CN117132478A - Orbit point cloud denoising method based on normal vector two-norm characteristic parameter - Google Patents
Orbit point cloud denoising method based on normal vector two-norm characteristic parameter Download PDFInfo
- Publication number
- CN117132478A CN117132478A CN202310454952.XA CN202310454952A CN117132478A CN 117132478 A CN117132478 A CN 117132478A CN 202310454952 A CN202310454952 A CN 202310454952A CN 117132478 A CN117132478 A CN 117132478A
- Authority
- CN
- China
- Prior art keywords
- point cloud
- normal vector
- characteristic
- point
- track
- 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
Links
- 239000013598 vector Substances 0.000 title claims abstract description 108
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000001914 filtration Methods 0.000 claims abstract description 32
- 230000002146 bilateral effect Effects 0.000 claims abstract description 30
- 230000014759 maintenance of location Effects 0.000 claims abstract description 12
- 238000012545 processing Methods 0.000 claims abstract description 10
- 238000004458 analytical method Methods 0.000 claims abstract description 6
- 230000008859 change Effects 0.000 claims description 13
- 230000008569 process Effects 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 20
- 238000012937 correction Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000012876 topography Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 2
- 238000011835 investigation Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 1
- 230000008602 contraction Effects 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000011002 quantification Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20028—Bilateral filtering
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Abstract
The invention discloses a method for denoising an orbit point cloud based on a normal vector two-norm characteristic parameter, which comprises the steps of performing grid-mesh processing on three-dimensional point cloud data; counting the number of point clouds under the current grid as the point cloud density, and removing discrete noise points by carrying out density threshold and connectivity analysis on the current grid; calculating the relationship between the normal vector and the normal vector of the orbital point cloud to construct a characteristic parameter of the normal vector and the normal vector of the point cloud, and dividing orbital point cloud data into a characteristic area and a non-characteristic area by using the characteristic parameter; the cosine of the normal vector included angle is used as a feature retention factor of the bilateral filtering factor to improve the bilateral filtering algorithm; carrying out improved bilateral filtering algorithm smooth denoising on the characteristic region, and carrying out least square-based plane fitting denoising treatment on the non-characteristic region; and denoising the point cloud is completed. The method can effectively remove noise data with different scales in the track point cloud data, can effectively keep the integrity of track surface area characteristic information which is easy to deform in the track point cloud, and is simple in calculation method and high in efficiency.
Description
Technical Field
The invention relates to removal of multi-scale noise of track point cloud data in track point cloud data processing, in particular to track point cloud partitioning by using a point cloud normal vector two-norm characteristic parameter, and denoising by an improved bilateral filtering algorithm and a least square fitting method.
Background
Along with the rapid development of the three-dimensional laser scanning technology, the three-dimensional laser scanning technology can rapidly acquire three-dimensional space information with high precision and high density on the surface of a target object, is widely applied to the fields of basic mapping construction, electric power facility investigation, urban building modeling, computer vision, agriculture and forestry yield investigation and the like, and realizes the reproduction of the object by reversely modeling the data acquired by scanning. However, in the process of acquiring the point cloud data of the target scene, the laser radar technology is often affected by the precision of the platform, the transceiver sensor, the detection environment, the reflection characteristic of the target and the complexity of the target scene, and the acquired point cloud data inevitably generates noise, so that the data processing precision and efficiency are seriously affected. Therefore, the point cloud denoising work becomes an indispensable ring in point cloud feature extraction and modeling.
In recent years, point cloud denoising is a research topic which is concerned by domestic and foreign scholars, zhou and the like process point cloud layering by using a layering iteration algorithm with double thresholds, and the method has the advantages of high denoising speed, high robustness and poor feature retaining effect. Jiang Tong et al divide the point cloud data into density grids of different sizes by voxelization, eliminate the cloud noise of the large-density grid points by using a region growing method, and directly remove the small-density grid points. Deng uses a DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering filtering method to divide the original point cloud into noise point cloud, ground point cloud and topography point cloud according to the space Density of the LiDAR point cloud, the characteristics of the ground point cloud and topography point cloud, so that the problem of discontinuous topography point cloud classification is solved, however, the method has a certain limitation in a relatively small railway space. Aiming at the problem that the existing point cloud filtering method can not simultaneously maintain sharp features and uneven distribution of filtering points, the Zhu provides a non-local low-rank point cloud denoising frame which is used for processing three-dimensional measurement surfaces with different scales and types of noise, but the algorithm is high in complexity. X Watanabe R provides a point cloud denoising method combining spectrum wavelet transformation and contraction for reducing errors caused in measurement and point cloud normal estimation processes, but the method has defects in removing large-range noise. For the track point cloud processing method, the track point cloud type and the noise classification problem are not fully considered, and the track point cloud processing method cannot be effectively applied to track point cloud data.
