CN110807781A - Point cloud simplification method capable of retaining details and boundary features - Google Patents

Point cloud simplification method capable of retaining details and boundary features Download PDF

Info

Publication number
CN110807781A
CN110807781A CN201911016983.7A CN201911016983A CN110807781A CN 110807781 A CN110807781 A CN 110807781A CN 201911016983 A CN201911016983 A CN 201911016983A CN 110807781 A CN110807781 A CN 110807781A
Authority
CN
China
Prior art keywords
point cloud
cloud data
point
neighborhood
boundary
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.)
Granted
Application number
CN201911016983.7A
Other languages
Chinese (zh)
Other versions
CN110807781B (en
Inventor
肖恭兵
刘伟东
刘屿
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.)
South China University of Technology SCUT
Original Assignee
South China University of Technology SCUT
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 South China University of Technology SCUT filed Critical South China University of Technology SCUT
Priority to CN201911016983.7A priority Critical patent/CN110807781B/en
Publication of CN110807781A publication Critical patent/CN110807781A/en
Application granted granted Critical
Publication of CN110807781B publication Critical patent/CN110807781B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Image Generation (AREA)
  • Image Processing (AREA)

Abstract

The invention discloses a point cloud simplification method for retaining details and boundary characteristics, which comprises the following steps of reading original point cloud model data, obtaining original point cloud data, carrying out rasterization processing, and establishing a local k-d tree; fitting the minimum tangent plane, and calculating the weighted equivalent resultant force F of the projection points of the point cloud dataiAnd its average value F when FiIf the number is more than F, the boundary point is determined and reserved, and a boundary reduced subset PSet1 is obtained; for non-boundary points, estimate their local density ρiPartitioning to obtain corresponding subsets; calculating normal vector of non-boundary point according to local density rhoiCalculating a local feature factor DiAnd their average values D, respectively, for the respective subsets to select the threshold value muTHWhen D is presenti/D＞μTHKeeping the point cloud data to obtain a feature point reduced subset PSet 2; combining the boundary reduced subset PSet1 and the characteristic point reduced subset PSet2 to obtain a reduced point cloud data set PSet,the method is convenient for obtaining the point cloud model with proper data volume and no distortion, improves the point cloud simplification efficiency, and enables the description of the detail characteristics and the boundary characteristics of the curved surface to be more accurate.

