to be segmented by using a filtering algorithm to obtain a point cloud main body includes:
each point p in the three-dimensional laser point cloud to be segmentediCalculate k thereof1Fitting a local plane with the neighborhood points to calculate piAnd removing points with the distance exceeding a certain distance threshold value from the distance to the local plane of the point cloud main body to obtain the point cloud main body.
In one embodiment, the down-sampling the point cloud main body by using a three-dimensional cube mesh filter to obtain a down-sampled point cloud main body includes:
and establishing a three-dimensional cubic grid of the point cloud main body, combining points in the same cubic body into a single point, and obtaining the point cloud main body after down-sampling.
In one embodiment, the step of repartitioning the face areas into corresponding three-dimensional cubic grids, determining uniformity of point cloud distribution in each grid of each face area based on a feature value of a k-neighborhood point covariance matrix, and determining a grid type to which each grid of each face area belongs, where the grid type includes a uniform point grid and a non-uniform point grid, includes:
re-dividing each surface domain into corresponding three-dimensional cubic grids, and taking k of each point of the point cloud main body after down-sampling2Each neighborhood point forms a neighborhood point set, and a neighborhood point covariance matrix C of each point is determined, which is expressed as:
Figure BDA0003275153420000031
wherein X, Y and Z are k2The vectors of row 1 and column represent k2Three-dimensional rectangular coordinates of the neighborhood points, cov (X, Y), cov (Y, X) represent the covariance of the two vectors X and Y, cov (X, Z), cov (Z, X) represent the covariance of the two vectors X and Z, respectively, cov (Y, Z), cov (Z, Y) represent the covariance of the two vectors Z and Y, respectively;
solving the eigenvalue of the covariance matrix of the neighborhood points of each point, determining the maximum eigenvalue of the covariance matrix corresponding to each point, and determining the eigenvalue threshold according to the calculation result of the maximum eigenvalue of each point;
and counting the maximum characteristic value corresponding to each point in each grid of the three-dimensional cubic grids of each area, wherein the number of points exceeding the characteristic threshold accounts for the total number of points in the grid, the grids exceeding the proportional threshold are non-uniform point grids, and the grids not exceeding the proportional threshold are uniform point grids.
In one embodiment, the step of formulating a growth rule based on curvature and normal vector information, extracting a growing interior point of each surface area according to an initial seed point corresponding to each surface area, calculating a ratio of the growing interior point number of each surface area to an area of the corresponding surface area, and determining a growing interior point hair density of each surface area includes:
calculating normal vectors and curvatures of each point of each area by adopting a principal component analysis method;
taking the initial seed points corresponding to each area as initial seed points in the seed point set to be searched of each area;
sequentially taking various sub-points in the seed point set to be searched as current searching seed points, and searching k of the current searching seed points3Each neighborhood point;
determining the neighborhood points of which the included angles with the normal vectors of the current search seed points are smaller than an angle threshold value in each neighborhood point as the points in the target surface area;
judging whether the curvature of each neighborhood point is smaller than a curvature threshold value or not according to the curvature threshold value, adding the neighborhood points smaller than the curvature threshold value to the seed point set to be searched, and deleting the current searching seed point from the seed point set to be searched;
returning to execute the steps of sequentially taking various sub-points in the seed point set to be searched as the current searching seed point, and searching k of the current searching seed point3And (5) performing neighborhood point counting until the seed point set to be searched is an empty set.
In one embodiment, the density threshold is determined according to the edge length of the three-dimensional cubic grid, and the density threshold is calculated according to the formula:
Density_th＝a*1/(Edge1^2)
where Density _ th denotes a Density threshold, a denotes a Density reduction coefficient, and Edge1 denotes the Edge length of the three-dimensional cubic grid used for down-sampling.
In one embodiment, the step of performing consistency check by using the growing interior points corresponding to the surface area smaller than the density threshold as target surface area interior points, extracting non-growing interior points in a non-uniform point grid of the surface area, and forming surface area interior points of the surface area by the non-growing interior points and the growing interior points together includes:
taking the growing interior points corresponding to the surface area smaller than the density threshold value as interior points of the target surface area, adding the interior points into the interior point set, and obtaining a fitting plane equation for the points in the interior point set by adopting a least square method;
extracting all points in the non-uniform point grid in the area space to form an alternative point data set;
sequentially popping up points in the data set of the alternative points for consistency check;
when the distance between the current popped point and the fitting plane is smaller than a distance threshold value and the included angle between the normal vector of the current popped point and the normal vector of the fitting plane is smaller than an angle threshold value, the current popped point is a non-growing interior point and is added into an interior point set;
and repeating the step of sequentially popping up the points in the candidate point data set for consistency check until all the points in the candidate point data set pop up, and forming the surface region interior points of the surface region by the non-growing interior points and the growing interior points in the final interior point set.
