CN104732581A - Mobile context point cloud simplification algorithm based on point feature histogram - Google Patents
Mobile context point cloud simplification algorithm based on point feature histogram Download PDFInfo
- Publication number
- CN104732581A CN104732581A CN201410851762.2A CN201410851762A CN104732581A CN 104732581 A CN104732581 A CN 104732581A CN 201410851762 A CN201410851762 A CN 201410851762A CN 104732581 A CN104732581 A CN 104732581A
- Authority
- CN
- China
- Prior art keywords
- point
- standard deviation
- histogram
- patterns
- simplifying
- 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
Links
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a mobile context point cloud simplification algorithm based on a point feature histogram (namely a PFH operator). The algorithm mainly comprises the steps of receiving a point cloud set from a point cloud obtaining device; calculating the PFH operator of each point; calculating the standard deviation of the PFH of each point, and designing simplification rules; calculating out a corresponding relation table between the threshold values and the simplification rates of different standard deviations; according to the requirements to the simplification rate from a user, determining the threshold value; reserving the point sets with standard deviations less than or equal to the threshold value in point cloud data, in point cloud data, and eliminating other point sets. According to the mobile context point cloud simplification algorithm based on the point feature histogram, the simplification rate is high, the realtime performance is great, and adequate geometrical characteristics are reserved for scene reconstruction.
Description
Technical field
The present invention relates to a kind of mobile context point cloud compressing algorithm based on point patterns histogram (English full name is Point Feature Histogram, referred to as PFH).
Background technology
Along with the development of SLAM (English full name is Simultaneous Localization and Mapping) technology, increasing application relates to and carries imaging device by mobile robot and walk in scene, by process, the information analyzing imaging device Real-time Obtaining, obtain the complete description to whole scene, i.e. three-dimensionalreconstruction.Because mobile context cloud data amount is huge, requirement of real-time is high, thus just seems very necessary in point cloud registering, splicing advance row data compaction.Meanwhile, the cloud data for three-dimensionalreconstruction only need retain good geometric properties, makes the point set after by registration can depict the profile of original model, without the need to retaining exhibiting high surface detailed information.
Therefore, simplifying algorithm and not only will have and higher simplify ratio under mobile context, also needs to retain abundant unique point.The problem of simplifying can be defined as: the k moment is from vision point
kthe point obtained converges and is combined into M
k, ask for M
ka subset P
k, meeting default ratio:
When, make P
kretain abundant unique point, that is:
In formula (2), Fc () is unique point quality metric function, and δ is the unique point retention rate of setting.
Common point cloud compressing algorithm comprises stochastic sampling method, Triangular meshes method, simplifies algorithm and clustering methodology etc. based on point geometry feature.Wherein, based on geometric properties to simplify algorithm the most common, if publication number is 101373540, name is called the patent of " point cloud compressing system and method ", publication number is 102750730A, the patent that name is called " a kind of Cloud Points Reduction method that feature keeps ", publication number is 102890828A, name is called the patent of " the Cloud Points Reduction method based on normal direction angle ", publication number is 101021954, name is called the patent of " point cloud compressing method of 3-D scanning ", publication number is 103701466A, the patent that name is called " the dispersion point cloud compression algorithm that feature based retains " is all carry out point cloud compressing based on normal vector or curvature feature, although this kind of algorithm calculates fast, easily, but only use little several parameter values to carry out the k neighborhood geometric properties of an approximate representation point, too many information cannot be obtained, and in most of scene, comprise many unique points, adopt normal vector, curvature characteristic representation, these unique points are made to have identical or very close eigenwert, its direct result just decreases the characteristic information of the overall situation.
In addition, publication number is 103065354A, name is called that the patent of " method for point cloud optimization and device thereof " is by characteristic recovery sharp-pointed after the pre-service such as point cloud compressing and enhancing, the point cloud after optimization is enable to be convenient to the realization of reconfiguration technique, but its calculating is very complicated, and under can not meeting mobile context, Real-time Reconstruction is to ageing requirement.
Summary of the invention
The object of the invention is to propose the point cloud compressing algorithm for three-dimensionalreconstruction under a kind of mobile context.
In order to achieve the above object, technical scheme of the present invention there is provided a kind of based on point patterns histogrammic mobile context point cloud compressing algorithm, comprises the steps:
Step 1, a utilization point cloud acquisition device obtain point and converge conjunction, it is characterized in that:
Step 2, by the spatial diversity between each calculation level of parametrization and neighborhood point, calculation level converges the point patterns histogram of each point in conjunction, and wherein, point patterns histogram is the three-dimensional feature descriptor for weighing each point geometric properties;
Step 3, analysis are positioned at the histogrammic distributional difference of point patterns of scene geometric object diverse location place point cloud, and can find out: on flat surfaces, the point patterns histogram distribution of each point is more concentrated; And be positioned at edge, angle point point patterns histogram distribution more even;
Step 4, calculate the histogrammic standard deviation of point patterns of each point, wherein, the point patterns histogram of planar point due to distribution more concentrated, thus standard deviation is larger; The point patterns histogram of edge, angle point is more even owing to distributing, thus standard deviation is less, bamboo product simplifies criterion, because be mostly present in edge, corner point for the crucial geological information reconstructing solid, and the geological information that plane comprises is less, thus design is simplified criterion and is: reject the planar point that point patterns histogram criteria difference is larger, the edge that retention point feature histogram standard deviation is less, angle point;
Step 5, basis are simplified criterion and are arranged standard deviation threshold method, will put the some rejecting of converging the standard deviation of closing mid point feature histogram and being greater than standard deviation threshold method, and realize simplifying of cloud data.
Preferably, in described step 5, after realizing the simplifying of cloud data, calculate and simplify rate;
Also comprise after described step 5:
Step 6, the difference adopting step described in step 5 to calculate corresponding to various criterion difference limen value simplify rate, thus obtain standard deviation threshold method and the mapping table of simplifying rate, and wherein, standard deviation threshold method is less, then rate of simplifying is larger;
Simplify the requirement of rate needed for step 7, known experiment, standard deviation threshold method is set according to the mapping table self-adaptation that step 6 obtains, completes simplifying of cloud data.
The present invention compared with prior art, has following advantage and good effect:
What a () was higher simplifies rate, and the rate of simplifying can reach 99%.
B () is known simplifies rate requirement, and self-adaptation arranges threshold value: calculated by this algorithm and obtain various criterion difference limen value and the mapping table of simplifying rate, can be implemented in difference rate of simplifying and requires that lower self-adaptation arranges the function of threshold value.
(c) real-time: utilize the histogrammic standard deviation of point patterns to classify, not only there is good classifying quality, avoid the training process needed for structural classification device and complicated calculations simultaneously, improve algorithm counting yield, for the real-time of scene reconstruction provides possibility.
D () retains the abundant geometric properties for scene reconstruction: because PFH operator has good recognition reaction to the geometrical property of object in scene, the point namely in plane, and its point patterns histogram distribution is more concentrated; And the point patterns histogram distribution of edge, angle point is comparatively even, comprises the more angle point of geological information by retaining, rejecting and comprising the less planar point of information, making the point set after simplifying can retain the abundant geometric properties for scene reconstruction.
Accompanying drawing explanation
Fig. 1 is point cloud compressing algorithm main flow chart;
Fig. 2 is PFH calculation flow chart;
Fig. 3 is the range of influence that the PFH of calculation level calculates;
Fig. 4 is the relative deviation between 2;
Fig. 5 is the PFH feature of certain point;
Fig. 6 is the FPFH figure that physical plane is put;
Fig. 7 is the FPFH figure of actual angle point;
Fig. 8 is the algorithm flow chart carrying out data compaction according to PFH;
Fig. 9 in subset PFH standard deviation curve a little;
Figure 10 is the PFH standard deviation distribution histogram of subset.