Description

Point cloud simplification method capable of retaining details and boundary features
Technical Field
The invention relates to the research field of reverse engineering and surveying and mapping, in particular to a point cloud simplification method for retaining details and boundary characteristics.
Background
With the continuous development of three-dimensional scanning technology, machine vision and other technologies, three-dimensional point cloud data is widely applied in the fields of industrial detection, reverse engineering, medical diagnosis, cultural relic protection and the like. Meanwhile, the accuracy and speed of the non-contact scanning equipment are higher and higher, and the amount of acquired point cloud data is also increased sharply. In the actual measurement process, dense point clouds of hundreds of thousands or even millions of data points can be obtained, and the generated redundant data points greatly increase the operation load of a computer. Therefore, it is necessary to reduce the three-dimensional point cloud data without affecting the characteristics of the object to be measured.
In recent years, the simplification of point cloud data is mainly divided into two main categories of methods based on spatial segmentation and on curvature itself. Methods based on spatial segmentation include bounding box methods, uniform mesh methods, and triangular mesh methods. The bounding box method achieves the purpose of simplification by retaining points closest to the center of the grid in each equally divided grid of the minimum bounding box of the point cloud data. The method has a good processing result for models with large data volume and simple structures, but many details can be lost for models with complex structures and various curved surfaces. The uniform grid method is an optimization method of a bounding box method, and carries out simplification processing by calculating a median point of a Z coordinate in a cube by using median filtering. The method based on triangular meshing needs to triangulate original point cloud data and remove part of triangles through normal vectors of the triangles so as to achieve the purpose of simplification. The curvature-based method generally requires constructing a point cloud topological structure through K neighborhood search and fitting a least square surface, and finally, formulating different reduction strategies through curvature thresholds. The method has a good reduction effect, but the calculation of the curvature is usually time-consuming, and other schemes for indirectly replacing the curvature calculation are usually required to be searched for optimization.
In the prior art, CN101373540B discloses a point cloud reduction method based on paraboloid fitting, which is essentially a reduction algorithm based on curvature threshold, and the time cost in curvature calculation is a disadvantage that the method cannot avoid; CN101021954A discloses a point cloud simplification method for simplifying curvature calculation, which replaces curvature calculation by the ratio of the distance from a point to be measured to a neighborhood point to the distance from the point to a fitting tangent plane, and uses the average value of approximate curvatures and the difference value of the approximate curvatures of each point as a judgment threshold value to carry out point cloud simplification; CN102890828A discloses a point cloud data simplification method based on normal included angle, which is characterized in that the method calculates the average value of dot products of the normal vector of each point cloud data point and the normal vector of the point cloud data points in the neighborhood, divides the regions by using the change of the value and sets a sampling ratio, and deletes the points which do not accord with the sampling ratio, thereby avoiding the complex quadric surface fitting and curvature estimation, but in the practical application, the parameters which need to be artificially set are too much, and the practical effect is often different due to experience; CN106373118A discloses a method for simplifying complex curved surface part point clouds capable of effectively retaining boundary and local features, which includes roughly classifying point cloud normal vectors and neighborhood point cloud normal vector dot products as threshold values, then completing point cloud data subdivision through a k-means clustering method and Hausdorff distance to complete point cloud data simplification, and because distance weight factors are not considered during neighborhood searching and point cloud normal vector calculation, the result of the k-means clustering method is not stable enough, so that unpredictable errors occur in boundary feature retention finally.
Disclosure of Invention
The invention aims to overcome the defects of the prior art, provides the point cloud simplification method for reserving the details and the boundary characteristics, is convenient for obtaining a point cloud model with proper data volume and no distortion, improves the point cloud simplification efficiency and enables the description of the detail characteristics and the boundary characteristics of the curved surface to be more accurate.
The purpose of the invention is realized by the following technical scheme:
a point cloud simplification method for retaining details and boundary features is characterized by comprising the following steps:
s1, reading the original point cloud model data to obtain original point cloud data;
s2, rasterizing the original point cloud data, and establishing a local k-d tree, wherein the local k-d tree is established specifically only for each point cloud dataEstablishing a k-d tree for the grid where the grid is located and the neighborhood grid; wherein the neighborhood grid of each point cloud data is defined as (l)x±i,ly±j,lz±k)，i,j,k∈[-1,1]；