According to the non-uniform point cloud surface area segmentation method based on the improved region growing method, the outlier data points of the three-dimensional laser point cloud to be segmented are removed by adopting a filtering algorithm, and a point cloud main body is obtained; adopting a three-dimensional cubic grid filter to perform down-sampling on the point cloud main body to obtain a down-sampled point cloud main body; roughly dividing the area of the point cloud main body after down-sampling, roughly defining the space coordinate range of each area, estimating the area of each area, and numbering each area; re-dividing each area into corresponding three-dimensional cubic grids, judging the uniformity of point cloud distribution in each grid of each area based on the characteristic value of the k neighborhood point covariance matrix, and determining the grid type to which each grid of each area belongs, wherein the grid type comprises a uniform point grid and a non-uniform point grid; forming an average point data set corresponding to each area according to points in all average point grids of each area, and determining the average point data set of each area to a point with the minimum Euclidean distance from a centroid point as an initial seed point corresponding to each area; formulating a growth rule based on curvature and normal vector information, extracting growth interior points of each surface area according to initial seed points corresponding to each surface area, calculating the ratio of the number of the growth interior points of each surface area to the area of the corresponding surface area, and determining the growth interior point hair density of each surface area; judging whether the growing inner point hair density of each surface area is greater than or equal to a density threshold value, stopping performing surface area inner point extraction on the surface area greater than or equal to the density threshold value, wherein the growing inner point of the surface area is the surface area inner point of the surface area; and taking the growing interior points corresponding to the surface area smaller than the density threshold as the interior points of the target surface area, carrying out consistency check, extracting the non-growing interior points in the non-uniform point grid of the surface area, and forming the surface area interior points of the surface area by the non-growing interior points and the growing interior points together to realize the extraction of the surface area interior points with poor uniformity and continuity.
Drawings
FIG. 1 is a schematic flow chart of a non-uniform point cloud surface region segmentation method based on an improved region growing method in one embodiment;
FIG. 2 is a schematic flow chart of a non-uniform point cloud surface region segmentation method based on an improved region growing method in another embodiment;
FIG. 3 is a schematic view of a surface area of a precast concrete box girder casting formwork in one embodiment;
FIG. 4 is a result of dividing the area No. 1 into a uniform point mesh and a non-uniform point mesh in one embodiment;
FIG. 5 is a result of dividing the area equal-point mesh and the area non-equal-point mesh of No. 2-3 in one embodiment;
FIG. 6 is a face segmentation result of left and right webs in one embodiment;
FIG. 7 is a result of floor area segmentation in one embodiment;
FIG. 8 is a diagram illustrating the overall effect of three surface segmentation in one embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
In one embodiment, as shown in fig. 1, there is provided a non-uniform point cloud surface region segmentation method based on an improved region growing method, including the following steps:
step S220, removing outlier data points of the three-dimensional laser point cloud to be segmented by adopting a filtering algorithm to obtain a point cloud main body.
The three-dimensional laser point cloud to be segmented can be the three-dimensional laser point cloud after blocky noise in the point cloud is removed by manually adopting professional point cloud processing software, such as removing noise of scanned field sundries and the like.
In one embodiment, the step of obtaining the point cloud main body by removing outlier data points of the three-dimensional laser point cloud to be segmented by using a filtering algorithm includes: each point p in the three-dimensional laser point cloud to be segmentediCalculate k thereof1Fitting a local plane with the neighborhood points to calculate piAnd removing points with the distance exceeding a certain distance threshold value from the distance to the local plane of the point cloud main body to obtain the point cloud main body.
Wherein k is1Is a positive integer, the value is selected according to actual conditions, such as k1Take 8, 15, 20, etc. The distance threshold is set according to practical conditions, such as 0.1m, 0.2m, 0.01m and the like.
And S240, performing down-sampling on the point cloud main body by adopting a three-dimensional cubic grid filter to obtain the down-sampled point cloud main body.