Embodiment
For making the present invention become apparent, hereby with preferred embodiment, and accompanying drawing is coordinated to be described in detail below.
As shown in Figure 1, the invention provides a kind of based on point patterns histogrammic mobile context point cloud compressing algorithm, comprise the steps:
Step 1, a utilization point cloud acquisition device obtain point and converge conjunction, it is characterized in that:
Step 2, by the spatial diversity between each calculation level of parametrization and neighborhood point, calculation level converges the PFH operator (i.e. point patterns histogram) of each point in conjunction, wherein, point patterns histogram is the three-dimensional feature descriptor for weighing each point geometric properties.
Step 3, analysis are positioned at the histogrammic distributional difference of point patterns of scene geometric object diverse location place point cloud, and can find out: on flat surfaces, the point patterns histogram distribution of each point is more concentrated; And be positioned at edge, angle point point patterns histogram distribution more even.
Step 4, calculate the histogrammic standard deviation of point patterns of each point, wherein, the point patterns histogram of planar point due to distribution more concentrated, thus standard deviation is larger; The point patterns histogram of edge, angle point is relatively more even owing to distributing, and thus standard deviation is less.Bamboo product simplifies criterion, because be mostly present in edge, corner point for the crucial geological information reconstructing solid, and the geological information that plane comprises is less, thus design is simplified criterion and is: reject the planar point that point patterns histogram criteria difference is larger, the edge that retention point feature histogram standard deviation is less, angle point.
Step 5, calculate and simplify rate according to simplifying criterion corresponding to various criterion difference limen value, thus obtain standard deviation threshold method and the mapping table (as shown in the table) of simplifying rate, wherein, standard deviation threshold method is less, then rate of simplifying is larger.
| Standard deviation threshold method | 5 | 8 | 10 | 15 | 17 | 20 | 25 |
| Simplify rate | 99％ | 92％ | 83％ | 69％ | 55％ | 41％ | 7.3％ |
Step 6, determine the requirement of user to the rate of simplifying, standard deviation threshold method is set according to the mapping table that step 5 obtains.
Step 7, selected point converge any one point in conjunction, judge the magnitude relationship of the histogrammic standard deviation of its point patterns and standard deviation threshold method, if be less than or equal to standard deviation threshold method, then retain this point, otherwise reject this point;
Step 8, repeat step 7 until traversal point converge in conjunction institute a little, complete simplifying of cloud data.
Consult shown in Fig. 2, for calculating the point patterns histogram of each point, i.e. the concrete implementing procedure figure of PFH operator:
First acquisition point is converged and is closed M
k, calculation level converges and closes M respectively
kin the normal vector of each point, then determine the range of influence that the PFH of calculation level calculates.As shown in Figure 3, if current calculation level is P
q, by P
qbe placed on the centre position of ball, crown radius is r, P
qall k neighbors (namely with a P
qdistance be less than radius r institute a little) be all connected to each other in one network.P
qpFH operator be deviation by 2 correspondent method vectors all in parametrization neighborhood and this information formed histogram according to statistical method and obtains.For describing 2 P
sand P
tcorresponding normal n
sand n
tbetween relative deviation, the uvw coordinate system that on a point wherein, definition one is fixing, wherein the coordinate axis of uvw coordinate system press formula (3) definition:
Use the uvw coordinate system of definition in formula (3), normal n
sand n
tbetween deviation can represent by one group of angle, define by formula (4):
Choose this 3 angle character α,
θ measures normal difference, and each range of characteristic values is divided into 5 sub-ranges by normalized, and statistics drops on the number of each sub-range point, so just constitutes the proper vector of one 125 dimension, if Fig. 5 is the point patterns histogram of certain point.