S3, fitting a minimum tangent plane to the rasterized point cloud data and k neighborhood points thereof, projecting the point cloud data and the k neighborhood points thereof onto the tangent plane, calculating vectors from projection points of the point cloud data to neighborhood projection points, and normalizing the vectors to obtain normalized data
Figure BDA0002246010340000021
Calculating a weighted equivalent resultant force F of the projected points of the point cloud dataiFurther calculating the average value F of the weighted equivalent resultant force, if FiIf the point cloud data is more than F, the point cloud data is regarded as boundary points, and the boundary points are reserved to obtain a reduced boundary subset PSet 1;
s4, for the non-boundary points, estimating the local density rho in the k neighborhood to which each point cloud data belongsiAnd partitioning to obtain an under-feature point set S1Feature point set S2And rich feature point set S3；
S5, calculating normal vectors of non-boundary points, and calculating a weighted average value theta of included angles between each point cloud data and normal vectors of each point in the neighborhood according to distance weightsi；
S6, according to the local density rhoiCalculating local characteristic factors and further calculating the average values of the local characteristic factors, which are respectively an underdetermined characteristic point set S1Feature point set S2And rich feature point set S3Selecting a suitable threshold value muTHWhen D is presenti/D＞μTHIf so, retaining the point cloud data, otherwise deleting the point cloud data, and combining the point cloud data sets retained by each subset to obtain a characteristic point reduced subset PSet 2;
and S7, merging the boundary reduced subset PSet1 and the feature point reduced subset PSet2 to obtain a reduced point cloud data set PSet.
Further, the step S2 is specifically: dividing the side length L of a cube grid according to the random sampling density rho of the point cloud data, wherein the minimum side length L of the cube grid is m times of the sampling density rho, dividing the point cloud data bounding box into space grids, and establishing a k neighborhood for each point cloud data.
Further, the dividing of the point cloud data bounding box into spatial grids specifically includes: grid number M for dividing point cloud data according to x, y and z directionsx,My,MzThe calculation is as follows:
Figure BDA0002246010340000031
wherein ,xmaxIs the maximum value in the x direction of the coordinate, xminIs the minimum value in x direction of the coordinate, ymaxIs the maximum value in the y direction of the coordinate, yminIs the minimum value in the y direction of the coordinate, zmaxIs the maximum value in the z direction of the coordinate, zminIs the minimum value in the z direction of the coordinate;
Figure BDA0002246010340000032
to round, δ is the expansion number.
Further, the sampling density ρ is obtained by calculating:
Figure BDA0002246010340000033
wherein n is the number of point cloud data taken out randomly, the value range of n is 20-30, and diIs the interval from the closest point of the point cloud data.
Further, the step S3 is specifically: fitting a tangent plane to each point cloud data and k neighborhood points thereof, wherein the fitting of the minimum tangent plane uses a least square method to project the point cloud data and the k neighborhood points thereof onto the tangent plane, the vector from the projection point of the point cloud data to the neighborhood projection point is obtained, the vector is normalized, and the normalized data is obtainedWhere j ∈ [1, k ]](ii) a And calculating the weighted equivalent resultant force of the projection points of the point cloud data as follows:
Figure BDA0002246010340000035
the resultant force value indicates the distribution condition of neighborhood point cloud data of a projection point of the point cloud data, and if the resultant force value is large, the point cloud data is deviated to one side;
Figure BDA0002246010340000041
wherein ,wijIs composed ofThe weighting coefficient of (2); dijIs composed of
Figure BDA0002246010340000043
The distance from the projection point of the point cloud data to the neighborhood projection point of the point cloud data;
the average of the weighted equivalent resultant forces is calculated as follows:
Figure BDA0002246010340000044
wherein, the resultant force average value represents the average distribution condition of the point cloud in the neighborhood of each projection point of the point cloud data set;
if FiIf the point cloud data is more than F, the point cloud data is regarded as boundary points and is reserved, and a boundary reduced subset PSet1 is obtained.
Further, the step S4 is specifically: for non-boundary points, estimating local density rho in k neighborhood to which each point cloud data belongsi(ii) a According to rhoiMaximum value of (p)maxAnd minimum value ρminDividing point cloud data of non-boundary points into under-feature point set S1Feature point set S2And rich feature point set S3The division is calculated as follows:
Figure BDA0002246010340000045
wherein ,PiIs random non-boundary point cloud data.
Further, the local density ρ i is calculated specifically as follows:
Figure BDA0002246010340000046
wherein ,
Figure BDA0002246010340000047
is a vector starting from the point cloud data to the neighborhood point cloud,
Figure BDA0002246010340000048
the smaller the mean value of (1) is, the denser the neighborhood where the point cloud data is located is, the larger rho i is.