In one embodiment, the down-sampling of the point cloud body by using the three-dimensional cubic mesh filter to obtain the down-sampled point cloud body includes: and establishing a three-dimensional cubic grid of the point cloud main body, combining points in the same cubic body into a single point, and obtaining the point cloud main body after down-sampling.
The method comprises the steps of establishing a three-dimensional cubic grid of a point cloud main body, combining points in the same cubic body into a single point, achieving the purpose of down-sampling, well retaining an original boundary of the point cloud, removing noise points above and below a plane, and being regarded as a further smoothing and filtering means.
Step S260, roughly dividing the surface area of the point cloud main body after down-sampling, roughly defining the space coordinate range of each surface area, estimating the area of each surface area, and numbering each surface area.
Step S280, each surface area is divided into corresponding three-dimensional cubic grids again, the uniformity of point cloud distribution in each grid of each surface area is judged based on the characteristic value of the k neighborhood point covariance matrix, and the grid type of each grid of each surface area is determined, wherein the grid type comprises a uniform point grid and a non-uniform point grid.
In one embodiment, the method comprises the steps of re-dividing each face area into corresponding three-dimensional cubic grids, judging the uniformity of point cloud distribution in each grid of each face area based on the characteristic value of a k neighborhood point covariance matrix, and determining the grid type of each grid of each face area, wherein the grid type comprises a uniform point grid and a non-uniform point grid, and the method comprises the following steps:
each area is divided into corresponding three-dimensional cubic grids again, and k of each point of the point cloud main body after down-sampling is taken2Each neighborhood point forms a neighborhood point set, and a neighborhood point covariance matrix C of each point is determined, which is expressed as:
Figure BDA0003275153420000081
wherein X, Y and Z are k2The vectors of row 1 and column represent k2Three-dimensional rectangular coordinates of the neighborhood points, cov (X, Y), cov (Y, X) represent the covariance of the two vectors X and Y, cov (X, Z), cov (Z, X) represent the covariance of the two vectors X and Z, respectively, cov (Y, Z), cov (Z, Y) represent the covariance of the two vectors Z and Y, respectively;
solving the eigenvalue of the covariance matrix of the neighborhood points of each point, determining the maximum eigenvalue of the covariance matrix corresponding to each point, and determining the eigenvalue threshold according to the calculation result of the maximum eigenvalue of each point; and (3) counting the maximum characteristic value corresponding to each point in each grid of the three-dimensional cubic grids of each area, wherein the number of points exceeding a characteristic threshold accounts for the total number of points in the grid, the grid exceeding the proportional threshold is a non-uniform point grid, and the grid not exceeding the proportional threshold is a uniform point grid.
And if the point exceeding the feature threshold value shows that the neighborhood point of the point has obvious dispersity in the direction of the feature vector corresponding to the maximum feature value, the point is considered to be distributed unevenly in all directions in the neighborhood range.
Step S300, forming a uniform point data set corresponding to each area according to points in all uniform point grids of each area, calculating a centroid point of each uniform point data set, and determining the uniform point data set of each area to a point with the minimum Euclidean distance to the centroid point as an initial seed point corresponding to each area.
And S320, formulating a growth rule based on the curvature and normal vector information, extracting the growing interior points of each surface area according to the initial seed points corresponding to each surface area, calculating the ratio of the growing interior points of each surface area to the area of the corresponding surface area, and determining the growing interior point hair density of each surface area.
In one embodiment, the method for determining the growing rule based on the curvature and normal vector information includes the steps of extracting growing interior points of each surface area according to initial seed points corresponding to each surface area, calculating the ratio of the growing interior points of each surface area to the corresponding surface area, and determining the growing interior point hair density of each surface area, including:
calculating normal vectors and curvatures of each point of each area by adopting a principal component analysis method; taking the initial seed points corresponding to each area as initial seed points in the seed point set to be searched of each area; sequentially taking various sub-points in the seed point set to be searched as the current searching seed point, and searching k of the current searching seed point3Each neighborhood point; determining the neighborhood points of which the included angle with the normal vector of the current search seed point is smaller than an angle threshold value in each neighborhood point as the points in the target surface area; judging whether the curvature of each neighborhood point is smaller than a curvature threshold value or not according to the curvature threshold value, adding the neighborhood points smaller than the curvature threshold value to the seed point set to be searched, and deleting the current seed point to be searched from the seed point set to be searched; returning to execute the k for searching the current searching seed point by taking various sub-points in the seed point set to be searched as the current searching seed point3And (5) performing neighborhood point counting until the seed point set to be searched is an empty set.