Composition graphs 6,7, analyzes point patterns histogrammic distributional difference in geometric object diverse location place in scene.The cloud data used in experiment is obtained by 3D scanner single sweep operation scene, the point cloud quantity that it comprises is 2,000,000, in order to convenient data processing, point set is divided into some subsets, each subset 1000 points, utilize and tall and handsomely reach CUDA technology, parallel processing is carried out to each subset, by PFH of each point of 128 ALU parallel computations.As shown in Figure 6,7, being respectively the point patterns histogram of planar point and corner points, in order to improve computing velocity, taking FPFH (Quick-Point feature histogram) to calculate here.On a flat surface, because the normal direction the change of divergence scope of calculation level and surrounding point is smaller, so all angles component of FPFH all concentrates on specific interval, the point patterns histogram distribution of thus adding up acquisition is more concentrated; And those are positioned at edge, angle point, because the normal direction the change of divergence scope of its calculation level and surrounding point is comparatively large, all angles component of FPFH can occur on whole interval, and thus point patterns histogram distribution is more even.
Consult shown in Fig. 8, for carrying out the concrete implementing procedure figure of Cloud Points Reduction according to above PFH result of calculation:
Because PFH is a multi-dimensions histogram, and the PFH feature of solid diverse location point has significant difference in distribution, thus chooses the PFH standard deviation of each point as characteristic of division.First calculate the PFH standard deviation of each point, as shown in Figure 9, for a son concentrate PFH standard deviation curve a little.Wherein, the point corresponding to two that indicate different PFH standard deviations, be the calculation level corresponding to Fig. 6, Fig. 7 respectively, what standard deviation was little is angle point, and what standard deviation was large is planar point.
Criterion is simplified in design: the some substantial amounts being positioned at flat surfaces, and the geometric properties comprised is less, should reject as far as possible, and again because the PFH distribution in plane is concentrated, standard deviation is comparatively large, thus should reject the larger point set of PFH standard deviation.And be positioned at solid edge, shape that the point of vertex position can define solid, comprise more geometrical property, should retain as much as possible, again because the PFH at edge, summit place is evenly distributed, standard deviation is less, thus should retain the less point set of PFH standard deviation.According to above analysis, by rejecting the larger point set of standard deviation, the point set that retention criteria difference is less, simplifying of cloud data can be realized.
By the number put in statistics with histogram various criterion difference interval, as shown in Figure 10, be the PFH standard deviation distribution histogram of subset.According to above simplify criterion, arrange standard deviation threshold method, reject standard deviation and be greater than the point set of threshold value, retention criteria difference is less than or equal to the point set of threshold value, the relation that can obtain standard deviation threshold method and simplify between rate.Simplify rate and standard deviation threshold method is negative correlation: it is less that standard deviation threshold method sets, and rate of simplifying is larger; It is larger that standard deviation threshold method sets, and rate of simplifying is less.By test of many times, calculate and obtain various criterion difference limen value and the mapping table of simplifying rate, can be implemented under known difference rate of simplifying requires, self-adaptation arranges threshold value to complete simplifying of cloud data.
Claims (2)
1., based on a point patterns histogrammic mobile context point cloud compressing algorithm, comprise the steps:
Step 1, a utilization point cloud acquisition device obtain point and converge conjunction, it is characterized in that:
Step 2, by the spatial diversity between each calculation level of parametrization and neighborhood point, calculation level converges the point patterns histogram of each point in conjunction, and wherein, point patterns histogram is the three-dimensional feature descriptor for weighing each point geometric properties;
Step 3, analysis are positioned at the histogrammic distributional difference of point patterns of scene geometric object diverse location place point cloud, and can find out: on flat surfaces, the point patterns histogram distribution of each point is more concentrated; And be positioned at edge, angle point point patterns histogram distribution more even;
Step 4, calculate the histogrammic standard deviation of point patterns of each point, wherein, the point patterns histogram of planar point due to distribution more concentrated, thus standard deviation is larger; The point patterns histogram of edge, angle point is relatively more even owing to distributing, and thus standard deviation is less; Bamboo product simplifies criterion, because be mostly present in edge, corner point for the crucial geological information reconstructing solid, and the geological information that plane comprises is less, thus design is simplified criterion and is: reject the planar point that point patterns histogram criteria difference is larger, the edge that retention point feature histogram standard deviation is less, angle point;
Step 5, basis are simplified criterion and are arranged standard deviation threshold method, will put the some rejecting of converging the standard deviation of closing mid point feature histogram and being greater than standard deviation threshold method, and realize simplifying of cloud data.