Further, the step S5 is specifically: respectively calculating the normal vector of each point cloud data in each subset through a principal component analysis methodAnd calculating the weighted average value theta of the included angle between each point cloud data and the normal vector of each point in the neighborhood according to the distance weightiThe calculation is as follows:
Figure BDA00022460103400000410
wherein ,||pij| | is the Euclidean distance from the point cloud data to each point in the neighborhood,normal vectors of all points in the point cloud data neighborhood are obtained;
further, the step S6 is specifically: calculating a local characteristic factor of the point cloud data according to the local density rho i, wherein the calculation is as follows:
Figure BDA0002246010340000051
wherein ,
Figure BDA0002246010340000052
representing belonging to subset SuAnd u belongs to [1,2,3 ]]；DiLocal characteristic factors of the point cloud data;
calculating the average local characteristic factor of the subset to which each point cloud data belongs according to the local characteristic factor of each point cloud data:
Figure BDA0002246010340000053
wherein ,
Figure BDA0002246010340000054
is a subset SuAverage local feature factor of (2);
respectively as an under-feature point set S1Feature point set S2And rich feature point set S3Selecting a threshold value muTHI.e. corresponding to thresholds respectively of
Figure BDA0002246010340000055
The threshold value muTHThe value is specifically as follows: under feature point set S1Threshold value of
Figure BDA0002246010340000056
Preferably 0.2, feature point set S2Threshold value of
Figure BDA0002246010340000057
Preferably 0.4, and a feature-rich point set S3Threshold value of
Figure BDA0002246010340000058
Preferably 0.6, the non-feature points, feature points and feature-rich points are excessively smooth, and excessive aggregation of holes and details is avoided.
When D is presenti/D＞μTHAnd if not, deleting the point cloud data, and combining the point cloud data sets of which the subsets are reserved to obtain the characteristic point reduced subset PSet 2.
Compared with the prior art, the invention has the following advantages and beneficial effects:
the method adopts rasterization and local k-d trees to establish the topological structure of the point cloud, avoids overlong time for establishing the global k-d tree, adopts distance weighting to avoid the judgment influence of long-distance neighborhood point cloud data on boundary points and characteristic points, adopts density weighting to avoid the hole phenomenon possibly generated by nonuniform point cloud data characteristic point extraction, and is suitable for a closed point cloud model with obvious boundary characteristics.
Drawings
FIG. 1 is a flow chart of a point cloud reduction method for preserving detail and boundary features according to the present invention;
FIG. 2 is a schematic diagram illustrating non-boundary point determination in the embodiment of the present invention;
FIG. 3 is a schematic diagram illustrating boundary point determination according to the embodiment of the present invention;
FIG. 4 is a diagram of an original model of a point cloud according to an embodiment of the present invention;
FIG. 5 is a simplified point cloud model diagram according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to examples and drawings, but the present invention is not limited thereto.
Example (b):
a point cloud reduction method with details and boundary features retained, the flow of which is shown in FIG. 1, includes the following steps,
firstly, reading original point cloud data with the data point number of N, rasterizing the point cloud data, and establishing a local k-d tree, namely dividing the side length L of a cube grid according to the random sampling density rho of the point cloud data, wherein the side length L of the cube grid is the minimum m times of the sampling density rho, and the value range of m is 1.2-1.5; the method comprises the following steps of dividing a point cloud data bounding box into space grids, wherein the point cloud data bounding box is divided into the space grids specifically: grid number M for dividing point cloud data according to x, y and z directionsx,My,MzThe calculation is as follows:
Figure BDA0002246010340000061
wherein ,xmaxIs the maximum value in the x direction of the coordinate, xminIs the minimum value in x direction of the coordinate, ymaxIs the maximum value in the y direction of the coordinate, yminIs the minimum value in the y direction of the coordinate, zmaxIs the maximum value in the z direction of the coordinate, zminIs the minimum value in the z direction of the coordinate;
Figure BDA0002246010340000062
to get the whole, δ is the amount of dilation;
the grid coordinate calculation of the point cloud data after rasterization is as follows:
Figure BDA0002246010340000063
wherein (x, y, z) is the geometrical coordinate of the point cloud data,
Figure BDA0002246010340000064
to get the whole, δ is the amount of dilation;
establishing a k neighborhood for each point cloud data, wherein the k value range is 15-20;
the sampling density ρ is obtained by the following calculation:
Figure BDA0002246010340000065
wherein n is the number of point cloud data taken out randomly, diIs the interval from the closest point of the point cloud data. The value range of n is 20-30.
Establishing a local k-d tree, specifically establishing a k-d tree only in a grid where point cloud data is located and a neighborhood grid of the point cloud data; wherein the neighborhood grid of each point cloud data is defined as (l)x±i,ly±j,lz±k)，i,j,k∈[-1,1]。
Second, for each point cloud data piAnd its k-neighborhood points fit a tangent plane, where the fitting of the smallest tangent plane uses the least squares method. And the point cloud data piAnd point projection of k neighborhood thereofTo the micro-tangent plane, the point cloud data p is obtainediStarting from the projection point P to the vector of the projection point in the neighborhood, normalizing the vector to obtain normalized dataWhere j ∈ [1, k ]]；