Step S340, judging whether the growing inner point hair density of each surface area is greater than or equal to the density threshold value, and stopping performing surface area inner point extraction on the surface area greater than or equal to the density threshold value, wherein the growing inner point of the surface area is the surface area inner point of the surface area.
In one embodiment, the density threshold is determined according to the edge length of the three-dimensional cubic grid, and the density threshold is calculated by the following formula:
Density_th＝a*1/(Edge1^2)
where Density _ th denotes a Density threshold, a denotes a Density reduction coefficient, and Edge1 denotes the Edge length of the three-dimensional cubic grid used for down-sampling.
And step S360, taking the growing interior points corresponding to the surface area smaller than the density threshold as the target surface area interior points, carrying out consistency check, extracting the non-growing interior points in the non-uniform point grid of the surface area, and forming the surface area interior points of the surface area by the non-growing interior points and the growing interior points.
In one embodiment, the step of performing consistency check by using growing interior points corresponding to a surface area smaller than a density threshold as target surface area interior points, extracting non-growing interior points in a non-uniform point grid of the surface area, and forming surface area interior points of the surface area by the non-growing interior points and the growing interior points together includes:
taking growing interior points corresponding to a surface area smaller than a density threshold value as interior points of a target surface area, adding the interior points to an interior point set corresponding to the surface area, and obtaining a fitting plane equation for the points in the interior point set by adopting a least square method; extracting all points in the non-uniform point grid in the area space to form an alternative point data set; sequentially popping up points in the data set of the alternative points for consistency check; when the distance between the current popped point and the fitting plane is smaller than a distance threshold value and the included angle between the normal vector of the current popped point and the normal vector of the fitting plane is smaller than an angle threshold value, the current popped point is a non-growing interior point and is added into an interior point set; and repeating the step of sequentially popping up the points in the candidate point data set for consistency check until all the points in the candidate point data set pop up, and forming the surface region interior points of the surface region by the non-growing interior points and the growing interior points in the final interior point set.
In one embodiment, as shown in fig. 2, the non-uniform point cloud surface area segmentation method based on the improved area growing method is specifically applied to the surface area segmentation of the three-dimensional laser point cloud (i.e. the three-dimensional laser point cloud to be segmented) of the web plate and the bottom plate on both sides of the precast concrete box girder pouring template (hereinafter referred to as a precast template) shown in fig. 3, for example:
step 1, each point P in the point cloud P of the prefabricated templateiCalculate k thereof1Fitting a local plane with the neighborhood points to calculate piAnd if the distance to the local plane exceeds a certain distance threshold Dist _ th _ dense, defining the point as an outlier and removing the outlier to obtain a point cloud Q, and performing the rest steps on the point cloud Q. In this example k1And taking 8, wherein the value of Dist _ th _ dense is 0.1 m.
The prefabricated template point cloud P is a three-dimensional laser point cloud of webs and a bottom plate on two sides of a prefabricated concrete box girder pouring template (hereinafter referred to as a prefabricated template), and the point cloud is obtained after block noise in the point cloud is removed in advance through professional point cloud processing software, wherein the block noise in the point cloud is like a pedestal under the prefabricated template, field sundries scanning and the like.
Step 2: and (4) downsampling the point cloud main body by adopting a three-dimensional cubic grid filter to obtain the downsampled point cloud main body.
A three-dimensional uniform cubic grid is established for the point cloud Q, the Edge length Edge1 of the cubic grid is 0.02m, and points in the same cube are combined into a single point, so that the purpose of downsampling is achieved. The method well keeps the original boundary of the point cloud, simultaneously removes noise points above and below the plane, and can be regarded as a further smoothing and filtering means.
And step 3: roughly demarcating the space coordinate range of each area of the point cloud main body after down-sampling and estimating the area A of the point cloud main bodyiEach area is numbered (i ═ 1,2,3 …).
In this embodiment, the center of the bottom surface of the prefabricated template is used as the origin of coordinates, and the number, name, spatial range and area of each surface are listed in table 1.