2. as claimed in claim 1 a kind of based on point patterns histogrammic mobile context point cloud compressing algorithm, it is characterized in that, in described step 5, after realizing the simplifying of cloud data, calculate and simplify rate;
Also comprise after described step 5:
Step 6, adopt step described in step 5 to calculate corresponding to various criterion difference limen value different simplify rate, thus obtain standard deviation threshold method and the mapping table of simplifying rate, wherein, standard deviation threshold method is less, then rate of simplifying is larger;
Step 7, known difference simplify rate require in, standard deviation threshold method is set according to the mapping table self-adaptation that step 6 obtains, completes simplifying of cloud data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410851762.2A CN104732581A (en) | 2014-12-26 | 2014-12-26 | Mobile context point cloud simplification algorithm based on point feature histogram |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410851762.2A CN104732581A (en) | 2014-12-26 | 2014-12-26 | Mobile context point cloud simplification algorithm based on point feature histogram |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104732581A true CN104732581A (en) | 2015-06-24 |
Family
ID=53456449
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410851762.2A Pending CN104732581A (en) | 2014-12-26 | 2014-12-26 | Mobile context point cloud simplification algorithm based on point feature histogram |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104732581A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105488459A (en) * | 2015-11-23 | 2016-04-13 | 上海汽车集团股份有限公司 | Vehicle-mounted 3D road real-time reconstruction method and apparatus |
| CN106021177A (en) * | 2016-05-19 | 2016-10-12 | 兰州交通大学 | Scanning line-based three-dimensional laser scanning data compaction method |
| CN106296693A (en) * | 2016-08-12 | 2017-01-04 | 浙江工业大学 | Based on 3D point cloud FPFH feature real-time three-dimensional space-location method |
| CN107843261A (en) * | 2017-10-31 | 2018-03-27 | 国网黑龙江省电力有限公司检修公司 | A kind of method and system based on laser scanning data positioning robot position |
| CN108133191A (en) * | 2017-12-25 | 2018-06-08 | 燕山大学 | A kind of real-time object identification method suitable for indoor environment |
| CN109559340A (en) * | 2018-11-29 | 2019-04-02 | 东北大学 | A kind of parallel three dimensional point cloud automation method for registering |
| CN111629193A (en) * | 2020-07-28 | 2020-09-04 | 江苏康云视觉科技有限公司 | Live-action three-dimensional reconstruction method and system |
| CN112396080A (en) * | 2019-08-17 | 2021-02-23 | 山东理工大学 | Complex surface point cloud normal characteristic clustering and grading estimation method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102944174A (en) * | 2012-11-28 | 2013-02-27 | 北京矿冶研究总院 | Point cloud data processing method and system |
| US20130147972A1 (en) * | 2011-12-13 | 2013-06-13 | Fujitsu Limited | User detecting apparatus, user detecting method and computer-readable recording medium storing a user detecting program |
-
2014
- 2014-12-26 CN CN201410851762.2A patent/CN104732581A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130147972A1 (en) * | 2011-12-13 | 2013-06-13 | Fujitsu Limited | User detecting apparatus, user detecting method and computer-readable recording medium storing a user detecting program |
| CN102944174A (en) * | 2012-11-28 | 2013-02-27 | 北京矿冶研究总院 | Point cloud data processing method and system |
Non-Patent Citations (2)
| Title |
|---|
| 孙殿柱等: ""基于型面特征的三维散乱点云精简算法"", 《中国机械工程》 * |
| 马兆建: ""虚拟现实辅助遥操作机器人的非结构化环境建模研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105488459A (en) * | 2015-11-23 | 2016-04-13 | 上海汽车集团股份有限公司 | Vehicle-mounted 3D road real-time reconstruction method and apparatus |
| CN106021177A (en) * | 2016-05-19 | 2016-10-12 | 兰州交通大学 | Scanning line-based three-dimensional laser scanning data compaction method |
| CN106021177B (en) * | 2016-05-19 | 2018-10-23 | 兰州交通大学 | A kind of laser scanning data compressing method based on scan line |
| CN106296693A (en) * | 2016-08-12 | 2017-01-04 | 浙江工业大学 | Based on 3D point cloud FPFH feature real-time three-dimensional space-location method |
| CN106296693B (en) * | 2016-08-12 | 2019-01-08 | 浙江工业大学 | Based on 3D point cloud FPFH feature real-time three-dimensional space-location method |
| CN107843261A (en) * | 2017-10-31 | 2018-03-27 | 国网黑龙江省电力有限公司检修公司 | A kind of method and system based on laser scanning data positioning robot position |
| CN108133191A (en) * | 2017-12-25 | 2018-06-08 | 燕山大学 | A kind of real-time object identification method suitable for indoor environment |
| CN109559340A (en) * | 2018-11-29 | 2019-04-02 | 东北大学 | A kind of parallel three dimensional point cloud automation method for registering |
| CN112396080A (en) * | 2019-08-17 | 2021-02-23 | 山东理工大学 | Complex surface point cloud normal characteristic clustering and grading estimation method |
| CN111629193A (en) * | 2020-07-28 | 2020-09-04 | 江苏康云视觉科技有限公司 | Live-action three-dimensional reconstruction method and system |
| CN111629193B (en) * | 2020-07-28 | 2020-11-10 | 江苏康云视觉科技有限公司 | Live-action three-dimensional reconstruction method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104732581A (en) | Mobile context point cloud simplification algorithm based on point feature histogram | |
| CN109993748B (en) | Three-dimensional grid object segmentation method based on point cloud processing network | |
| Yang et al. | Automatic 3d facial expression recognition using geometric scattering representation | |
| US20180165511A1 (en) | Three-dimensional facial recognition method and system | |
| CN108107444B (en) | Transformer substation foreign matter identification method based on laser data | |
| CN111079685B (en) | 3D target detection method | |
| WO2016023264A1 (en) | Fingerprint identification method and fingerprint identification device | |
| CN108320329A (en) | A kind of 3D map creating methods based on 3D laser | |
| CN105806315B (en) | Noncooperative target relative measurement system and measuring method based on active coding information | |
| CN105957076A (en) | Clustering based point cloud segmentation method and system | |
| CN111723721A (en) | Three-dimensional target detection method, system and device based on RGB-D | |
| CN104123746B (en) | The computational methods of real-time normal vector in a kind of 3-D scanning point cloud | |
| CN110807781B (en) | Point cloud simplifying method for retaining details and boundary characteristics | |
| CN103824050A (en) | Cascade regression-based face key point positioning method | |
| CN104504709B (en) | Feature ball based classifying method of three-dimensional point-cloud data of outdoor scene | |
| CN102339376A (en) | Classifying and processing method based on active shape model and K nearest neighbor algorithm for facial forms of human faces | |
| CN102708364B (en) | Cascade-classifier-based fingerprint image classification method | |
| CN102708370A (en) | Method and device for extracting multi-view angle image foreground target | |
| CN104038792B (en) | For the video content analysis method and apparatus of IPTV supervision | |
| CN105354578A (en) | Multi-target object image matching method | |
| CN106980812A (en) | Three-dimensional face features' independent positioning method based on concatenated convolutional neutral net | |
| CN106231553B (en) | Multinode information based on wireless acoustic sensor network merges sound localization method | |
| CN111145129A (en) | Point cloud denoising method based on hyper-voxels | |
| CN104537353A (en) | Three-dimensional face age classifying device and method based on three-dimensional point cloud | |
| CN103902992A (en) | Human face recognition method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20150624 |