Thirdly, calculating the weighted equivalent resultant force F of the projection point P of the point cloud dataiThe calculation is as follows:
Figure BDA0002246010340000072
the resultant force value represents the distribution condition of the point cloud in the neighborhood of the projection point P, and if the resultant force value is large, the point cloud data is deviated to one side;
Figure BDA0002246010340000073
wherein ,wijIs composed ofThe weighting coefficient of (2); dijIs composed of
Figure BDA0002246010340000075
The distance from the projection point of the point cloud data to the neighborhood projection point of the point cloud data;
the non-boundary point determination schematic diagram is shown in fig. 2, the boundary point determination schematic diagram is shown in fig. 3, if the resultant force value of the k neighborhood of the projection point is larger, the projection point is deviated to one side, the projection point is determined to be the boundary point at the moment, otherwise, the distribution of the point is more uniform, the point is considered to be an inner point, and therefore whether the target point is the boundary point is judged according to the result;
calculating a weighted equivalent resultant force FiIs calculated as follows:
Figure BDA0002246010340000076
wherein, the resultant force average value represents the average distribution condition of the point cloud in the neighborhood of each projection point of the point cloud data set;
if FiIf the point cloud data is more than F, the point cloud data is regarded as boundary points, and the boundary points are reserved to obtain a reduced boundary subset PSet 1;
fourthly, for the non-boundary points, estimating the local density rho in the k neighborhood to which each point cloud data belongsi(ii) a The local density ρiThe calculation of (a) is specifically as follows:
wherein ,
Figure BDA0002246010340000078
is a vector starting from the point cloud data to the neighborhood point cloud,
Figure BDA0002246010340000079
the smaller the mean value of (a), the denser the neighborhood where the point cloud data is located, the rhoiThe larger;
according to rhoiMaximum value of (p)maxAnd minimum value ρminDividing point cloud data of non-boundary points into under-feature point set S1Feature point set S2And rich feature point set S3The division is calculated as follows:
Figure BDA0002246010340000081
wherein ,PiIs random non-boundary point cloud data corresponding to neighborhood density as rhoiPoint (2) of (c).
Fifthly, respectively calculating the normal vector of each point cloud data in each subset
Figure BDA0002246010340000082
Where the subset includes a set S of under-feature points1Feature point set S2And rich feature point set S3(ii) a (ii) a Here, the normal vector of the point cloud data
Figure BDA0002246010340000083
The estimation method adopts a principal component analysis method, and calculates the weighted average value theta of the included angle between each point cloud data and the normal vector of each point in the neighborhood according to the distance weightiThe calculation is as follows:
wherein ,||pij| | is the Euclidean distance from the point cloud data to each point in the neighborhood,
Figure BDA0002246010340000085
normal vectors of all points in the point cloud data neighborhood are obtained;
sixthly, according to the local density rhoiCalculating the local characteristic factor of the point cloud data as follows:
Figure BDA0002246010340000086
wherein ,
Figure BDA0002246010340000087
representing belonging to subset SuAnd u belongs to [1,2,3 ]]；DiLocal characteristic factors of the point cloud data;
calculating the average local characteristic factor of the subset to which each point cloud data belongs according to the local characteristic factor of each point cloud data:
Figure BDA0002246010340000088
wherein ,
Figure BDA0002246010340000089
is a subset SuAverage local feature factor of (2);
respectively as an under-feature point set S1Feature point set S2And rich feature point set S3Selecting a threshold value muTHI.e. corresponding to thresholds respectively ofThe threshold value muTHThe value is specifically as follows: under feature point set S1Threshold value ofPreferably 0.2, feature point set S2Threshold value of
Figure BDA00022460103400000812
Preferably 0.4, and a feature-rich point set S3Threshold value of
Figure BDA00022460103400000813
Preferably 0.6.
When D is presenti/D＞μTHIf so, retaining the point cloud data, otherwise deleting the point cloud data, and combining the point cloud data sets retained by each subset to obtain a characteristic point reduced subset PSet 2;
and seventhly, combining the boundary reduced subset PSet1 and the characteristic point reduced subset PSet2, wherein the problem of deleting repeated points does not exist in the combination, and the combination is directly carried out to obtain the reduced point cloud data set PSet.
The point cloud with 10000 cats shown in fig. 4 is subjected to 4400 point reduction by combining the above steps, and the final reduction result is shown in fig. 5. As can be seen from the figure, the method provided by the invention well keeps the local and boundary characteristics of the point cloud data and has beneficial effects on precision and simplicity.
The above embodiments are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments, and any other changes, modifications, substitutions, combinations, and simplifications which do not depart from the spirit and principle of the present invention should be construed as equivalents thereof, and all such changes, modifications, substitutions, combinations, and simplifications are intended to be included in the scope of the present invention.

Claims (10)

1. A point cloud simplification method for retaining details and boundary features is characterized by comprising the following steps:
s1, reading the original point cloud model data to obtain original point cloud data;
s2, rasterizing the original point cloud data, and establishing a local k-d tree, wherein the k-d tree is established for each point cloud data only in the grid where the point cloud data is located and the neighborhood grid of the point cloud data; wherein the neighborhood grid of each point cloud data is defined as (l)x±i,ly±j,lz±k)，i,j,k∈[-1,1]；
S3, fitting a minimum tangent plane to the rasterized point cloud data and k neighborhood points thereof, projecting the point cloud data and the k neighborhood points thereof onto the tangent plane, calculating vectors from projection points of the point cloud data to neighborhood projection points, and normalizing the vectors to obtain normalized data
Figure FDA0002246010330000012
Calculating a weighted equivalent resultant force F of the projected points of the point cloud dataiFurther calculating the average value F of the weighted equivalent resultant force, if FiIf the point cloud data is more than F, the point cloud data is regarded as boundary points, and the boundary points are reserved to obtain a reduced boundary subset PSet 1;
s4, for the non-boundary points, estimating the local density rho in the k neighborhood to which each point cloud data belongsiAnd partitioning to obtain an under-feature point set S1Feature point set S2And rich feature point set S3；
S5, calculating normal vectors of non-boundary points, and calculating a weighted average value theta of included angles between each point cloud data and normal vectors of each point in the neighborhood according to distance weightsi；
S6, according to the local density rhoiCalculating a local feature factor DiFurther, the average values D are calculated and are respectively the under feature point set S1Feature point set S2And rich feature point set S3Selecting a threshold value muTHI.e. corresponding to thresholds respectively of
Figure FDA0002246010330000013
When D is presenti/D＞μTHIf not, deleting the point cloud data, and combining the retained subsetsThe point cloud data set obtains a reduced subset of feature points PSet 2;
and S7, merging the boundary reduced subset PSet1 and the feature point reduced subset PSet2 to obtain a reduced point cloud data set PSet.
2. The method of claim 1, wherein the step S2 specifically includes: dividing the side length L of a cube grid according to the random sampling density rho of the point cloud data, wherein the minimum side length L of the cube grid is m times of the sampling density rho, dividing the point cloud data bounding box into space grids, and establishing a k neighborhood for each point cloud data.
3. The method of claim 2, wherein the point cloud data bounding box is partitioned into spatial grids, specifically: grid number M for dividing point cloud data according to x, y and z directionsx,My,MzThe calculation is as follows:
Figure FDA0002246010330000011
wherein ,xmaxIs the maximum value in the x direction of the coordinate, xminIs the minimum value in x direction of the coordinate, ymaxIs the maximum value in the y direction of the coordinate, yminIs the minimum value in the y direction of the coordinate, zmaxIs the maximum value in the z direction of the coordinate, zminIs the minimum value in the z direction of the coordinate;
Figure FDA0002246010330000021
to round, δ is the expansion number.
4. The method of claim 1, wherein the sampling density p is obtained by calculating:
wherein n is the number of point cloud data taken out randomly, the value range of n is 20-30, and diIs the interval from the closest point of the point cloud data.
5. The method of claim 1, wherein the step S3 specifically includes: fitting a tangent plane to each point cloud data and k neighborhood points thereof, wherein the fitting of the minimum tangent plane uses a least square method to project the point cloud data and the k neighborhood points thereof onto the tangent plane, the vector from the projection point of the point cloud data to the neighborhood projection point is obtained, the vector is normalized, and the normalized data is obtained
Figure FDA0002246010330000023
Where j ∈ [1, k ]](ii) a And calculating the weighted equivalent resultant force of the projection points of the point cloud data as follows:
Figure FDA0002246010330000024
wherein ,wijIs composed of
Figure FDA0002246010330000026
The weighting coefficient of (2); dijIs composed of
Figure FDA0002246010330000027
The distance from the projection point of the point cloud data to the neighborhood projection point of the point cloud data;
the average of the weighted equivalent resultant forces is calculated as follows:
wherein N is the total number of the point cloud data;
if FiIf the point cloud data is more than F, the point cloud data is regarded as boundary points and is reserved, and a boundary reduced subset PSet1 is obtained.
6. The method of claim 1, wherein the step S4 specifically includes: for non-boundary points, estimating local density rho in k neighborhood to which each point cloud data belongsi(ii) a According to rhoiMaximum value of (p)maxAnd minimum value ρminDividing point cloud data of non-boundary points into under-feature point set S1Feature point set S2And rich feature point set S3The division is calculated as follows:
Figure FDA0002246010330000031
wherein ,PiIs random non-boundary point cloud data.
7. The method of claim 1, wherein the local density p is a point cloud reduction method with details and boundary features preservediThe calculation of (a) is specifically as follows:
Figure FDA0002246010330000032
wherein ,
Figure FDA0002246010330000033
is a vector starting from the point cloud data to the neighborhood point cloud.
8. The method of claim 1, wherein the step S5 specifically includes: respectively calculating the normal vector of each point cloud data in each subset through a principal component analysis method
Figure FDA0002246010330000034
And calculating the weighted average value theta of the included angle between each point cloud data and the normal vector of each point in the neighborhood according to the distance weightiThe calculation is as follows:
wherein ,||pij| | is the Euclidean distance from the point cloud data to each point in the neighborhood,
Figure FDA0002246010330000036
and (4) obtaining normal vectors of all points in the point cloud data neighborhood.
9. The method of claim 1, wherein the step S6 specifically includes: according to local density ρiCalculating the local characteristic factor of the point cloud data as follows:
Figure FDA0002246010330000037
wherein ,
Figure FDA0002246010330000038
representing belonging to subset SuAnd u belongs to [1,2,3 ]]；DiLocal characteristic factors of the point cloud data;
calculating the average local characteristic factor of the subset to which each point cloud data belongs according to the local characteristic factor of each point cloud data:
wherein ,
Figure FDA00022460103300000310
is a subset SuAverage local feature factor of (2);
respectively as an under-feature point set S1Feature point set S2And rich feature point set S3Selecting a threshold value muTHI.e. corresponding to thresholds respectively of
Figure FDA00022460103300000311
When D is presenti/D＞μTHAnd if not, deleting the point cloud data, and combining the point cloud data sets of which the subsets are reserved to obtain the characteristic point reduced subset PSet 2.
10. The method of claim 1, wherein the threshold μ is set to a value that is less than a threshold valueTHThe value is specifically as follows: under feature point set S1Threshold value of
Figure FDA0002246010330000041
Is 0.2, feature point set S2Threshold value of
Figure FDA0002246010330000042
Is 0.4, and a feature-rich point set S3Threshold value ofIs 0.6.
CN201911016983.7A 2019-10-24 2019-10-24 Point cloud simplifying method for retaining details and boundary characteristics Active CN110807781B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911016983.7A CN110807781B (en) 2019-10-24 2019-10-24 Point cloud simplifying method for retaining details and boundary characteristics