Table 1 table of results of area delineation
Figure BDA0003275153420000111
Figure BDA0003275153420000121
And 4, step 4: dividing each rough dividing area into a plurality of three-dimensional cubic grids again, judging the uniformity of point cloud distribution in each grid based on the characteristic value of the k neighborhood point covariance matrix, and accordingly dividing each grid into two types, namely a uniform point grid and a non-uniform point grid;
wherein, taking k of each point of the point cloud subject Q2Each neighborhood point forms a neighborhood point set, and a neighborhood point covariance matrix C is made:
Figure BDA0003275153420000122
wherein X, Y and Z are all k2The vectors of row 1 and column represent k2The three-dimensional rectangular coordinates of the neighborhood points cov (X, Y), cov (Y, X) each represent the covariance of the two vectors X and Y, cov (X, Z), cov (Z, X) each represent the covariance of the two vectors X and Z, and cov (Y, Z), cov (Z, Y) each represent the covariance of the two vectors Z and Y.
Further, the feature value of the covariance matrix of the neighborhood point of each point is calculated, the maximum feature value of each point is determined, the feature threshold value is determined according to the calculation result of the maximum feature value, and if the maximum feature value of a certain point exceeds the feature threshold value, the feature threshold value is determined. Then, it is indicated that the neighborhood point of the point has a relatively obvious dispersion in the direction of the eigenvector corresponding to the maximum eigenvalue, and the point is considered to be distributed relatively unevenly in each direction in the neighborhood range.
Further, a plurality of cubic grids newly divided into 3 surface domain spaces are assigned with Edge length as Edge2, the maximum characteristic value corresponding to each point in each cubic grid is counted, the number of points exceeding the characteristic threshold accounts for the total number of points in the grid, if the number exceeds a certain ratio threshold, the grid is defined as a non-uniform point grid, otherwise, the grid is a uniform point grid, and the Edge2 of the embodiment takes 0.1m, k2And 8, taking the characteristic threshold value as 0.0004 and taking the proportion threshold value as 80%. In the embodiment, only one measuring station is arranged for field scanning, and the point cloud density is farther away from the laser scannerThe lower the point cloud uniformity and continuity, the worse. As shown in fig. 4 and 5, the results of the division of the uniform point mesh and the non-uniform point mesh in the surface areas of nos. 1 to 3 of this embodiment are shown.
And 5: points in all the uniform point grids of each area form a uniform point data set DiCalculating the centroid point M of the mean point data seti，DiMiddle to centroid Euclidean distance minimum point SiThe initial seed point defined as face field i.
Wherein, the points in the uniform point grid in 3 surface domain spaces are respectively extracted to form 3 uniform point data sets Di(i ═ 1,2,3), calculate DiCentroid point Mi(i ═ 1,2,3) coordinates, finding the mean-point data set DiCenter to center of mass point MiS is the point with the smallest euclidean distanceiAnd (i ═ 1,2 and 3) defined as the initial seed point of the i-number face domain, so as to ensure that the point cloud distribution around the initial seed point is uniform and continuous.
Step 6: establishing a growth rule based on curvature and normal vector information, extracting interior points of each surface area, namely growth interior points, and calculating the number N of the growth interior pointsiCorresponding to the area A of the surface areaiThe ratio of (A) to (B) is called the growing inner point hair Density _ gro.
Calculating normal vectors and curvatures of each point by adopting a principal component analysis method; taking the initial seed points corresponding to each area as initial seed points in the seed point set to be searched of each area; sequentially taking various sub-points in the seed point set to be searched as the current searching seed point, and searching k of the current searching seed point3And if the curvature of the neighborhood point is smaller than the curvature threshold value, the neighborhood point is added to the seed point set to be searched, k3Deleting the current seed point after the judgment of each neighborhood point is finished, executing the steps of taking various seed points in the seed point set to be searched as the current searching seed point in sequence, and searching k of the current searching seed point3And (5) performing neighborhood point counting until the seed point set to be searched is an empty set. In this example k3Take 8, the curvature threshold 0.05, and the angle threshold 5 °.
Further, the number N of growing interior points of 3 surface domains is countedi(i-1, 2,3), calculating the number of growing inliers NiAnd the area A of the surfaceiThe ratio of (a) to (b), i.e., the grown inner spot wool Density _ gro, and the calculation of the areal grown inner spot wool Density of this example 3 are shown in Table 2.
TABLE 2 growth inner point wool density
Figure BDA0003275153420000141
And 7: and if the hair density of the growing inner points is greater than the density threshold value, the extraction of the inner points of the target surface area is considered to be finished, and the growing inner points are the inner points of the target surface area.
Further, this density threshold is determined in this embodiment by the Edge length Edge1 of the cube mesh in step 2.