Disclosure of Invention
In order to solve the problems in the background technology, the invention provides a track point cloud denoising method based on normal vector two-norm characteristic parameters. Firstly, performing grid processing on track point cloud data, solving the density of the point cloud grid, performing density threshold value and connectivity analysis, removing discrete noise deviating from a point cloud main body, and then dividing the track point cloud data by applying a method based on classification of two-norm characteristic parameters of a normal vector of the point cloud, and dividing the track point cloud into a characteristic region and a non-characteristic region. And finally, in order to keep the integrity of the deformation sensitive area information as much as possible, smoothing the track point cloud of the characteristic area by using an improved bilateral filtering algorithm, and denoising the non-characteristic area based on least square plane fitting.
In view of the above, the invention provides an orbit point cloud denoising method based on normal vector two-norm characteristic parameters.
The technical solution for realizing the purpose of the invention is as follows: an orbit point cloud denoising method based on normal vector two-norm characteristic parameters. The method comprises the following steps:
s1: performing grid processing on the whole track point cloud data;
s2: calculating the grid point cloud data density, carrying out grid density threshold value and connectivity analysis, and removing point cloud discrete noise;
s3: calculating the relationship between the normal vector of the point cloud and the two norms of the normal vector to construct a characteristic parameter of the two norms of the normal vector of the point cloud;
s4: dividing the track point cloud data into a characteristic area and a non-characteristic area by utilizing the parameter;
s5: the method comprises the steps of using a normal vector included angle cosine as a feature retention factor of a bilateral filtering factor to improve a bilateral filtering algorithm;
s6: carrying out improved bilateral filtering smooth denoising on the track characteristic region data;
s7: performing least square-based plane fitting denoising on the non-characteristic region data of the track;
s8: and (5) ending.
Preferably, in step 2, the set threshold d is: d= (d) x +d y +d z )/3
Wherein d x Is divided intoA dot; d, d y Is the median value; d, d z Is the inferior division point.
Preferably, in step 3, a point cloud normal vector two-norm characteristic parameter is utilized, and the parameter isWherein (1)>A normal vector representing point p; />Representing the in-field point q i Normal vector of (2); r is the set k neighborhood radius; n represents the number of all point clouds in the neighborhood; />Normal vector and q representing point p i The normal of the point measures a binary norm. />The value range is [0,1 ]]When the normal vector difference between the target point and the point in its domain is small +.>When the normal difference between the target point and the point in the field is large
Preferably, in step 4, all the track point cloud data points are traversed, normal vectors of the track point cloud data points are obtained, correction of the normal vector direction is performed, and the normal vector two-norm characteristic parameters in step 3 are used for dividing the whole track point cloud data into characteristic areas and non-characteristic areas.
Preferably, in step 5, the projection distance of the adjacent point on the tangent plane is used as a weighting factor for the traditional bilateral filtering factor, which has a certain limitation in a sharp area with more severe feature change, and the invention uses the cosine of the normal vector included angle as the feature retention factor of the bilateral filtering factor, and when the included angle of two vectors is 90 degrees, the weight is minimum; when the normal vector of each point in the k neighborhood is relatively consistent with the normal vector direction of the point to be searched or the included angle between the two vectors is smaller, the weight is larger. Therefore, the robustness of the track point cloud data of the sharp region with the sharp characteristic change can be better guaranteed.
Compared with the prior art, the invention has the remarkable advantages that: the invention provides a track point cloud denoising method based on normal vector two-norm characteristic parameters, which constructs characteristic parameters according to the normal vector two-norm of the point cloud, then sets a threshold value to divide track point cloud data, and uses an improved bilateral filtering algorithm to smooth and denoise a characteristic region.
Drawings
The drawings in the following description are only examples of the present invention and other drawings may be obtained from the drawings provided without inventive effort for a person of ordinary skill in the art.
FIG. 1 is a flow chart of an orbit point cloud denoising method based on normal vector two-norm characteristic parameters;
fig. 2 is a schematic diagram of original point cloud data according to the present invention, wherein (a) is a schematic diagram of original data of a railway track, and (b) is a schematic diagram of original data of a subway track;
FIG. 3 is a diagram of the data after removing the discrete noise according to the present invention, wherein (a) is a diagram of the data after removing the discrete noise for a railway track, and (b) is a diagram of the data after removing the discrete noise for a subway track;
fig. 4 is a schematic view of a feature area after performing point cloud division by using normal vector two norms, where (a) is a schematic view of a feature area of a railway track, and (b) is a schematic view of a feature area of a subway track;
fig. 5 is a schematic diagram of a non-characteristic area after performing point cloud division by using normal vector two norms, wherein (a) is a schematic diagram of a non-characteristic area of a railway track, and (b) is a schematic diagram of a non-characteristic area of a subway track;
FIG. 6 is a graph comparing denoising effects of a railway track and a subway track point cloud feature region, wherein (a) is a first denoising result graph of a railway track feature region algorithm, (b) is a second denoising result graph of a railway track feature region algorithm, (c) is a text algorithm denoising result graph of a railway track feature region, (d) is a first denoising result graph of a subway track feature region algorithm, (e) is a second denoising result graph of a subway track feature region algorithm, and (f) is a text denoising result graph of a subway track feature region;
FIG. 7 is a model comparison diagram of denoising of a point cloud feature region of a subway track according to the present invention, wherein (a) is a subway track original data model diagram, (b) is a point cloud feature region algorithm one denoising result model diagram, (c) is a point cloud feature region algorithm two denoising result model diagram, and (d) is a point cloud feature region text algorithm denoising result model diagram
FIG. 8 is a table of quantification of the comparison of different methods when denoising the point cloud characteristic areas of the railway track and the subway track according to the invention;
fig. 9 is a denoising result diagram of a non-characteristic area point cloud of a railway track, wherein (a) is a denoising result diagram of the non-characteristic area point cloud of the railway track, and (b) is a denoising result diagram of the non-characteristic area point cloud of the railway track;
fig. 10 is a denoising final result diagram according to the present invention, wherein (a) is a railway track denoising final result diagram and (b) is a subway track denoising final result diagram.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, the embodiment of the invention discloses a method for denoising an orbit point cloud based on a normal vector two-norm characteristic parameter, which comprises the following steps:
s1: performing grid processing on the whole point cloud data;
s2: calculating the grid point cloud data density, carrying out grid density threshold value and connectivity analysis, and removing point cloud discrete noise;
s3: calculating the relationship between the normal vector of the point cloud and the two norms of the normal vector to construct a characteristic parameter of the two norms of the normal vector of the point cloud;
s4: dividing the point cloud data into a characteristic area and a non-characteristic area by utilizing the parameter;
s5: the method comprises the steps of using a normal vector included angle cosine as a feature retention factor of a bilateral filtering factor to improve a bilateral filtering algorithm;
s6: carrying out improved bilateral filtering smooth denoising on the track characteristic region data;
s7: performing least square-based plane fitting denoising on the non-characteristic region data of the track;
s8: and (5) ending.
In order to further optimize the above technical solution, the threshold d set in step 2 is:
d＝(d x +d y +d z )/3
wherein d x Is divided into upper points; d, d y Is the median value; d, d z Is the inferior division point.
In order to further optimize the technical scheme, in step 3, a point cloud normal vector two-norm characteristic parameter is utilized, wherein the parameter isWherein (1)>A normal vector representing point p; />Representing the in-field point q i Normal vector of (2); r is the set k neighborhood radius; n represents the number of all point clouds in the neighborhood; />Normal vector and q representing point p i The normal of the point measures a binary norm. />The value range is [0,1 ]]When the normal vector difference between the target point and the point in its domain is small +.>When the normal difference between the target point and the point in the fieldWhen the difference is large
In order to further optimize the technical scheme, in the step 4, all the orbit point cloud data points are traversed by using a k-d tree, normal vectors of the orbit point cloud data points are obtained, correction of the normal vector direction is carried out, and the whole orbit point cloud data are divided into a characteristic area and a non-characteristic area by using the normal vector two-norm characteristic parameters in the step 3.
In order to further optimize the technical scheme, in step 5, the projection distance of the adjacent points on the tangent plane of the conventional bilateral filter factor is used as a weighting factor, which has a certain limitation in a sharp area with more severe characteristic change, and the invention uses the normal vector included angle cosine as the characteristic retention factor of the bilateral filter factor, and when the included angle of two vectors is 90 degrees, the weight is minimum; when the normal vector of each point in the k neighborhood is relatively consistent with the normal vector direction of the point to be searched or the included angle between the two vectors is smaller, the weight is larger. Therefore, the robustness of the track point cloud data of the sharp region with the sharp characteristic change can be better guaranteed.
The method according to the invention is further described below.
Examples
The track point cloud denoising method based on the normal vector two-norm characteristic parameters is selected and constructed, and because track point cloud data contains multi-scale noise, the method is difficult to apply to deformation monitoring and model reconstruction of tracks, and better characteristic integrity is required to be maintained for denoising characteristic region data.
Step one, performing grid processing on the whole point cloud data;
step two, calculating the grid point cloud data density, carrying out grid density threshold value and connectivity analysis, and removing point cloud discrete noise;
and counting the number of point clouds in a single grid of the data subjected to the previous step of grid mesh, calculating the grid density, calculating a grid density threshold d according to a formula (1), judging the data as effective data points if the single grid density is more than 0.25d, and otherwise, judging the data as discrete noise points.
d＝(d x +d y +d z )/3 (1)
Wherein d x Is divided into upper points; d, d y Is the median value; d, d z Is the inferior division point.
After the removal of the discrete noise points is completed, for any grid g, 6 grids coplanar with it are counted. Assuming that the coordinates of g are (x, y, z), counting whether the grids at 6 coordinates of (x-1, y, z), (x+1, y, z), (x, y+1, z), (x, y, z-1) and (x, y, z+1) exist point cloud data, if the grid number N of the data exists i If it is 2 or more, the mesh g is considered to be an effective mesh. If N i Equal to 1, i.e. g has only 1 closely adjacent grids g 1 If data exists, then pair g 1 Repeating the steps to obtain N i+1 At this time, the original grid g is not counted, if N i+1 Greater than or equal to 1 g 1 For effective grid, otherwise g 1 I.e. the grid where the discrete noise clusters are located. If N i If 0, g can be directly obtained 1 Is the grid where the discrete noise clusters are located, and the data of the grid where the discrete noise clusters are located is removed. The above steps are repeated until all grids are traversed.
Thirdly, calculating the relationship between the normal vector of the point cloud and the two norms of the normal vector to construct a characteristic parameter of the two norms of the normal vector of the point cloud;
(1) Assuming that the track point cloud has m data points in total, performing zero-mean processing on the track point cloud, and calculating a covariance matrix C of the track point cloud:
(2) And (3) obtaining the eigenvalue and eigenvector of the matrix C, wherein the eigenvector corresponding to the minimum eigenvalue is the normal vector of the point. Normally, the normal vector direction of the unordered point cloud in the neighborhood is discontinuous, in order to unify the normal vector direction, a topological structure is constructed by using a k-d tree, all data points are traversed, any one point is selected, the normal vector is calculated, when the normal vector of the point is negative with the normal vector inner products of other points in the k neighborhood, namely, the two vector directions are opposite, the direction correction needs to be carried out, otherwise, the direction correction is kept unchanged, as shown in a formula (5):
in the method, in the process of the invention,is the normal vector at the center feature point; />And the central feature point corresponds to the normal vector of the k neighborhood points.
(3) After the normal vector of the track point cloud with the corrected direction is obtained, the track point cloud data is divided in a normal vector two-norm mode. In general, in the region with rich features, the plane variation amplitude is larger, the distance between normal vectors of corresponding data points is larger, and in the corresponding non-feature flat region, the distance between normal vectors is smaller. As shown in formula (6):
in the method, in the process of the invention,a normal vector representing point p; />Representing the in-field point q i Normal vector of (2); r is the set k neighborhood radius; n represents the number of all point clouds in the neighborhood; />Normal vector and q representing point p i Normal vector of points two norms。/>The value range is [0,1 ]]When the normal vector difference between the target point and the point in the field is smallWhen the normal difference between the target point and the point in the field is large +.>
Dividing the point cloud data into a characteristic area and a non-characteristic area by utilizing the parameter;
fifthly, using a normal vector included angle cosine as a feature retention factor of the bilateral filtering factor to improve the bilateral filtering algorithm;
conventional bilateral filtering factors use the projection distance of neighboring points on their tangential planes as a weighting factor, which has a limitation in sharp areas where the features change more severely. The improved bilateral filtering is adopted for filtering the characteristic region, and the improved bilateral filtering factor is shown as a formula (7):
in the method, in the process of the invention,the Euclidean distance from point p to point q; />Distance vector and normal vector for points p and q>Is an inner product of (2); x is the neighborhood set of data points p; omega s And omega r Respectively regarding the degree of fairing sigma d And the degree of feature retention sigma n Is a gaussian kernel function of (c).
In the neighborhood range of the data point p, the normal change of the point can reflect the characteristic change, the part with large normal change has large curvature change, the characteristic change is obvious, and otherwise, the characteristic change is smaller. The cosine of the normal vector included angle is used as a feature retention factor of the bilateral filtering factor, and when the included angle of the two vectors is 90 degrees, the weight is minimum; when the normal vector of each point in the k neighborhood is relatively consistent with the normal vector direction of the point to be searched or the included angle between the two vectors is smaller, the weight is larger. This can better ensure the robustness of the data in sharp areas where the features are highly variable.
Step six, carrying out improved bilateral filtering smooth denoising on the characteristic region data;
the near signal point noise is removed from the characteristic region track point cloud data by using the filtering algorithm provided in the last step, and the track point cloud data is processed by using an original bilateral filtering algorithm, a literature improvement algorithm and a text improvement algorithm (hereinafter, the track point cloud data are collectively referred to as algorithm 1, algorithm 2 and text Wen Suanfa). In order to further measure the denoising performance of the algorithm, the method introduces Hausdorff distance from point to point (P2 point) and from point to plane (P2 plane) to reflect the quality of the denoising effect, and in the aspect of feature retention, the method introduces information entropy as a judgment standard, and the information entropy of the point depends on the disorder degree of the area and can be used for describing the feature richness of the point. The denoising time determines the operation efficiency of the algorithm, so that four evaluation indexes of P2point, P2plane, average information entropy and denoising time are used for more objectively analyzing the denoising result.
Step seven, carrying out least square-based plane fitting denoising on the non-characteristic region data;
and carrying out least square plane fitting denoising on the non-characteristic region, counting the distance between the data point and the fitting plane, and setting a threshold value at the 95% confidence degree to remove near-signal point noise in order to maximize denoising efficiency.
Claims (5)
1. The orbit point cloud denoising method based on the normal vector two-norm characteristic parameter is characterized by dividing orbit point cloud data by solving the normal vector of the orbit point cloud data and the normal vector two-norm characteristic parameter and denoising different types of data by using different methods, and comprises the following steps of:
s1: performing grid processing on the whole track point cloud data;
s2: calculating grid point cloud data density, carrying out grid density threshold value and connectivity analysis, and removing track point cloud discrete noise;
s3: calculating the relationship between the normal vector of the point cloud and the two norms of the normal vector to construct a characteristic parameter of the two norms of the normal vector of the point cloud;
s4: dividing the track point cloud data into a characteristic area and a non-characteristic area by utilizing the parameter;
s5: the method comprises the steps of using a normal vector included angle cosine as a feature retention factor of a bilateral filtering factor to improve a bilateral filtering algorithm;
s6: carrying out improved bilateral filtering smooth denoising on the track characteristic region data;
s7: performing least square-based plane fitting denoising on the non-characteristic region data of the track;
s8: and (5) ending.
2. The orbit point cloud denoising method based on normal vector two-norm characteristic parameter according to claim 1, wherein the method comprises the following steps:
in step 2, the set threshold d is:
d＝(d x +d y +d z )/3
wherein d x Is divided into upper points; d, d y Is the median value; d, d z Is the inferior division point.
3. The orbit point cloud denoising method based on normal vector two-norm characteristic parameter according to claim 1, wherein the method comprises the following steps:
in the step 3, the point cloud data of the track is divided by utilizing the point cloud normal vector two-norm characteristic parameters, the point cloud data comprising the track surface, the track head and the track connection part are divided into characteristic areas, and the area comprising the track surface is divided into non-characteristic areas. The parameter is
Wherein,a normal vector representing point p; />Representing the in-field point q i Normal vector of (2); r is the set k neighborhood radius; n represents the number of all point clouds in the neighborhood; />Normal vector and q representing point p i The normal of the point measures a binary norm. Parameter +.>Parameter +_when the normal difference between the target point and its field point is large>
4. The orbit point cloud denoising method based on normal vector two-norm characteristic parameter according to claim 1, wherein the method comprises the following steps:
in step 4, traversing all the orbit point cloud data points by using a k-d tree, solving the normal vector of the orbit point cloud data points, correcting the normal vector direction, and dividing the whole orbit point cloud data into a characteristic area and a non-characteristic area by using the normal vector two-norm characteristic parameters in step 3.
5. The orbit point cloud denoising method based on normal vector two-norm characteristic parameter according to claim 1, wherein the method comprises the following steps:
in step 5, the projection distance of the adjacent points on the tangential plane of the conventional bilateral filtering factor is used as a weighting factor, which has a certain limitation in a sharp area with more severe feature change, and the invention uses the normal vector included angle cosine as the feature retention factor of the bilateral filtering factor, and when the included angle of two vectors is 90 degrees, the weight is minimum; when the normal vector of each point in the k neighborhood is relatively consistent with the normal vector direction of the point to be searched or the included angle between the two vectors is smaller, the weight is larger. Therefore, the robustness of the track point cloud data of the sharp region with the sharp characteristic change can be better guaranteed. The improved bilateral filter factor is as follows
In the method, in the process of the invention,the Euclidean distance from point p to point q; />Distance vector and normal vector for points p and q>Is an inner product of (2); x is the neighborhood set of data points p; omega s And omega r Respectively regarding the degree of fairing sigma d And the degree of feature retention sigma n Is a gaussian kernel function of (c).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310454952.XA CN117132478B (en) | 2023-04-25 | Orbit point cloud denoising method based on normal vector two-norm characteristic parameter |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310454952.XA CN117132478B (en) | 2023-04-25 | Orbit point cloud denoising method based on normal vector two-norm characteristic parameter |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117132478A true CN117132478A (en) | 2023-11-28 |
| CN117132478B CN117132478B (en) | 2024-05-17 |
Family
ID=
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109241978A (en) * | 2018-08-23 | 2019-01-18 | 中科光绘（上海）科技有限公司 | The rapid extracting method of planar chip in ground three-dimensional laser point cloud |
| CN111986115A (en) * | 2020-08-22 | 2020-11-24 | 王程 | Accurate elimination method for laser point cloud noise and redundant data |
| CN112132840A (en) * | 2020-09-01 | 2020-12-25 | 济南市房产测绘研究院(济南市房屋安全检测鉴定中心) | Vehicle-mounted street tree point cloud classification and characteristic information extraction method |
| WO2021129317A1 (en) * | 2019-12-26 | 2021-07-01 | 华南理工大学 | Point cloud smoothing filtering method based on normal vector |
| CN113177897A (en) * | 2021-05-20 | 2021-07-27 | 聚时科技(上海)有限公司 | Rapid lossless filtering method for disordered 3D point cloud |
| CN114581331A (en) * | 2022-03-14 | 2022-06-03 | 深圳市凌云视迅科技有限责任公司 | Point cloud noise reduction method and device suitable for multiple scenes |
| CN115131424A (en) * | 2022-06-13 | 2022-09-30 | 武汉科技大学 | Clustering and boundary feature-preserving automobile trunk point cloud volume calculation method |
| 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 |
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109241978A (en) * | 2018-08-23 | 2019-01-18 | 中科光绘（上海）科技有限公司 | The rapid extracting method of planar chip in ground three-dimensional laser point cloud |
| WO2021129317A1 (en) * | 2019-12-26 | 2021-07-01 | 华南理工大学 | Point cloud smoothing filtering method based on normal vector |
| CN111986115A (en) * | 2020-08-22 | 2020-11-24 | 王程 | Accurate elimination method for laser point cloud noise and redundant data |
| CN112132840A (en) * | 2020-09-01 | 2020-12-25 | 济南市房产测绘研究院(济南市房屋安全检测鉴定中心) | Vehicle-mounted street tree point cloud classification and characteristic information extraction method |
| CN113177897A (en) * | 2021-05-20 | 2021-07-27 | 聚时科技(上海)有限公司 | Rapid lossless filtering method for disordered 3D point cloud |
| CN114581331A (en) * | 2022-03-14 | 2022-06-03 | 深圳市凌云视迅科技有限责任公司 | Point cloud noise reduction method and device suitable for multiple scenes |
| CN115131424A (en) * | 2022-06-13 | 2022-09-30 | 武汉科技大学 | Clustering and boundary feature-preserving automobile trunk point cloud volume calculation method |
| 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 |
Non-Patent Citations (4)
| Title |
|---|
| LANG ZHOU等: "Point cloud denoising review: from classical to deep learning-based approaches", 《GRAPHICAL MODELS》, 2 May 2022 (2022-05-02), pages 1 - 16 * |
| 戴雷宇: "基于地基与机载LiDAR数据的阔叶林LAD反演", 《中国优秀硕士学位论文全文数据库 农业科技辑》, no. 2020, 15 January 2020 (2020-01-15), pages 049 - 121 * |
| 王井利等: "基于地铁隧道点云数据的组合滤波算法", 《沈阳建筑大学学报(自然科学版)》, vol. 36, no. 3, 31 May 2020 (2020-05-31), pages 520 - 528 * |
| 郭进等: "采用密度k-means和改进双边滤波的点云自适应去噪算法", 《传感器与微系统( TRANSDUCER AND MICROSYSTEM TECHNOLOGIES)》, vol. 35, no. 7, 31 December 2016 (2016-12-31), pages 147 - 153 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111299815B (en) | Visual detection and laser cutting trajectory planning method for low-gray rubber pad | |
| CN103870845B (en) | Novel K value optimization method in point cloud clustering denoising process | |
| CN111986115A (en) | Accurate elimination method for laser point cloud noise and redundant data | |
| CN110807781B (en) | Point cloud simplifying method for retaining details and boundary characteristics | |
| CN109272524B (en) | Small-scale point cloud noise denoising method based on threshold segmentation | |
| CN110580705B (en) | Method for detecting building edge points based on double-domain image signal filtering | |
| CN111598780B (en) | Terrain adaptive interpolation filtering method suitable for airborne LiDAR point cloud | |
| CN109886937B (en) | Insulator defect detection method based on super-pixel segmentation image recognition | |
| CN113628263A (en) | Point cloud registration method based on local curvature and neighbor characteristics thereof | |
| CN105118090A (en) | Adaptive point-cloud filtering method for complex terrain structure | |
| CN111340723B (en) | Terrain-adaptive airborne LiDAR point cloud regularization thin plate spline interpolation filtering method | |
| CN109100788B (en) | Seismic data non-local mean de-noising method | |
| CN115222625A (en) | Laser radar point cloud denoising method based on multi-scale noise | |
| CN111968224A (en) | Ship 3D scanning point cloud data processing method | |
| CN111062960A (en) | Aircraft skin butt joint characteristic line extraction method based on scattered point cloud | |
| CN114663373A (en) | Point cloud registration method and device for detecting surface quality of part | |
| CN114898118A (en) | Automatic statistical method and system for power transmission line house removal amount based on multi-source point cloud | |
| CN116433707A (en) | Accurate extraction method and system for optical center sub-pixels of line structure under complex background | |
| CN116538996A (en) | Laser radar-based topographic mapping system and method | |
| CN111639560A (en) | Finger vein feature extraction method and device based on dynamic fusion of vein skeleton line and topographic relief characteristic | |
| CN108734731B (en) | Machine vision-based quasi-circular particle target counting method | |
| CN108520550B (en) | Point cloud data Poisson surface reconstruction method based on noise classification and MLS | |
| CN117132478B (en) | Orbit point cloud denoising method based on normal vector two-norm characteristic parameter | |
| CN110458938B (en) | Real-time three-dimensional reconstruction method and system for bulk material pile | |
| CN102314687B (en) | Method for detecting small targets in infrared sequence images |
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 | ||
| GR01 | Patent grant |