CN113628263A - Point cloud registration method based on local curvature and neighbor characteristics thereof - Google Patents

Point cloud registration method based on local curvature and neighbor characteristics thereof Download PDF

Info

Publication number
CN113628263A
CN113628263A CN202111040513.1A CN202111040513A CN113628263A CN 113628263 A CN113628263 A CN 113628263A CN 202111040513 A CN202111040513 A CN 202111040513A CN 113628263 A CN113628263 A CN 113628263A
Authority
CN
China
Prior art keywords
point
points
point cloud
registration
feature
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
CN202111040513.1A
Other languages
Chinese (zh)
Inventor
李明昊
王毅
张明理
何新
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.)
Xian Thermal Power Research Institute Co Ltd
Original Assignee
Xian Thermal Power Research Institute Co Ltd
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 Xian Thermal Power Research Institute Co Ltd filed Critical Xian Thermal Power Research Institute Co Ltd
Priority to CN202111040513.1A priority Critical patent/CN113628263A/en
Publication of CN113628263A publication Critical patent/CN113628263A/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
    • G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
    • G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
    • 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)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a point cloud registration method based on local curvature and neighbor characteristics thereof, which introduces the local curvature and related characteristic information of a point cloud into a point cloud characteristic extraction process and improves the precision and speed of point cloud registration. Firstly, a point cloud down-sampling algorithm is completed based on point density distribution to reduce the density of the point cloud, and a radius filtering algorithm is used for eliminating point cloud noise points; secondly, by extracting curvature values of the detection points, the quasi-plane distance formed by the detection points and the adjacent points of the detection points and the distance standard deviation of the fitting plane of the detection points and the adjacent points, combining three characteristic parameters to complete the parameter construction work of point cloud characteristics, and using a fast point characteristic histogram FPFH as a characteristic descriptor; and finally, solving the optimal rotation matrix R and translation matrix t by combining the SAC-IA sampling consistency initial registration algorithm and the ICP iterative closest point algorithm to complete the whole point cloud registration process.

Description

Point cloud registration method based on local curvature and neighbor characteristics thereof
Technical Field
The invention belongs to the technical field of computer vision, and particularly relates to a point cloud registration method based on local curvature and neighbor characteristics of the local curvature.
Background
The application of the point cloud in the three-dimensional vision technology is more and more extensive, the point cloud data acquisition tool is developed rapidly, the data acquisition cost is reduced, and the efficiency is improved, so that the point cloud technology is applied to the fields of machine vision, cultural relic restoration, automatic driving, unmanned aerial vehicles and the like. The main work of point cloud registration is to rotate and translate point clouds in different coordinate systems, and finally integrate the point clouds into a complete point cloud in the same coordinate system. The method is characterized in that the method is divided into non-rigid registration and rigid registration according to the properties of target point cloud and source point cloud, the non-rigid registration refers to non-rigid transformation such as deformation and stretching between the target point cloud and the source point cloud, in an actual application scene, partial registration problems need to use the non-rigid registration, such as medical diagnosis and face recognition, and a great development space is left for the research of the non-rigid registration; rigid body registration means that deformation does not exist between source point cloud and target point cloud, only rigid transformation of rotation and translation occurs, and the actual application range is wider, but the existing point cloud registration technology has the defects of low registration speed, easy falling into local optimal solution and the like. How to complete point cloud registration quickly and with high quality is still a problem with important research value.
Disclosure of Invention
In order to overcome the problems in the prior art, the invention aims to provide a point cloud registration method based on local curvature and neighbor characteristics thereof.
In order to achieve the purpose, the invention adopts the following technical scheme:
a point cloud registration method based on local curvature and neighbor characteristics thereof comprises the following steps:
step 1, point cloud data are preprocessed, so that the original characteristics of the point cloud are not damaged, the density of the point cloud is reduced, and noise points in the point cloud are removed;
step 2, point cloud characteristic parameter construction, wherein characteristic parameters are constructed jointly by combining the average curvature of the detection points, the Euclidean distance between the detection points and the adjacent points to a fitting plane and the standard deviation of the distance between the detection points and the adjacent points to the fitting plane, and characteristic points are extracted;
step 3, point cloud registration, namely calculating a feature descriptor by using an FPFH (fast point feature histogram) algorithm and finishing point cloud initial registration work by combining with an SAC-IA (sample consensus-IA) sampling consistency initial registration algorithm, wherein point cloud fine registration uses an ICP (iterative closest point) algorithm;
the point cloud pair in the point cloud registration is divided into a target point cloud and a source point cloud, and the target point cloud and the source point cloud use complete paired point clouds or multi-view point clouds collected by a laser radar, so that the method is suitable for small-size point cloud data or large-size industrial point cloud data.
The step 1 specifically comprises the following steps:
step 1.1: performing three-dimensional voxel grid processing on the point cloud, scattering point cloud data in voxel grids, and replacing other points in the voxel grids by using points closest to the center of gravity of the voxel grids in each voxel grid, wherein the method reduces the number of the point clouds and ensures that the microscopic form of the point cloud is unchanged;
step 1.2: and setting different search radiuses r according to the density distribution of the neighboring point cloud, calculating the number of points in the sphere with the radius r, and if the number is smaller than a threshold value, rejecting the noise points.
The step 2 specifically comprises the following steps:
step 2.1: calculating the average curvature H of the probe points, described as equation (1):
Figure BDA0003248932950000021
wherein:
k1-a maximum principal curvature;
k2-a minimum principal curvature;
step 2.2: calculating a fitting plane formed by k adjacent points of the detection points according to the second-order multiplication, wherein the plane is as shown in an equation (2):
Ax+By+Cz+D＝0 (2)
wherein:
A. b, C, D-is a known constant and A, B, C is not simultaneously zero;
and (3) calculating the Euclidean distance d from the detection point to the fitting plane, wherein the Euclidean distance d is expressed by the formula (3):
Figure BDA0003248932950000031
wherein:
d is the Euclidean distance from the detection point to the fitting plane of the adjacent points;
x1,y1,z1-as probe point coordinates;
step 2.3: calculating the standard deviation sigma of the projection distance between the detected point and the k adjacent points to the fitting plane, and describing the standard deviation sigma as formula (4):
Figure BDA0003248932950000032
wherein:
ri-the distance of the probe point to each of the neighboring points to the fitting plane;
mu is the average value of the distance from each detection point to each adjacent point to the fitting plane;
k is the sum of the number of the detection points and the adjacent points;
step 2.4: the feature parameter β is constructed from the three parameters of steps 2.1, 2.2, 2.3 together, as described by equation (5):
β＝σ·H·d，Δd>0 (5)
wherein:
sigma is the standard deviation of the projection distance from the detected point and k adjacent points to the fitting plane;
h-mean curvature;
Δ d-the difference between d and μ;
step 2.5: according to the characteristic parameter beta provided in the step 2.4, when delta d is larger than 0, the detection point is the outermost point in the adjacent points of the area and has characteristics; when the delta d is smaller than 0, the detection point and the adjacent points are mutually interwoven and cannot be defined as characteristic points, so that the points with the delta d smaller than 0 are firstly removed in the calculation; in addition, points where β is smaller than the average value of β are eliminated, and the remaining points are used as feature points.
The step 3 specifically comprises the following steps:
step 3.1: calculating a Fast Point Feature Histogram (FPFH) according to the spatial difference between the extracted feature points and the neighboring feature points thereof, accurately describing the spatial geometric attributes in the point neighborhood, firstly defining a three-dimensional coordinate system uvw, calculating the deviation between each point in the k neighborhood of the point P and an estimated normal, wherein the coordinate system uvw is described as formula (6):
Figure BDA0003248932950000041
wherein:
ns——Psa surface normal of the point;
Pt-a probe point;
Ps-any close neighborhood of the probe point;
under the uvw coordinate system, n is calculateds、ntThe angular deviations α, β, θ between, referred to as PsFeature of point three elements, putting three elements into histogram to form PsSimplified Point Feature Histogram (SPFH) of points, described as formula (7):
Figure BDA0003248932950000051
wherein:
q——Psand PtThe Euclidean distance between two points;
nt——Pta surface normal of the point;
α——ntthe included angle with the v axis;
β——Ptto PsVector of points and nsThe included angle of (A);
θ——ntthe included angle between the projection vector on the u plane and the w plane and the u axis;
calculating a probe point PtAnd calculating the corresponding simplified point feature histogram SPFH (specific pathogen free) with the feature three elements of each adjacent point in the k neighborhood, and calculating the feature histogram of each simplified point in the neighborhoodThe graph SPFH is weighted and counted to obtain a fast point feature histogram FPFH, which is described as formula (8):
Figure BDA0003248932950000052
wherein:
FPFH(Pt)——Pta fast point feature histogram of points;
SPFH(Pt)——Ptsimplified point feature histograms of points;
m——Ptnumber of neighbors of a point;
Pm——Pteach of the neighbors of a point;
wm——Ptand neighbor point PmThe distance of (d);
SPFH(Pm)——Pmsimplified point feature histograms of points;
step 3.2: the SAC-IA sampling consistency initial registration algorithm uses an FPFH (flat-panel display and frequency hopping) feature descriptor as input, reduces iteration times, searches for similar features between source point cloud and target point cloud as corresponding points through a KD (K-dimension) tree, traverses and calculates the minimum error of the registration result of the SAC-IA algorithm each time, when the minimum error is obtained by calculating transformation quality, the rotation and translation matrix at the time is the optimal solution of initial registration, and the transformation quality D is described as the formula (9):
Figure BDA0003248932950000061
wherein:
D(ei) -transform quality of group i;
ei-representing an error value between corresponding points of the ith group;
te-a predetermined value;
step 3.3: searching points with the shortest distance in the two groups of point cloud sets by using an ICP iterative closest point algorithm, and iteratively solving an optimal rigid transformation matrix;
step 3.4: according to the RMSE root mean square error between corresponding point pairs after fine registration, the final completion degree of point cloud registration is measured, the larger the RMSE root mean square is, the larger the error is, otherwise, the smaller the error is, and the formula (10) is described;
Figure BDA0003248932950000062
wherein:
RMSE (P, Q) -P, Q Overall root mean Square error of the two sets of point clouds;
n-P, Q total number of corresponding point pairs of two groups of point clouds;
Xi-the euclidean distance between each set of corresponding points;
Figure BDA0003248932950000071
-is the true euclidean distance between each set of corresponding points.
Compared with the prior art, the invention has the following advantages:
(1) is more reliable
In the aspect of point cloud preprocessing, different search radiuses are set according to point cloud density distribution to carry out point cloud down-sampling operation, and the defect that cavities are easy to generate after down-sampling due to uneven density distribution is overcome.
(2) More accurate
Compared with a SAC-IA initial registration algorithm which directly uses the FPFH feature descriptors in combination with the SAC-IA algorithm, the feature points in the invention can reduce the root mean square error by about 70 percent by combining the FPFH feature descriptors with the SAC-IA algorithm; meanwhile, the accuracy of SAC-IA initial registration is higher than that of ISS, Harris and other characteristic points combined with FPFH characteristic descriptors.
(3) More rapidly
The feature point extraction algorithm is superior to ISS algorithm, Harris algorithm and SIFT algorithm in extraction speed, and the speed is respectively increased by 84%, 55% and 93%.
Drawings
Fig. 1a, fig. 1b and fig. 1c are respectively a gaussian noise effect graph added to a face point cloud, a noise reduction effect graph of the noisy face point cloud by using a fixed radius filtering algorithm, and an adaptive radius filtering effect graph of the noisy face point cloud according to density distribution.
FIG. 2 features a computational flow diagram.
Fig. 3a, 3b, 3c and 3d are diagrams of rabbit feature extraction effects using the feature extraction algorithm, the ISS feature extraction algorithm, the SIFT feature extraction algorithm and the Harris feature extraction algorithm of the present invention, respectively.
FIG. 4 is a time-consuming diagram of the feature extraction algorithm.
Fig. 5a, 5b, 5c and 5d are respectively a Bunny point cloud data registration effect graph using the original point cloud data, the feature point set extracted by the invention, the feature point set extracted by the ISS algorithm, and the feature point set extracted by the Harris algorithm as registration input sets.
Detailed Description
The invention is described in detail below with reference to the figures and examples.
The method comprises the following steps of taking point cloud data of a human face, Cat and Cow, point cloud data of Bunny and Dragon of Stanford university as input for research, comparing ISS, SIFT and Harris feature extraction algorithms with the feature extraction algorithm provided by the invention, carrying out feature description by using an FPFH feature descriptor, carrying out an initial registration experiment by using an SAC-IA algorithm, and carrying out a fine registration experiment by using a classical ICP algorithm to realize the whole point cloud registration process, wherein the method comprises the following steps:
step 1, carrying out point cloud pretreatment on experimental data.
Because the density distribution of the face data is uneven, and cavity areas are easy to appear when radius filtering is carried out by using a fixed radius value, the invention combines the density distribution condition of point cloud, and carries out radius filtering processing on different density areas by using different radii r, and the preprocessing effect display is carried out by using the noisy face point cloud, the fixed radius filtering and the density distribution radius filtering respectively in the graph 1a, the graph 1b and the graph 1 c.
And 2, extracting and describing features.
Firstly, a characteristic extraction stage is carried out, wherein the curvature of a detection point, the Euclidean distance from the detection point to a fitting plane of adjacent points and the distance standard from the detection point to the adjacent points to the fitting plane are extractedAnd (3) constructing feature parameters by using the three parameters, and then automatically screening feature points to obtain feature point sets P of the target point cloud and the source point cloud respectivelyt、Ps。
Secondly, a feature histogram is calculated for the spatial difference between the extracted feature point and the neighboring feature point by using a fast point feature histogram FPFH, so as to accurately describe the spatial geometric attribute in the point neighborhood, and a flow chart of the step 1 and the step 2 is shown in FIG. 2.
Fig. 3a, fig. 3b, fig. 3c and fig. 3d respectively show the effect of using the feature extraction algorithm, the ISS algorithm, the SIFT algorithm and the Harris algorithm of the present invention for Bunny point cloud data, and it can be seen from the diagrams: the extracted feature points are concentrated on the parts with obvious features, such as ears, necks, feet, tails and the like, of the Bunny point cloud data; the feature points extracted by the ISS algorithm are scattered in the whole point cloud space; most of feature points extracted by the SIFT algorithm are concentrated in a point cloud edge area; the characteristic points extracted by the Harris algorithm are scattered in the whole point cloud space and have no obvious characteristics.
Fig. 4 shows the time consumption of the algorithm for feature extraction of four groups of point cloud data of Bunny, Dragon, Cat and Cow by using the algorithm, the ISS algorithm, the Harris algorithm and the SIFT algorithm respectively, and it can be seen from the graph that: the calculation time of the feature extraction algorithm is superior to that of other three algorithms.
And 3, carrying out initial point cloud registration and fine point cloud registration.
After initial registration is carried out by using SAC-IA sampling consistency initial registration algorithm, the source point cloud and the target point cloud are basically overlapped, but a certain offset still exists between points at the moment, in order to achieve accurate registration and minimize distance error between the points, ICP iterative closest point algorithm is used, and the feature sets of the source point cloud and the target point cloud after initial registration are respectively set as Pi *And Qi *The general flow of ICP iterative closest point algorithm is as follows:
(1) computing P by constructing a KD TreeiAnd QiThe two point sets correspond to points, and the searching efficiency can be greatly improved by using the KD tree;
(2) calculating a rotation and translation matrix value according to the corresponding point pair obtained in the first step;
(3) according to the rotation translation matrix obtained in the second step, a source point cloud feature set P is obtainedi *Performing rigid body transformation to obtain a point set P after rotationi *；
(4) Calculating Pi *And Qi *When the corresponding point is smaller than a given threshold value, stopping iteration, otherwise, using Pi *’Iteration continues as a new input, stopping when the average distance meets the threshold.
Fig. 5 shows the registration results of Bunny when using different sets of points as registration input objects.
Through analysis, the feature point extraction algorithm designed by the invention reduces the calculation amount of huge point clouds, obviously improves the registration speed of the point clouds, has excellent registration precision, has better effect on small-scale point cloud data and large-scale industrial point cloud data, and has high efficiency, universality and robustness, and the feature extraction speed is superior to ISS, Harris, SIFT and other methods.

Claims (4)

1. A point cloud registration method based on local curvature and neighbor characteristics thereof is characterized in that: the method comprises the following steps:
step 1, point cloud data are preprocessed, so that the original characteristics of the point cloud are not damaged, the density of the point cloud is reduced, and noise points in the point cloud are removed;
step 2, point cloud characteristic parameter construction, wherein characteristic parameters are constructed jointly by combining the average curvature of the detection points, the Euclidean distance between the detection points and the adjacent points to a fitting plane and the standard deviation of the distance between the detection points and the adjacent points to the fitting plane, and characteristic points are extracted;
step 3, point cloud registration, namely calculating a feature descriptor by using an FPFH (fast point feature histogram) algorithm and finishing point cloud initial registration work by combining with an SAC-IA (sample consensus-IA) sampling consistency initial registration algorithm, wherein point cloud fine registration uses an ICP (iterative closest point) algorithm;
the point cloud pair in the point cloud registration is divided into a target point cloud and a source point cloud, and the target point cloud and the source point cloud use complete paired point clouds or multi-view point clouds collected by a laser radar, so that the method is suitable for small-size point cloud data or large-size industrial point cloud data.
2. The point cloud registration method based on local curvature and its neighboring features of claim 1, wherein: the step 1 specifically comprises the following steps:
step 1.1: performing three-dimensional voxel grid processing on the point cloud, scattering point cloud data in voxel grids, and replacing other points in the voxel grids by using points closest to the center of gravity of the voxel grids in each voxel grid, wherein the method reduces the number of the point clouds and ensures that the microscopic form of the point cloud is unchanged;
step 1.2: and setting different search radiuses r according to the density distribution of the neighboring point cloud, calculating the number of points in the sphere with the radius r, and if the number is smaller than a threshold value, rejecting the noise points.
3. The point cloud registration method based on local curvature and its neighboring features of claim 1, wherein: the step 2 specifically comprises the following steps:
step 2.1: calculating the average curvature H of the probe points, described as equation (1):
Figure FDA0003248932940000021
wherein:
k1-a maximum principal curvature;
k2-a minimum principal curvature;
step 2.2: calculating a fitting plane formed by k adjacent points of the detection points according to the second-order multiplication, wherein the plane is as shown in an equation (2):
Ax+By+Cz+D＝0 (2)
wherein:
A. b, C, D-is a known constant and A, B, C is not simultaneously zero;
and (3) calculating the Euclidean distance d from the detection point to the fitting plane, wherein the Euclidean distance d is expressed by the formula (3):
Figure FDA0003248932940000022
wherein:
d is the Euclidean distance from the detection point to the fitting plane of the adjacent points; x is the number of1,y1,z1-as probe point coordinates;
step 2.3: calculating the standard deviation sigma of the projection distance between the detected point and the k adjacent points to the fitting plane, and describing the standard deviation sigma as formula (4):
Figure FDA0003248932940000023
wherein:
ri-the distance of the probe point to each of the neighboring points to the fitting plane;
mu is the average value of the distance from each detection point to each adjacent point to the fitting plane;
k is the sum of the number of the detection points and the adjacent points;
step 2.4: the feature parameter β is constructed from the three parameters of steps 2.1, 2.2, 2.3 together, as described by equation (5):
β＝σ·H·d，Δd>0 (5)
wherein:
sigma is the standard deviation of the projection distance from the detected point and k adjacent points to the fitting plane;
h-mean curvature;
Δ d-the difference between d and μ;
step 2.5: according to the characteristic parameter beta provided in the step 2.4, when delta d is larger than 0, the detection point is the outermost point in the adjacent points of the area and has characteristics; when the delta d is smaller than 0, the detection point and the adjacent points are mutually interwoven and cannot be defined as characteristic points, so that the points with the delta d smaller than 0 are firstly removed in the calculation; in addition, points where β is smaller than the average value of β are eliminated, and the remaining points are used as feature points.
4. The point cloud registration method based on local curvature and its neighboring features of claim 1, wherein the step 3 is specifically:
step 3.1: calculating a fast point feature histogram FPFH according to the spatial difference between the extracted feature points and the neighboring feature points thereof, accurately describing the spatial geometric attributes in the point neighborhood, firstly defining a three-dimensional coordinate system uvw, calculating the deviation between each point in the k neighborhood of the point P and an estimated normal, wherein the coordinate system uvw is described as formula (6):
Figure FDA0003248932940000031
wherein:
ns——Psa surface normal of the point;
Pt-a probe point;
Ps-any close neighborhood of the probe point;
under the uvw coordinate system, n is calculateds、ntThe angular deviations α, β, θ between, referred to as PsFeature of point three elements, putting three elements into histogram to form PsSimplified point feature histogram SPFH of points, described as formula (7):
Figure FDA0003248932940000041
wherein:
α——ntthe included angle with the v axis;
q——Psand PtThe Euclidean distance between two points;
β——Ptto PsVector of points and nsThe included angle of (A);
nt——Pta surface normal of the point;
θ——ntthe included angle between the projection vector on the u plane and the w plane and the u axis;
calculating a probe point PtThree elements of the characteristics of each adjacent point in the k neighborhood are calculated, corresponding simplified point characteristic histogram SPFH is calculated, and the characteristics of each simplified point in the neighborhood are subjected to SPFHWeighting and counting the histogram SPFH to obtain a fast point feature histogram FPFH, which is described in a formula (8):
Figure FDA0003248932940000042
wherein:
FPFH(Pt)——Pta fast point feature histogram of points;
SPFH(Pt)——Ptsimplified point feature histograms of points;
m——Ptnumber of neighbors of a point;
Pm——Pteach of the neighbors of a point;
wm——Ptand neighbor point PmThe distance of (d);
SPFH(Pm)——Pmsimplified point feature histograms of points;
step 3.2: the SAC-IA sampling consistency initial registration algorithm uses an FPFH (flat-panel display and frequency hopping) feature descriptor as input, reduces iteration times, searches for similar features between source point cloud and target point cloud as corresponding points through a KD (K-dimension) tree, traverses and calculates the minimum error of the registration result of the SAC-IA algorithm each time, when the minimum error is obtained by calculating transformation quality, the rotation and translation matrix at the time is the optimal solution of initial registration, and the transformation quality D is described as the formula (9):
Figure FDA0003248932940000051
wherein:
D(ei) -transform quality of group i;
ei-representing an error value between corresponding points of the ith group;
te-a predetermined value;
step 3.3: searching points with the shortest distance in the two groups of point cloud sets by using an ICP iterative closest point algorithm, and iteratively solving an optimal rigid transformation matrix;
step 3.4: according to the RMSE root mean square error between corresponding point pairs after fine registration, the final completion degree of point cloud registration is measured, the larger the RMSE root mean square is, the larger the error is, otherwise, the smaller the error is, and the formula (10) is described;
Figure FDA0003248932940000052
wherein:
RMSE (P, Q) -P, Q Overall root mean Square error of the two sets of point clouds;
n-P, Q total number of corresponding point pairs of two groups of point clouds;
Xi-the euclidean distance between each set of corresponding points;
Figure FDA0003248932940000061
-is the true euclidean distance between each set of corresponding points.
CN202111040513.1A 2021-09-06 2021-09-06 Point cloud registration method based on local curvature and neighbor characteristics thereof Pending CN113628263A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111040513.1A CN113628263A (en) 2021-09-06 2021-09-06 Point cloud registration method based on local curvature and neighbor characteristics thereof

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111040513.1A CN113628263A (en) 2021-09-06 2021-09-06 Point cloud registration method based on local curvature and neighbor characteristics thereof

Publications (1)

Publication Number Publication Date
CN113628263A true CN113628263A (en) 2021-11-09

Family

ID=78389409

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111040513.1A Pending CN113628263A (en) 2021-09-06 2021-09-06 Point cloud registration method based on local curvature and neighbor characteristics thereof

Country Status (1)

Country Link
CN (1) CN113628263A (en)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114612529A (en) * 2022-03-18 2022-06-10 广东电网有限责任公司 Point cloud registration method and device, electronic equipment and storage medium
CN114627346A (en) * 2022-03-15 2022-06-14 电子科技大学 Point cloud data down-sampling method capable of retaining important features
CN115147471A (en) * 2022-06-28 2022-10-04 兰州交通大学 Laser point cloud automatic registration method based on curvature density characteristics
CN115797423A (en) * 2022-12-07 2023-03-14 中国矿业大学(北京) Registration method and system for optimizing near iteration of mine point cloud based on descriptor characteristics
CN115797418A (en) * 2022-09-27 2023-03-14 湖南科技大学 Complex mechanical part measurement point cloud registration method and system based on improved ICP
CN116563561A (en) * 2023-07-06 2023-08-08 北京优脑银河科技有限公司 Point cloud feature extraction method, point cloud registration method and readable storage medium
CN116665284A (en) * 2023-08-02 2023-08-29 深圳宇石科技有限公司 Face modeling and mask model partition matching method, device, terminal and medium
CN117111491A (en) * 2023-10-25 2023-11-24 北京赛目科技股份有限公司 Method, device and equipment for determining closest point of object in automatic driving simulation
US12094058B2 (en) 2022-08-19 2024-09-17 Fluid Dynamic Sciences, Llc Fast feature recognition and mesh generation in structural design

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114627346A (en) * 2022-03-15 2022-06-14 电子科技大学 Point cloud data down-sampling method capable of retaining important features
CN114612529B (en) * 2022-03-18 2024-09-03 广东电网有限责任公司 Point cloud registration method and device, electronic equipment and storage medium
CN114612529A (en) * 2022-03-18 2022-06-10 广东电网有限责任公司 Point cloud registration method and device, electronic equipment and storage medium
CN115147471A (en) * 2022-06-28 2022-10-04 兰州交通大学 Laser point cloud automatic registration method based on curvature density characteristics
US12094058B2 (en) 2022-08-19 2024-09-17 Fluid Dynamic Sciences, Llc Fast feature recognition and mesh generation in structural design
CN115797418A (en) * 2022-09-27 2023-03-14 湖南科技大学 Complex mechanical part measurement point cloud registration method and system based on improved ICP
CN115797418B (en) * 2022-09-27 2024-09-10 湖南科技大学 Complex mechanical part measurement point cloud registration method and system based on improved ICP
CN115797423A (en) * 2022-12-07 2023-03-14 中国矿业大学(北京) Registration method and system for optimizing near iteration of mine point cloud based on descriptor characteristics
CN116563561A (en) * 2023-07-06 2023-08-08 北京优脑银河科技有限公司 Point cloud feature extraction method, point cloud registration method and readable storage medium
CN116563561B (en) * 2023-07-06 2023-11-14 北京优脑银河科技有限公司 Point cloud feature extraction method, point cloud registration method and readable storage medium
CN116665284A (en) * 2023-08-02 2023-08-29 深圳宇石科技有限公司 Face modeling and mask model partition matching method, device, terminal and medium
CN116665284B (en) * 2023-08-02 2023-11-28 深圳宇石科技有限公司 Face modeling and mask model partition matching method, device, terminal and medium
CN117111491B (en) * 2023-10-25 2024-01-19 北京赛目科技股份有限公司 Method, device and equipment for determining closest point of object in automatic driving simulation
CN117111491A (en) * 2023-10-25 2023-11-24 北京赛目科技股份有限公司 Method, device and equipment for determining closest point of object in automatic driving simulation

Similar Documents

Publication Publication Date Title
CN113628263A (en) Point cloud registration method based on local curvature and neighbor characteristics thereof
CN114170279B (en) Point cloud registration method based on laser scanning
WO2024077812A1 (en) Single building three-dimensional reconstruction method based on point cloud semantic segmentation and structure fitting
CN107123164B (en) Three-dimensional reconstruction method and system for keeping sharp features
CN111696210A (en) Point cloud reconstruction method and system based on three-dimensional point cloud data characteristic lightweight
CN108830902A (en) A kind of workpiece identification at random and localization method based on points cloud processing
CN113298833A (en) Target object point cloud characteristic line and surface extraction method and system
CN110807781B (en) Point cloud simplifying method for retaining details and boundary characteristics
CN108022262A (en) A kind of point cloud registration method based on neighborhood of a point center of gravity vector characteristics
CN107481274B (en) Robust reconstruction method of three-dimensional crop point cloud
CN107818598B (en) Three-dimensional point cloud map fusion method based on visual correction
CN108133191A (en) A kind of real-time object identification method suitable for indoor environment
CN116402866A (en) Point cloud-based part digital twin geometric modeling and error assessment method and system
CN114240871B (en) Point cloud data processing method for contour detection in workpiece forming process
CN110222661B (en) Feature extraction method for moving target identification and tracking
CN111145129A (en) Point cloud denoising method based on hyper-voxels
CN104851095A (en) Workpiece image sparse stereo matching method based on improved-type shape context
CN115797418B (en) Complex mechanical part measurement point cloud registration method and system based on improved ICP
CN112164145A (en) Method for rapidly extracting indoor three-dimensional line segment structure based on point cloud data
CN114663373A (en) Point cloud registration method and device for detecting surface quality of part
CN117132630A (en) Point cloud registration method based on second-order spatial compatibility measurement
CN114648445B (en) Multi-view high-resolution point cloud splicing method based on feature point extraction and fine registration optimization
CN114463396B (en) Point cloud registration method utilizing plane shape and topological graph voting
CN116091727A (en) Complex Qu Miandian cloud registration method based on multi-scale feature description, electronic equipment and storage medium
CN116310355A (en) Laser point cloud denoising and defect detection method for complex structural member

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