Density_th＝a*1/(Edge1^2)
Edge1 is 0.02m in this embodiment. The distance from the scanner to the object to be measured affects the density of the point cloud, and the density reduction needs to be considered. According to the precision of the scanner and the on-site measurement distance, the density reduction coefficient a of the embodiment takes a value of 0.85. Density _ th is calculated as 2125.
The growing interior point hair density of the face areas of the left web and the right web is greater than 2125, and the face area interior point extraction is completed. The extraction of the points in the floor surface area, namely the No. 1 surface area, needs to be further completed through consistency check. As shown in fig. 6, the results of dividing the surface area of the left web and the right web, i.e., No. 2 and No. 3, in this embodiment are shown.
And 8: if the hair density of the growing inner points is smaller than the density threshold value, further consistency check is needed, and the obtained growing inner points are set as target area inner points. And verifying whether the points in the alternative data set have similar characteristics with the points in the set target surface area, namely consistency check, if the check conditions are met, the points are regarded as the points in the set target surface area, and thus non-growing interior points in the surface area are extracted to form the points in the target surface area together with the growing interior points.
Wherein the obtained growing interior pointSetting the interior points of the target surface area and adding the interior points to the interior point set I corresponding to the No. 1 surface area1In pair I1The points in (1) are subjected to a least square method to obtain a fitting plane equation. Extracting all points in non-uniform point grids in the floor surface domain space to form an alternative point data set E1Sequentially pop up E1The points in (1) were checked for consistency.
Further, the consistency check condition includes: the distance between the current pop-up point and the fitting plane is smaller than a certain distance threshold, and the included angle between the normal vector of the current pop-up point and the normal vector of the fitting plane is smaller than a certain angle threshold. If the current pop-up point meets the consistency test condition, the current pop-up point is regarded as the target surface region interior point, namely the non-growing interior point, through the consistency test and is added to the interior point set I1In this embodiment, the distance threshold is 0.01m, and the angle threshold is 10 °. Repeating the above steps until the alternative point data set E1All the points in the list are popped up, and the final inner point set I1Namely the target area inner point. As shown in fig. 7, the result of dividing the floor surface area, i.e., the surface area 1 in this embodiment is shown in fig. 8, which is a three-dimensional laser point cloud surface area division effect diagram of the two side webs and the floor of the precast concrete box girder casting formwork (hereinafter referred to as precast formwork) in this embodiment.
It should be understood that, although the steps in the flowchart of fig. 1 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a portion of the steps in fig. 1 may include multiple sub-steps or multiple stages that are not necessarily performed at the same time, but may be performed at different times, and the order of performance of the sub-steps or stages is not necessarily sequential, but may be performed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above examples only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims (7)

1. A non-uniform point cloud surface region segmentation method based on an improved region growing method is characterized by comprising the following steps:
removing outlier data points of the three-dimensional laser point cloud to be segmented by adopting a filtering algorithm to obtain a point cloud main body;
adopting a three-dimensional cubic grid filter to perform downsampling on the point cloud main body to obtain a downsampled point cloud main body;
roughly dividing the area of the point cloud main body after the down-sampling, roughly defining the space coordinate range of each area, estimating the area of each area, and numbering each area;
re-dividing the surface areas into corresponding three-dimensional cubic grids, judging the uniformity of point cloud distribution in each grid of each surface area based on the characteristic value of a k neighborhood point covariance matrix, and determining the grid types of the grids of each surface area, wherein the grid types comprise uniform point grids and non-uniform point grids;
forming a mean point data set corresponding to each area according to points in all the mean point grids of each area, calculating a center of mass point of each mean point data set, and determining the mean point data set of each area to a point with the minimum Euclidean distance to the center of mass point as an initial seed point corresponding to each area;
formulating a growth rule based on curvature and normal vector information, extracting growth interior points of each surface area according to initial seed points corresponding to each surface area, calculating the ratio of the growth interior points of each surface area to the area of the corresponding surface area, and determining the growth interior point hair density of each surface area;
judging whether the growing inner point hair density of each area is greater than or equal to a density threshold value, stopping carrying out area inner point extraction on the area greater than or equal to the density threshold value, wherein the growing inner point of the area is the area inner point of the area;
and taking the growing interior points corresponding to the surface area smaller than the density threshold as target surface area interior points, performing consistency check, extracting non-growing interior points in the non-uniform point grid of the surface area, and forming the surface area interior points of the surface area by the non-growing interior points and the growing interior points.