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911016983.7A CN110807781B (en) 2019-10-24 2019-10-24 Point cloud simplifying method for retaining details and boundary characteristics

Publications (2)

Publication Number Publication Date
CN110807781A true CN110807781A (en) 2020-02-18
CN110807781B CN110807781B (en) 2023-09-29

Family

ID=69489085

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911016983.7A Active CN110807781B (en) 2019-10-24 2019-10-24 Point cloud simplifying method for retaining details and boundary characteristics

Country Status (1)

Country Link
CN (1) CN110807781B (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111652855A (en) * 2020-05-19 2020-09-11 西安交通大学 Point cloud simplification method based on survival probability
CN111815611A (en) * 2020-07-14 2020-10-23 南京航空航天大学苏州研究院 Round hole feature extraction method for rivet hole measurement point cloud data
CN112614216A (en) * 2020-12-04 2021-04-06 大连理工大学 Variable-curvature self-adaptive point cloud data down-sampling method
CN113744389A (en) * 2021-08-24 2021-12-03 武汉理工大学 Point cloud simplification method for complex part curved surface feature retention
CN114022617A (en) * 2021-11-18 2022-02-08 中国科学院长春光学精密机械与物理研究所 Method for discriminating scattered point cloud hole boundary
CN114119628A (en) * 2021-10-25 2022-03-01 南京航空航天大学 Point cloud accurate segmentation method based on feature template
CN116310227A (en) * 2023-05-18 2023-06-23 海纳云物联科技有限公司 Three-dimensional dense reconstruction method, three-dimensional dense reconstruction device, electronic equipment and medium
CN114022617B (en) * 2021-11-18 2024-04-30 中国科学院长春光学精密机械与物理研究所 Method for judging hole boundaries of scattered point cloud

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104361625A (en) * 2014-07-24 2015-02-18 西北农林科技大学 Ray principle based cloud data compaction algorithm with boundary reservation
CN104616349A (en) * 2015-01-30 2015-05-13 天津大学 Local curved surface change factor based scattered point cloud data compaction processing method
CN108198244A (en) * 2017-12-20 2018-06-22 中国农业大学 A kind of Apple Leaves point cloud compressing method and device
CN109410342A (en) * 2018-09-28 2019-03-01 昆明理工大学 A kind of point cloud compressing method retaining boundary point

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104361625A (en) * 2014-07-24 2015-02-18 西北农林科技大学 Ray principle based cloud data compaction algorithm with boundary reservation
CN104616349A (en) * 2015-01-30 2015-05-13 天津大学 Local curved surface change factor based scattered point cloud data compaction processing method
CN108198244A (en) * 2017-12-20 2018-06-22 中国农业大学 A kind of Apple Leaves point cloud compressing method and device
CN109410342A (en) * 2018-09-28 2019-03-01 昆明理工大学 A kind of point cloud compressing method retaining boundary point

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111652855A (en) * 2020-05-19 2020-09-11 西安交通大学 Point cloud simplification method based on survival probability
CN111815611A (en) * 2020-07-14 2020-10-23 南京航空航天大学苏州研究院 Round hole feature extraction method for rivet hole measurement point cloud data
CN111815611B (en) * 2020-07-14 2024-03-22 南京航空航天大学 Round hole feature extraction method for rivet hole measurement point cloud data
CN112614216A (en) * 2020-12-04 2021-04-06 大连理工大学 Variable-curvature self-adaptive point cloud data down-sampling method
CN112614216B (en) * 2020-12-04 2022-10-04 大连理工大学 Variable-curvature self-adaptive point cloud data down-sampling method
CN113744389A (en) * 2021-08-24 2021-12-03 武汉理工大学 Point cloud simplification method for complex part curved surface feature retention
CN113744389B (en) * 2021-08-24 2023-10-10 武汉理工大学 Point cloud simplifying method for complex part curved surface feature preservation
CN114119628A (en) * 2021-10-25 2022-03-01 南京航空航天大学 Point cloud accurate segmentation method based on feature template
CN114022617A (en) * 2021-11-18 2022-02-08 中国科学院长春光学精密机械与物理研究所 Method for discriminating scattered point cloud hole boundary
CN114022617B (en) * 2021-11-18 2024-04-30 中国科学院长春光学精密机械与物理研究所 Method for judging hole boundaries of scattered point cloud
CN116310227A (en) * 2023-05-18 2023-06-23 海纳云物联科技有限公司 Three-dimensional dense reconstruction method, three-dimensional dense reconstruction device, electronic equipment and medium
CN116310227B (en) * 2023-05-18 2023-09-12 海纳云物联科技有限公司 Three-dimensional dense reconstruction method, three-dimensional dense reconstruction device, electronic equipment and medium

Also Published As

Publication number Publication date
CN110807781B (en) 2023-09-29

Similar Documents

Publication Publication Date Title
CN110807781B (en) Point cloud simplifying method for retaining details and boundary characteristics
CN108830931B (en) Laser point cloud simplification method based on dynamic grid k neighborhood search
CN111986115A (en) Accurate elimination method for laser point cloud noise and redundant data
Wang et al. Segmentation of LiDAR point clouds for building extraction
CN105139379B (en) Based on the progressive extracting method of classified and layered airborne Lidar points cloud building top surface
CN109685080B (en) Multi-scale plane extraction method based on Hough transformation and region growth
CN110838115B (en) Ancient cultural relic three-dimensional model change detection method by contour line extraction and four-dimensional surface fitting
CN112347550B (en) Coupling type indoor three-dimensional semantic graph building and modeling method
CN107862735B (en) RGBD three-dimensional scene reconstruction method based on structural information
CN111508073B (en) Method for extracting roof contour line of three-dimensional building model
CN110047133A (en) A kind of train boundary extraction method towards point cloud data
CN114200477A (en) Laser three-dimensional imaging radar ground target point cloud data processing method
CN111932669A (en) Deformation monitoring method based on slope rock mass characteristic object
CN111783722B (en) Lane line extraction method of laser point cloud and electronic equipment
CN114332291A (en) Oblique photography model building outer contour rule extraction method
CN110942077B (en) Feature line extraction method based on weight local change degree and L1 median optimization
CN107610174B (en) Robust depth information-based plane detection method and system
CN112634457A (en) Point cloud simplification method based on local entropy of Hausdorff distance and average projection distance
CN109345571B (en) Point cloud registration method based on extended Gaussian image
CN112767429B (en) Ground-snow surface point cloud rapid segmentation method
CN116721228B (en) Building elevation extraction method and system based on low-density point cloud
CN107122782B (en) Balanced semi-dense stereo matching method
Omidalizarandi et al. Segmentation and classification of point clouds from dense aerial image matching
CN112270746A (en) Aluminum alloy 3D printing point cloud simplification algorithm based on neighborhood covariance characteristic parameter threshold
Kim et al. New approach for planar patch segmentation using airborne laser data

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant