CN109389672B - Processing method of point cloud of reflective workpiece - Google Patents
Processing method of point cloud of reflective workpiece Download PDFInfo
- Publication number
- CN109389672B CN109389672B CN201811151979.7A CN201811151979A CN109389672B CN 109389672 B CN109389672 B CN 109389672B CN 201811151979 A CN201811151979 A CN 201811151979A CN 109389672 B CN109389672 B CN 109389672B
- Authority
- CN
- China
- Prior art keywords
- point
- point cloud
- points
- distance
- subsets
- 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.)
- Active
Links
- 238000003672 processing method Methods 0.000 title claims abstract description 6
- 238000012545 processing Methods 0.000 claims abstract description 22
- 238000000034 method Methods 0.000 claims abstract description 17
- 238000012952 Resampling Methods 0.000 claims abstract description 9
- 238000009827 uniform distribution Methods 0.000 claims abstract description 4
- 238000009826 distribution Methods 0.000 claims description 10
- 238000001914 filtration Methods 0.000 claims description 4
- 230000002146 bilateral effect Effects 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 230000002950 deficient Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 6
- 230000007547 defect Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20028—Bilateral filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30164—Workpiece; Machine component
Abstract
Hair brushThe invention provides a method for processing a point cloud of a reflective workpiece. A processing method of a reflective workpiece point cloud comprises the following steps: s1, inputting point clouds
Carrying out nearest neighbor resampling to obtain m subsets; s2, carrying out point cloud denoising, outlier removing and uniform distribution on each subset by using a Gaussian weighted local optimal projection operator (SWLOP); s3, processing each subset (x) processed by SWLOP operator 1 ,x 2 ,x 3 L x n ) Iterating for t times until the processed point cloud meets the requirement of subsequent processing, and then recombining the subsets into a new subset; s4. Multiple subsets (x) 1 ,x 2 ,x 3 L x n ) Are combined into a point set
And performing edge-aware resampling (EAR) upsampling on the merged point cloud. The invention can improve the quality of the defective point cloud, shorten the time for processing the point cloud and improve the efficiency and the precision of the three-dimensional reconstruction of the point cloud.
Description
Technical Field
The invention relates to the technical field of three-dimensional scanning, in particular to a method for processing a point cloud of a reflective workpiece.
Background
Three-dimensional point clouds are becoming more and more popular in many research fields such as object recognition and reconstruction due to their simplicity and high flexibility of representation. The rapid development of three-dimensional scanners such as laser three-dimensional scanners, structured light three-dimensional scanners, laser radars and the like enables us to rapidly and accurately acquire complete point cloud data. However, the point cloud obtained is inevitably contaminated by noise due to the defects of the acquisition devices themselves. Especially when the surface of the scanned object has a reflection characteristic, the acquired point cloud contains a large amount of noise and outliers, and the acquired point cloud data is not uniformly distributed. The stronger the reflection characteristic of the object surface, the more uneven the point cloud distribution, the more noise, and the point cloud loss exists at a plurality of positions, which seriously affect the precision of the subsequent three-dimensional reconstruction. In addition, processing a large amount of point cloud data takes much time, reducing the efficiency of point cloud processing.
The current processing method of point cloud data is mainly to directly carry out filtering and denoising operation on input point cloud and then carry out mesh reconstruction. Although satisfactory results can be obtained when the method is used for processing point cloud models with low noise, the defects of the method are obvious when a large amount of noise exists in the point cloud models. For example, the point cloud data processing efficiency is not high, and the noise points cannot be effectively removed.
Disclosure of Invention
The invention provides a method for processing a point cloud of a reflective workpiece, which aims to overcome at least one defect in the prior art. The invention can improve the quality of the defective point cloud, shorten the time for processing the point cloud and improve the efficiency and the precision of the three-dimensional reconstruction of the point cloud.
In order to solve the technical problems, the invention adopts the technical scheme that: a processing method of a reflective workpiece point cloud comprises the following steps:
s1, inputting point clouds
Carrying out nearest neighbor resampling to obtain m subsets;
s2, carrying out point cloud denoising, outlier removing and uniform distribution on each subset by using a Gaussian weighted local optimal projection operator (SWLOP);
s3, processing each subset (x) processed by SWLOP operator 1 ,x 2 ,x 3 L x n ) Iterating for t times until the processed point cloud meets the requirement of subsequent processing, and then recombining the subsets into a new subset;
s4. Multiple subsets (x) 1 ,x 2 ,x 3 L x n ) Are combined into a point set
And performing edge-aware resampling (EAR) upsampling on the merged point cloud.
Further, in step S1, a data set is given
Dividing the point set by a kd-tree, and inquiring each data point p by the kd-tree c R field of (a) (p) 0 ,p 1 ,p 2 L p k ) (ii) a Calculating the distance between each point in r field and query point p c The indexes of each point are arranged according to the distance, and the distance p is extracted c The index number of m points nearest to the point is respectively stored in (R) according to the index number 0 ,R 1 ,R 2 L R m ) Among the subsets, P is divided into m subsets by passing through all points.
Further, in step S2, the original point cloud is first processed
Expressed as P-N (mu, sigma) 2 ) Is calculated as a point p according to the formula (1) j R field of (1) average distance mu of all points j Sum variance
Then with point p j Point cloud obedience within r-domain as center
Gaussian distribution, and solving the average distance mu and the variance sigma of the global point cloud according to the formula (2) 2 ；
In formula (1) and formula (2):
p j -the jth point; p is a radical of jn -nth point in r field of jth point;
p Jnearest -is a distance p j The closest point; n is the total number of points in the r field of the j point;
σ j 2 -point p j r variance within the domain; mu.s j -point p j r mean in domain;
σ 2 -a global variance; μ — global mean;
comparison of
And σ 2 If, if
Then point p is identified j To distribute outliers and noise, set p j Gaussian weight of (T) j Equal to 0, otherwise T j Is equal to 1,T j The value is expressed as the following binary equation:
taking the value of the Gaussian weight value of each point of the original point cloud P, if the point P is j Variance of all points in r field of
Less than global variance σ 2 Then p is j The value of the Gaussian weight function is 1, otherwise, the value is 0;
projecting arbitrary set of points onto P-surface
After-projection point set is recorded as Q = { Q = i } y∈I The distribution of the set Q satisfies: 1. as close as possible to the surface represented by the point cloud P; 2. the points of Q are distributed as uniformly as possible; the fixed point of the formula (4) is solved through iteration;
Q＝G(C) (4)
in equations (4) and (5):
E 1 (Y, P, C) -for controlling the projected points to be as close as possible to the real surface represented by the point cloud P;
E 2 (Y,C)-for controlling the distribution between proxels as uniform as possible;
T j is a point p j A Gaussian weight function;
c-set of points Q = { Q ] for final iteration i } y∈I (ii) a P-is the original set of points
Y-is any set of points projected onto the P surface
P represents the distance between points;
η(r)＝1/(3r 3 ) -simulating the repulsive force between the projected points, r being the distance between two points;
Further, the step S4 specifically includes the following steps:
s41, calculating the local density of the point cloud, positioning the most sparse position of the current point set, and inserting a point;
s42, moving the base point to the implicit curved surface of the current local area by using a bilateral filtering projection method;
and S43, judging whether the inserted base point meets the condition of upsampling, and if so, returning to the step S41.
Compared with the prior art, the invention has the following beneficial effects:
according to the method, aiming at the scanning point cloud of the workpiece with the surface reflecting light, firstly, the original point cloud is segmented; then, in order to achieve the effect of improving the quality of the point clouds, carrying out noise reduction operation on each piece of point cloud respectively by using Gaussian weighted local projection algorithm; and finally, carrying out edge perception resampling operation on the point cloud in order to enhance and maintain the point cloud details. The invention can effectively process the scanning point cloud of the workpiece with the reflective surface, and the efficiency is greatly improved.
Drawings
FIG. 1 is a flow chart of a method for processing a cloud of retroreflective workpiece points according to the present invention.
FIG. 2 is a schematic diagram of the uniform sampling based on kd-tree in the present invention.
Fig. 3 is a schematic diagram of an example wafer point cloud processing in an embodiment of the invention.
Fig. 4 is a schematic diagram of a wafer point cloud session reconstruction result in the embodiment of the present invention.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the patent; for the purpose of better illustrating the embodiments, certain features of the drawings may be omitted, enlarged or reduced, and do not represent the size of an actual product; it will be understood by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted. The positional relationships depicted in the drawings are for illustrative purposes only and are not to be construed as limiting the present patent.
As shown in fig. 1 to 3, a method for processing a reflective workpiece point cloud includes the following steps:
s1, inputting point clouds
Nearest neighbor resampling into m subsets is performed. The method comprises the following specific steps: given a data set
Dividing the point set by a kd-tree, and inquiring each data point p by the kd-tree c R field of (a) (p) 0 ,p 1 ,p 2 L p k ) (ii) a Calculating the distance between each point in r field and query point p c The indexes of each point are arranged according to the distance, and the distance p is extracted c Index number of 4 points nearest to the point according to the index numberStore 4 points in (R) respectively 0 ,R 1 ,R 2 ,R 3 ) Among the subsets, P is divided into 4 subsets by passing all points.
And S2, carrying out point cloud denoising, outlier removing and uniform distribution on each subset by using a Gaussian weighted local optimal projection operator (SWLOP). The method comprises the following specific steps:
firstly, the original point cloud is obtained
Expressed as P-N (mu, sigma) 2 ) Is calculated as a point p according to the formula (1) j R field of (1) average distance mu of all points j Sum variance
Then with point p j Point cloud obedience within r-domain as center
Gaussian distribution, which is to calculate the average distance mu and variance sigma of the global point cloud according to the formula (2) 2 。
In formula (1) and formula (2):
p j -the jth point; p is a radical of jn -nth point in r field of jth point;
p Jnearest -is a distance p j The closest point; n is the total number of points in the r field of the j point;
σ j 2 -point p j r variance within the domain; mu.s j -point p j r mean in domain;
σ 2 -a global variance; μ — global mean;
comparison of
And σ 2 If, if
Then point p is identified j To distribute outliers and noise, set p j Gaussian weight of (T) j Equal to 0, otherwise T j Is equal to 1,T j The value is expressed as the following binary equation:
taking the value of the Gaussian weight value of each point of the original point cloud P, if the point P is j Variance of all points in r field of
Less than global variance σ 2 Then p is j The gaussian weight function takes a value of 1, otherwise it takes a value of 0.
Projecting arbitrary set of points onto P-surface
After-projection point set is recorded as Q = { Q = i } y∈I The distribution of the set Q satisfies: 1. as close as possible to the surface represented by the point cloud P; 2. the points of Q are distributed as uniformly as possible; the fixed point of the formula (4) is solved through iteration;
Q＝G(C) (4)
in equations (4) and (5):
E 1 (Y, P, C) -for controlling the projected points to be as close as possible to the real surface represented by the point cloud P;
E 2 (Y, C) -for controlling the distribution between the proxels as uniform as possible;
T j is a point p j A Gaussian weight function;
c-set of points Q = { Q ] for final iteration i } y∈I (ii) a P-is the original set of points
Y-is any set of points projected onto the P surface
P.P-represents the distance between points;
η(r)＝1/(3r 3 ) -simulating the repulsive force between the projected points, r being the distance between two points;
And S3, iterating each subset processed by the SWLOP operator for 5 times until the processed point cloud meets the requirement of subsequent processing, and then combining the processed subsets into a complete point cloud again.
And S4, performing edge-aware resampling (EAR) upsampling on the merged point cloud. The method comprises the following specific steps:
s41, calculating the local density of the point cloud, positioning the most sparse position of the current point set, and inserting a point;
s42, moving the base point to the implicit curved surface of the current local area by using a bilateral filtering projection method;
and S43, judging whether the inserted base point meets the condition of upsampling, and if so, returning to the step S41.
And finally, reconstructing the processed point cloud by using a Session reconstruction algorithm, wherein the reconstruction result is shown in FIG. 4. The enlarged part can be seen, the noise point of the wafer point cloud is effectively removed, the part with small point cloud loss is filled, and the fine structure and the edge wire in the middle part of the wafer are maintained.
TABLE 1 treatment time for each stage of the process
The algorithm processing time of each model is shown in table 1, and it can be seen from the table that the nearest neighbor sampling through the kd-tree is greatly shortened in algorithm processing time, wherein the processing time of the wafer point cloud is improved by 43.8%, and the efficiency is effectively improved.
It should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.
Claims (2)
1. A processing method of a reflective workpiece point cloud is characterized by comprising the following steps:
s1, inputting point clouds
Carrying out nearest neighbor resampling to obtain m subsets;
s2, carrying out point cloud denoising, outlier removing and uniform distribution on each subset by using a Gaussian weighted local optimal projection operator (SWLOP);
s3, processing each subset (x) processed by SWLOP operator 1 ,x 2 ,x 3 …x n ) Iterating for t times until the processed point cloud meets the requirement of subsequent processing, and recombining the subsetsA new subset;
s4. Multiple subsets (x) 1 ,x 2 ,x 3 …x n ) Are combined into a point set
Performing edge-aware resampling (EAR) upsampling on the merged point cloud;
in said step S1, a data set is given
Dividing the point set by a kd-tree, and inquiring each data point p by using the kd-tree c R field of (a) (p) 0 ,p 1 ,p 2 …p k ) (ii) a Calculating the distance between each point in r field and query point p c The indexes of each point are arranged according to the distance, and the distance p is extracted c The index numbers of m points with the nearest points are respectively stored in (R) 0 ,R 1 ,R 2 …R m ) Among the subsets, dividing P into m subsets by passing through all points;
in the step S2, the original point cloud is firstly processed
Expressed as P-N (mu, sigma) 2 ) Is calculated as a point p according to the formula (1) j R field of (1) average distance mu of all points j Sum variance
Then with point p j Point cloud obedience within r-domain as center
Gaussian distribution, and solving the average distance mu and the variance sigma of the global point cloud according to the formula (2) 2 ；
In formula (1) and formula (2):
p j -the jth point; p is a radical of jn -nth point in r field of jth point;
p Jnearest is a distance p j The closest point; n-the total number of points in the r field of the j point;
σ j 2 -point p j r variance within the domain; mu.s j -point p j r mean in domain;
σ 2 -a global variance; μ — global mean;
comparison of
And σ 2 If, if
Then point p is identified j To distribute outliers and noise, set p j Gaussian weight of (T) j Equal to 0, otherwise T j Is equal to 1,T j The value is expressed as the following binary equation:
taking the value of the Gaussian weight value of each point of the original point cloud P, if the point P is j Variance of all points in r field of
Less than global variance σ 2 Then p is j The value of the Gaussian weight function is 1, otherwise, the value is 0;
projecting arbitrary set of points onto P-surface
After-projection point set is recorded as Q = { Q = i } y∈I The distribution of the set Q satisfies: 1. as close as possible to the surface represented by the point cloud P; 2. the points of Q are distributed as uniformly as possible; solving the fixed point of formula (4) through iteration;
Q＝G(C) (4)
in equations (4) and (5):
E 1 (Y, P, C) -for controlling the projected points to be as close as possible to the real surface represented by the point cloud P;
E 2 (Y, C) -for controlling the distribution between the proxels as uniform as possible;
T j is a point p j A Gaussian weight function;
c-set of points Q = { Q ] for final iteration i } y∈I (ii) a P-is the original set of points
Y-is any set of points projected onto the P surface
Iih | -represents the distance between the points;
η(r)＝1/(3r 3 ) -simulating the repulsive force between the projected points, r being the distance between two points;
2. The method of claim 1, wherein the step S4 comprises the following steps:
s41, calculating the local density of the point cloud, positioning the most sparse position of the current point set, and inserting a point;
s42, moving the base point to the implicit curved surface of the current local area by using a bilateral filtering projection method;
and S43, judging whether the inserted base point meets the condition of upsampling, and if so, returning to the step S41.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811151979.7A CN109389672B (en) | 2018-09-29 | 2018-09-29 | Processing method of point cloud of reflective workpiece |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811151979.7A CN109389672B (en) | 2018-09-29 | 2018-09-29 | Processing method of point cloud of reflective workpiece |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109389672A CN109389672A (en) | 2019-02-26 |
| CN109389672B true CN109389672B (en) | 2023-03-21 |
Family
ID=65418328
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811151979.7A Active CN109389672B (en) | 2018-09-29 | 2018-09-29 | Processing method of point cloud of reflective workpiece |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109389672B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114627020B (en) * | 2022-03-18 | 2023-06-20 | 易思维（杭州）科技有限公司 | Method for removing reflection noise point of curved surface workpiece |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103729872A (en) * | 2013-12-30 | 2014-04-16 | 浙江大学 | Point cloud enhancement method based on subsection resampling and surface triangularization |
| CN106504332A (en) * | 2016-10-19 | 2017-03-15 | 未来科技（襄阳）有限公司 | The curve reestablishing method and device of three-dimensional point cloud |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107223268B (en) * | 2015-12-30 | 2020-08-07 | 中国科学院深圳先进技术研究院 | Three-dimensional point cloud model reconstruction method and device |
-
2018
- 2018-09-29 CN CN201811151979.7A patent/CN109389672B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103729872A (en) * | 2013-12-30 | 2014-04-16 | 浙江大学 | Point cloud enhancement method based on subsection resampling and surface triangularization |
| CN106504332A (en) * | 2016-10-19 | 2017-03-15 | 未来科技（襄阳）有限公司 | The curve reestablishing method and device of three-dimensional point cloud |
Non-Patent Citations (1)
| Title |
|---|
| 特征敏感的点云重采样算法;陈永辉等;《小型微型计算机系统》;20170515(第05期);正文第1-5页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109389672A (en) | 2019-02-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | Hf-neus: Improved surface reconstruction using high-frequency details | |
| CN110390638B (en) | High-resolution three-dimensional voxel model reconstruction method | |
| CN110415166B (en) | Training method for fusion image processing model, image processing method, image processing device and storage medium | |
| Ming et al. | Rapid reconstruction of 3D neuronal morphology from light microscopy images with augmented rayburst sampling | |
| CN111931914A (en) | Convolutional neural network channel pruning method based on model fine tuning | |
| US8351654B2 (en) | Image processing using geodesic forests | |
| WO2013106984A1 (en) | Learning painting styles for painterly rendering | |
| US20220180036A1 (en) | Methods for modeling of a design in reticle enhancement technology | |
| CN113592711B (en) | Three-dimensional reconstruction method, system, equipment and storage medium for non-uniform point cloud data | |
| Djemame et al. | Solving reverse emergence with quantum PSO application to image processing | |
| CN105718943A (en) | Character selection method based on particle swarm optimization algorithm | |
| CN114663685B (en) | Pedestrian re-recognition model training method, device and equipment | |
| CN106504332A (en) | The curve reestablishing method and device of three-dimensional point cloud | |
| CN115222625A (en) | Laser radar point cloud denoising method based on multi-scale noise | |
| Chen et al. | Persistent memory residual network for single image super resolution | |
| CN111027590A (en) | Breast cancer data classification method combining deep network features and machine learning model | |
| CN109389672B (en) | Processing method of point cloud of reflective workpiece | |
| CN112419191A (en) | Image motion blur removing method based on convolution neural network | |
| CN108520550B (en) | Point cloud data Poisson surface reconstruction method based on noise classification and MLS | |
| CN112967296B (en) | Point cloud dynamic region graph convolution method, classification method and segmentation method | |
| CN107862749B (en) | Point cloud geometric detail feature removing method | |
| CN113963138A (en) | Complete and accurate extraction method of three-dimensional laser point cloud characteristic point line | |
| CN111488811A (en) | Face recognition method and device, terminal equipment and computer readable medium | |
| CN109409381A (en) | The classification method and system of furniture top view based on artificial intelligence | |
| Cudic et al. | Unpaired mesh-to-image translation for 3D fluorescent microscopy images of neurons |
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 | ||
| GR01 | Patent grant |