2. The method of claim 1, wherein the step of obtaining the point cloud body by removing outlier data points of the three-dimensional laser point cloud to be segmented using a filtering algorithm comprises:
each point p in the three-dimensional laser point cloud to be segmentediCalculate k thereof1Fitting a local plane with the neighborhood points to calculate piAnd removing points with the distance exceeding a certain distance threshold value from the distance to the local plane of the point cloud main body to obtain the point cloud main body.
3. The method of claim 1, wherein the down-sampling the point cloud body using a three-dimensional cube mesh filter to obtain a down-sampled point cloud body comprises:
and establishing a three-dimensional cubic grid of the point cloud main body, combining points in the same cubic body into a single point, and obtaining the point cloud main body after down-sampling.
4. The method according to claim 1, wherein the step of repartitioning each of the face regions into corresponding three-dimensional cubic grids, determining uniformity of point cloud distribution in each grid of each of the face regions based on eigenvalues of a k-neighborhood point covariance matrix, and determining a grid type to which each grid of each of the face regions belongs, the grid type including a uniform point grid and a non-uniform point grid, comprises:
re-dividing each surface domain into corresponding three-dimensional cubic grids, and taking k of each point of the point cloud main body after down-sampling2Each neighborhood point forms a neighborhood point set, and a neighborhood point covariance matrix C of each point is determined, which is expressed as:
Figure FDA0003275153410000021
wherein X, Y and Z are k2The vectors of row 1 and column represent k2Three-dimensional rectangular coordinates of the neighborhood points, cov (X, Y), cov (Y, X) represent the covariance of the two vectors X and Y, cov (X, Z), cov (Z, X) represent the covariance of the two vectors X and Z, respectively, cov (Y, Z), cov (Z, Y) represent the covariance of the two vectors Z and Y, respectively;
solving the eigenvalue of the covariance matrix of the neighborhood points of each point, determining the maximum eigenvalue of the covariance matrix corresponding to each point, and determining the eigenvalue threshold according to the calculation result of the maximum eigenvalue of each point;
and counting the maximum characteristic value corresponding to each point in each grid of the three-dimensional cubic grids of each area, wherein the number of points exceeding the characteristic threshold accounts for the total number of points in the grid, the grids exceeding the proportional threshold are non-uniform point grids, and the grids not exceeding the proportional threshold are uniform point grids.
5. The method according to claim 1, wherein the step of formulating a growth rule based on curvature and normal vector information, extracting growing interior points of each surface area according to initial seed points corresponding to each surface area, calculating a ratio of the number of the growing interior points of each surface area to the area of the corresponding surface area, and determining the growing interior point hair density of each surface area comprises:
calculating normal vectors and curvatures of each point of each area by adopting a principal component analysis method;
taking the initial seed points corresponding to each area as initial seed points in the seed point set to be searched of each area;
sequentially taking various sub-points in the seed point set to be searched as current searching seed points, and searching k of the current searching seed points3Each neighborhood point;
determining the neighborhood points of which the included angles with the normal vectors of the current search seed points are smaller than an angle threshold value in each neighborhood point as the points in the target surface area;
judging whether the curvature of each neighborhood point is smaller than a curvature threshold value or not according to the curvature threshold value, adding the neighborhood points smaller than the curvature threshold value to the seed point set to be searched, and deleting the current searching seed point from the seed point set to be searched;
returning to execute the steps of sequentially taking various sub-points in the seed point set to be searched as the current searching seed point, and searching k of the current searching seed point3And (5) performing neighborhood point counting until the seed point set to be searched is an empty set.
6. The method of claim 1, wherein the density threshold is determined according to the edge length of the three-dimensional cubic grid, and the density threshold is calculated according to the formula:
Density_th＝a*1/(Edge1^2)
where Density _ th denotes a Density threshold, a denotes a Density reduction coefficient, and Edge1 denotes the Edge length of the three-dimensional cubic grid used for down-sampling.
7. The method according to claim 1, wherein the step of performing a consistency check with the growing interior points corresponding to the surface area smaller than the density threshold as target surface area interior points to extract non-growing interior points in a non-uniform point grid of the surface area, and forming surface area interior points of the surface area by the non-growing interior points and the growing interior points comprises:
taking the growing interior points corresponding to the surface area smaller than the density threshold value as interior points of the target surface area, adding the interior points to an interior point set corresponding to the surface area, and obtaining a fitting plane equation for the points in the interior point set by adopting a least square method;
extracting all points in the non-uniform point grid in the area space to form an alternative point data set;
sequentially popping up points in the data set of the alternative points for consistency check;
when the distance between the current popped point and the fitting plane is smaller than a distance threshold value and the included angle between the normal vector of the current popped point and the normal vector of the fitting plane is smaller than an angle threshold value, the current popped point is a non-growing interior point and is added into an interior point set;
and repeating the step of sequentially popping up the points in the alternative point data set for consistency check until all the points in the alternative point data set pop up, and forming the surface region interior points of the surface region by the non-growing interior points and the growing interior points in the final interior point set.
CN202111114728.3A 2021-09-23 2021-09-23 Non-uniform point cloud surface area segmentation method based on improved region growing method Pending CN113902688A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111114728.3A CN113902688A (en) 2021-09-23 2021-09-23 Non-uniform point cloud surface area segmentation method based on improved region growing method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111114728.3A CN113902688A (en) 2021-09-23 2021-09-23 Non-uniform point cloud surface area segmentation method based on improved region growing method

Publications (1)

Publication Number Publication Date
CN113902688A true CN113902688A (en) 2022-01-07

Family

ID=79028928

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111114728.3A Pending CN113902688A (en) 2021-09-23 2021-09-23 Non-uniform point cloud surface area segmentation method based on improved region growing method

Country Status (1)

Country Link
CN (1) CN113902688A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115408549A (en) * 2022-08-31 2022-11-29 深圳前海瑞集科技有限公司 Workpiece point cloud filtering method and device, computer readable medium and electronic equipment
CN117218143A (en) * 2023-11-07 2023-12-12 法奥意威(苏州)机器人系统有限公司 Point cloud segmentation method and device

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115408549A (en) * 2022-08-31 2022-11-29 深圳前海瑞集科技有限公司 Workpiece point cloud filtering method and device, computer readable medium and electronic equipment
CN115408549B (en) * 2022-08-31 2024-04-12 深圳前海瑞集科技有限公司 Workpiece point cloud filtering method and device, computer readable medium and electronic equipment
CN117218143A (en) * 2023-11-07 2023-12-12 法奥意威(苏州)机器人系统有限公司 Point cloud segmentation method and device
CN117218143B (en) * 2023-11-07 2024-01-23 法奥意威(苏州)机器人系统有限公司 Point cloud segmentation method and device

Similar Documents

Publication Publication Date Title
Woo et al. A new segmentation method for point cloud data
CN106780524B (en) Automatic extraction method for three-dimensional point cloud road boundary
CN107610061B (en) Edge-preserving point cloud hole repairing method based on two-dimensional projection
CN110060338B (en) Prefabricated part point cloud identification method based on BIM model
CN109685080B (en) Multi-scale plane extraction method based on Hough transformation and region growth
CN113902688A (en) Non-uniform point cloud surface area segmentation method based on improved region growing method
US20190188856A1 (en) Systems and methods for block based edgel detection with false edge elimination
Zhu et al. Seed point selection method for triangle constrained image matching propagation
CN107644452A (en) Airborne LiDAR point cloud roof dough sheet dividing method and system
CN104616349A (en) Local curved surface change factor based scattered point cloud data compaction processing method
CN110807781A (en) Point cloud simplification method capable of retaining details and boundary features
CN112381862B (en) Full-automatic registration method and device for CAD (computer-aided design) model and triangular mesh
CN110047133A (en) A kind of train boundary extraction method towards point cloud data
CN114332291A (en) Oblique photography model building outer contour rule extraction method
CN110544308A (en) Transformer substation modeling method and device, computer equipment and storage medium
CN115690184A (en) Tunnel face displacement measurement method based on three-dimensional laser scanning
CN113409332B (en) Building plane segmentation method based on three-dimensional point cloud
CN114898118A (en) Automatic statistical method and system for power transmission line house removal amount based on multi-source point cloud
Liu et al. Application of three-dimensional laser scanning in the protection of multi-dynasty ceramic fragments
CN113838069B (en) Point cloud segmentation method and system based on flatness constraint
CN113971718A (en) Method for performing Boolean operation on three-dimensional point cloud model
CN104821015B (en) Surface sampling point α-shape curved surface topology rebuilding methods in kind
CN111881964A (en) Linear building mode identification method and system based on Delaunay triangulation network
CN116012613B (en) Method and system for measuring and calculating earthwork variation of strip mine based on laser point cloud
CN111598941A (en) Method, device and equipment for measuring gradient of tower and storage medium

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination