CN106952297A - A kind of laser scanning data point cloud degree compression method - Google Patents
A kind of laser scanning data point cloud degree compression method Download PDFInfo
- Publication number
- CN106952297A CN106952297A CN201710173764.4A CN201710173764A CN106952297A CN 106952297 A CN106952297 A CN 106952297A CN 201710173764 A CN201710173764 A CN 201710173764A CN 106952297 A CN106952297 A CN 106952297A
- Authority
- CN
- China
- Prior art keywords
- point
- data
- cloud
- compression
- point cloud
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T9/00—Image coding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
Abstract
A kind of laser scanning data point cloud degree compression method of the invention, belongs to Point Cloud Data from Three Dimension Laser Scanning compression method, and cloud data compression is implemented in combination with point-connectivity theory in graph theory.This compression method is applicable not only to dispersion point cloud, is equally applicable to the orderly point cloud of distribution rule；Single site point cloud before registration is applicable not only to, multi-site point cloud after registration is equally applicable to.Four functions of this method can be used alone, and can also repeat and mutually undertaking is used, directive property compression is with conspicuous characteristics.It as open cloud data compressed platform, can continue to extend other compression functions, it is adaptable to the cloud data compression of different geometric shapes and compression purpose, expand space larger.
Description
Technical field
The invention belongs to Point Cloud Data from Three Dimension Laser Scanning compression method, it is implemented in combination with a little with point-connectivity theory in graph theory
Cloud data compression.
Technical background
With the sustainable development of three-dimensional laser scanning technique and being substantially improved for scanning device precision, the distribution of Bian sampling points is more
Crypto set is with accurately, and obtained point cloud quantity can be quite huge, also just inevitably generates mass of redundancy data point.These
Redundant digit strong point brings inconvenience to the further application of the cloud data in later stage, and takes big quantity space in the data transmission
And resource, reduce these redundant points imperative.In addition, the target compressed sometimes is not necessarily redundant points cloud, for non-superfluous
Remaining cloud, according to later stage application target difference, it is necessary to be compressed to significant initial cloud data, reservation presumptive area,
Predetermined point is studied.Such compression algorithm only one of which purpose, meets the different compression requirement of user to greatest extent.
The content of the invention
The present invention proposes the stronger point cloud degree compression method of practicality.This compression method is applicable not only to dispersion point cloud,
It is equally applicable to the orderly point cloud of distribution rule；Single site point cloud before registration is applicable not only to, multistation after registration is equally applicable to
Point point cloud.Four functions of this method be can be used alone, and can also repeat and mutually undertaking is used, and directive property compression feature is dashed forward
Go out.As open cloud data compressed platform, can continue to extend other compression functions, it is adaptable to different geometric shapes and
The cloud data compression of purpose is compressed, is expanded space larger.
The yulan tree scattered point cloud data chosen herein for the three observation station for acquiring in ground carries out registration and obtains original
Cloud data is illustrated.It is first according to each point digital complex demodulation in data (the z values in three-dimensional coordinate) from low to high
Order carries out lossless arrangement again；Then carried out according to the point cloud of rearrangement according to user's request piecemeal, in each cloud mass
A virtual point is inside obtained respectively, and the coordinate of this virtual point is exactly the average value of each point coordinates in block, then obtained respectively in block
Distance of each point to this block virtual point；Finally sorted from small to large again in each piece according to this distance value.Two minor sorts
Sequence number allows for each point and imparts new coordinate, the lookup and deletion that can be pinpointed according to this sequence number.
Technical solution of the present invention is a kind of laser scanning data point cloud degree compression method, and this method includes：
Step 1：Vegetation cloud data is obtained from different orientation by multiple websites, then by the cloud data of each website
It is unified to arrive under same coordinate system, realize the registration of each website cloud data；
Step 2：First according to x in three-dimensional coordinate, y, z is worth size to formulate data point sort method, will own in step 1
Data point is ranked up according to the sort method of formulation, obtains sequence cloud data collection N；
Step 3：Required according to compression, the suitable point cloud number n of selection carries out piecemeal, the row for obtaining step 2 in units of n
Sequence cloud data collection N, carries out piecemeal in order, and that is finally discontented with n data point is individually for one piece；
Step 4：The mean point V of all data points in each piece is calculated first0, each data point is calculated in block to mean point
V0Distance and angle, each data point in block is ranked up again further according to distance, each point sequence sequence number is obtained；
Step 5：By each point sequence sequence number in block and the point to mean point V0Distance and angle set up corresponding relation, from
And each piece obtains one by sequence number, the corresponding distance of each sequence number and angle, this block mean point V0The packet of composition, realizes number
According to compression.
Further, the specific method of the step 1 is：From a point converge P in choose a number of data point, and
The corresponding data point of these points is searched out in consecutive points converge Q；Then, by least square method iteration, optimal seat is calculated
Transformation parameter is marked, even if error function：In R and t it is minimum, wherein R is spin moment
Battle array, t is translation vector, and n is the sum for the data point chosen, qkK-th of data point in Q, p are converged for the point of selectionkTo choose
Point converge k-th of data point in P；By the smart registration result at preceding two station by identical step and the 3rd adjacent site cloud data
Registration is carried out, comprehensive complete individual plant vegetation cloud data can be finally obtained.
Point cloud degree compression method is applicable not only to dispersion point cloud compression, is also applied for orderly point cloud compression；It is applicable not only to match somebody with somebody
Single site point cloud compression, is equally applicable to multi-site point cloud compression after registration before accurate.Point cloud degree compression method can just be realized multiple
Function, and replace other a variety of compression methods, user will be not used in wasting twists and turns on selection compression method, repeat alternately or
Undertaking can just complete the effect that other compression methods reach using difference in functionality.
Brief description of the drawings
Fig. 1 is yulan tree original point cloud data schematic diagram；
Fig. 2 is a cloud degree structural representation；
Fig. 3 is leaf original point cloud data schematic diagram；
Fig. 4 is schematic diagram data after the compression of 30 points of execution distal end；
Fig. 5 is schematic diagram data after the 40 points of compressions of execution near-end；
Fig. 6 pinpoints schematic diagram data after 49.9% compression to perform；
Fig. 7 is present invention point cloud degrees of data compression method flow chart.
Specific implementation step
1) data acquisition：
In order to obtain complete vegetation cloud data, it is necessary to be scanned from different azimuth to vegetation, and by will be adjacent
Target in website cloud data carries out pressure as control point and met, and can complete the rough registration to multistation cloud data, i.e.,
Point cloud under different coordinates is substantially unified under the same coordinate system.Then, by smart registration, it can make between multi-site cloud
Stitching error reaches minimum.If P, Q converge for the point of the same object in different station for acquiring, and point p, q are measured object body surface
Conformation of the face any point in different coordinates point converges, i.e. p (xp,yp,zp) ∈ P, q (xq,yq,zq) ∈ Q, cloud data
Smart registration seeks to make the point of two points any expression body surface same point in converging to (p, q), meets identical conversion,
I.e.：
Wherein, R is spin matrix, and t is translation vector.The present invention use ICP algorithm (iterative closet point algorithm,
Iterative Closest Point), it has main steps that：First, it is assumed that an initial position and state estimation, i.e., from
Point converges and chosen in P a number of point, and searches out in consecutive points converge Q the corresponding points of these points；Then, minimum is passed through
Square law iteration, calculates optimal coordinate conversion parameter, even if error function
Minimum R and t.The registration process of two station vegetation cloud datas is as shown in Figure 2.The smart registration result at preceding two station is pressed
Identical step is registering with the 3rd site cloud data progress, can finally obtain comprehensive complete individual plant vegetation cloud data.
2) point cloud arrangement：
The object of scanner scanning any shape, the data of acquisition are all the cloud data collection of a limited range, are obtained
Individual plant yulan tree point cloud chart picture such as Fig. 1.Because the arrangement of these cloud data collection is unordered, study for convenience, by this
A little point clouds re-start once lossless arrangement according to the orders of DEM (the z values in three-dimensional coordinate) from low to high.Point cloud after arrangement
Rearranged according to new sequence number, this sequence number is just to confer to the repositioning information of dispersion point cloud.With z values from low to high
In alignment processes, if there is z value identical situations, just compare x values, x values are larger to be arranged in front, it is less be arranged in after,
If x values are still identical, continue to compare y values, y values are larger to be arranged in front, it is less be listed in after.
3) cloud piecemeal is put：
The cloud data collection N rearranged, requires that the suitable point cloud number n of selection carries out piecemeal, with n according to the compression of user
Started counting up for unit from DEM minimum values, full n restarts next count and circulated.Unit often comprising a n cloud we regard
For one piece, the unit of a remaining discontented n clouds is individually considered as one piece, and whole point cloud thus is divided into M blocks.It is expressed as：
M=int (N/n)+1 (3)
4) arranged again in block：
To the x of an every piece of n clouds (last block is less than n), y, z averages respectively:
According to Dist values, (each point is apart from block center virtual point V0Distance value) n point in block is carried out again once from
It is small to arrive big lossless arrangement.V0With being formed after n line as shown in Fig. 2 V in figure0Point is just V in n degree point, virtual point cloud mass0
Point cloud degree (point cloud degree) just be n.Virtual point V0Connect remaining each point to greatest extent in figure, and answering
Play a part of control in, associate remaining each real point.Monoblock point cloud is constituted after point cloud degree network, it is meant that huge three-dimensional
Each point (being n point in each block altogether) in each block of scattered point cloud data can be with virtual center point V0Between all can
There is the relation of a distance, be also n distance value of correspondence altogether so between minimum distance to maximum distance, so that
In per module unit, with virtual point V0The n that is connected clouds have been assigned the location information from 1 to n, at this moment from 1 to any of n
One numeral and V0Number combine just represent a point, just can more facilitate to enter any block point cloud by such method
Row two-dimensional digitalization is retrieved, and is that next step compression lays the foundation.
5) method is realized：
Monoblock point cloud constitute point cloud degree network after, it is meant that huge unorganized point cloud repartitioned into by
V0The two-dimensional coordinate system that number is constituted with n values.And in every module unit, with virtual point V0The n that is connected clouds have been assigned from
1 arrives n location information, at this moment any one numeral and V from 1 to n0Number combine just represent a point, by such
Method just can more facilitate carries out two-dimensional digital retrieval to any block point cloud, is that next step compression lays the foundation.
Below by taking the cloud data (4583 points) of yulan tree leaf as an example, such as Fig. 3.Each function is presented in order to protrude
Effect, the point cloud of leaf is divided into 46 pieces by us according to every piece of 100 points here.
A. remote point compresses
According to user's needs, to V0Remote point is compressed, and this function has denoising, deburring, integer for cloud data
Function.Such as Fig. 4, when reducing 30 of distal end, the noise of blade distal end is deleted, and the peripheral shape of blade is closer
In actual blade.
B. proximal points are compressed
For V0The compression of proximal points, this function has for cloud data to be put cloud, extracts point cloud framework in sparse piece
Function, perspectiveization is had more in the case where profile is integrally constant.Such as Fig. 5, when the proximal points of compression are, blade is being kept at 40
On the premise of profile, the translucency of blade interior is more obvious, and perspectiveization is protruded.
C. interval point compresses
Needed to set respective bins according to user, and the point cloud in interval is compressed.With the pressure to designated area
Contracting function, compression possesses certain directive property.
D. specified point compresses
It is compressed, is reduce to the unwanted specified point of user for specified point in cloud data, directive property is more dashed forward
Go out, compression effectiveness is obvious.The function is enough when cloud data piecemeal, such as carries out 2292 pieces of divisions to 4583 points of point cloud, often
Block only has at 2 points, when a point to every piece carries out specified point compression, rate 49.9891% is exactly compressed to a cloud and is compressed,
Because in 2 points of adjoint points each other in upper piece, the global feature after compression will not be lost, and compression efficiency is high.Such as Fig. 6.
This 4 kinds of functions can carry out repeatedly or accept according to different requirements to compress to same data, to reach expection
Effect.Because destroying the quantity of original point cloud and being distributed limited and process control, compression process belongs to Lossless Compression.Point
Cloud degree compression method flow is as shown in Figure 7.Algorithm is realized using Windows7, VS2010 platform herein, C++ application programs
Exploitation is completed, auxiliary Geomagic Studio 12 complete point cloud and shown.
Claims (2)
1. a kind of laser scanning data point cloud degree compression method, this method includes：
Step 1：Vegetation cloud data is obtained from different orientation by multiple websites, it is then that the cloud data of each website is unified
To under same coordinate system, the registration of each website cloud data is realized；
Step 2：First according to x in three-dimensional coordinate, y, z is worth size to formulate data point sort method, by all data in step 1
Point is ranked up according to the sort method of formulation, obtains sequence cloud data collection N；
Step 3：Required according to compression, the suitable point cloud number n of selection carries out piecemeal, the sequence point for obtaining step 2 in units of n
Cloud data set N, carries out piecemeal in order, and that is finally discontented with n data point is individually for one piece；
Step 4：The mean point V of all data points in each piece is calculated first0, each data point is calculated in block to mean point V0's
Distance and angle, are ranked up to each data point in block again further according to distance, obtain each point sequence sequence number；
Step 5：By each point sequence sequence number in block and the point to mean point V0Distance and angle set up corresponding relation so that respectively
Block obtains one by sequence number, the corresponding distance of each sequence number and angle, this block mean point V0The packet of composition, realizes data pressure
Contracting.
2. a kind of laser scanning data point cloud degree compression method as claimed in claim 1, it is characterised in that the step 1
Specific method be：From a point converge P in choose a number of data point, and search out these in consecutive points converge Q
The corresponding data point of point；Then, by least square method iteration, optimal coordinate conversion parameter is calculated, even if error function：In R and t it is minimum, wherein R is spin matrix, and t is translation vector, and n is the number chosen
The sum at strong point, qkK-th of data point in Q, p are converged for the point of selectionkK-th of data point in P is converged for the point of selection；Will
The smart registration result at preceding two station is registering with the 3rd adjacent site cloud data progress by identical step, can finally obtain comprehensive
Complete individual plant vegetation cloud data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710173764.4A CN106952297A (en) | 2017-03-22 | 2017-03-22 | A kind of laser scanning data point cloud degree compression method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710173764.4A CN106952297A (en) | 2017-03-22 | 2017-03-22 | A kind of laser scanning data point cloud degree compression method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106952297A true CN106952297A (en) | 2017-07-14 |
Family
ID=59473631
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710173764.4A Pending CN106952297A (en) | 2017-03-22 | 2017-03-22 | A kind of laser scanning data point cloud degree compression method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106952297A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019210531A1 (en) * | 2018-05-03 | 2019-11-07 | 北京大学深圳研究生院 | Point cloud attribute compression method based on deleting 0 elements in quantisation matrix |
| CN111598803A (en) * | 2020-05-12 | 2020-08-28 | 武汉慧点云图信息技术有限公司 | Point cloud filtering method based on variable resolution voxel grid and sparse convolution |
| WO2021248339A1 (en) * | 2020-06-09 | 2021-12-16 | 深圳市大疆创新科技有限公司 | Point cloud encoding/decoding method and apparatus |
| US11535400B2 (en) * | 2020-05-09 | 2022-12-27 | Nanjing University Of Aeronautics And Astronautics | Fairing skin repair method based on measured wing data |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103886625A (en) * | 2014-01-09 | 2014-06-25 | 北京工业大学 | Point cloud data sparse representation method based on compressed sensing |
| CN105551015A (en) * | 2015-12-02 | 2016-05-04 | 南京邮电大学 | Scattered-point cloud image registering method |
| CN105931234A (en) * | 2016-04-19 | 2016-09-07 | 东北林业大学 | Ground three-dimensional laser scanning point cloud and image fusion and registration method |
-
2017
- 2017-03-22 CN CN201710173764.4A patent/CN106952297A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103886625A (en) * | 2014-01-09 | 2014-06-25 | 北京工业大学 | Point cloud data sparse representation method based on compressed sensing |
| CN105551015A (en) * | 2015-12-02 | 2016-05-04 | 南京邮电大学 | Scattered-point cloud image registering method |
| CN105931234A (en) * | 2016-04-19 | 2016-09-07 | 东北林业大学 | Ground three-dimensional laser scanning point cloud and image fusion and registration method |
Non-Patent Citations (1)
| Title |
|---|
| 吴宝峰: "阔叶树激光雷达点云数据压缩方法研究", 《中国优秀硕士学位论文全文数据库 农业科技辑》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019210531A1 (en) * | 2018-05-03 | 2019-11-07 | 北京大学深圳研究生院 | Point cloud attribute compression method based on deleting 0 elements in quantisation matrix |
| US11216985B2 (en) | 2018-05-03 | 2022-01-04 | Peking University Shenzhen Graduate School | Point cloud attribute compression method based on deleting 0 elements in quantisation matrix |
| US11535400B2 (en) * | 2020-05-09 | 2022-12-27 | Nanjing University Of Aeronautics And Astronautics | Fairing skin repair method based on measured wing data |
| CN111598803A (en) * | 2020-05-12 | 2020-08-28 | 武汉慧点云图信息技术有限公司 | Point cloud filtering method based on variable resolution voxel grid and sparse convolution |
| CN111598803B (en) * | 2020-05-12 | 2023-05-09 | 武汉慧点云图信息技术有限公司 | Point cloud filtering method based on variable-resolution voxel grid and sparse convolution |
| WO2021248339A1 (en) * | 2020-06-09 | 2021-12-16 | 深圳市大疆创新科技有限公司 | Point cloud encoding/decoding method and apparatus |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106952297A (en) | A kind of laser scanning data point cloud degree compression method | |
| CN106600683B (en) | A kind of adaptive method for reconstructing of grid model towards Bone CT sequence image | |
| CN109919206A (en) | A kind of remote sensing image ground mulching classification method based on complete empty convolutional neural networks | |
| CN104408055B (en) | The storage method and device of a kind of laser radar point cloud data | |
| Aardt et al. | Forest volume and biomass estimation using small-footprint lidar-distributional parameters on a per-segment basis | |
| CN102368237B (en) | Image retrieval method, device and system | |
| CN109446986B (en) | Effective feature extraction and tree species identification method for tree laser point cloud | |
| CN110070567A (en) | A kind of ground laser point cloud method for registering | |
| CN105654483A (en) | Three-dimensional point cloud full-automatic registration method | |
| CN104679831A (en) | Method and device for matching human model | |
| CN112990228B (en) | Image feature matching method, related device, equipment and storage medium | |
| CN108537865A (en) | A kind of the pseudo-classic architecture model generation method and device of view-based access control model three-dimensional reconstruction | |
| CN106093875A (en) | Near field WB-RCS data acquisition and the compression method of denoising is followed the trail of based on base | |
| CN111091030A (en) | Tree species identification method and device, computer equipment and readable storage medium | |
| CN106777133A (en) | A kind of similar connection processing method of metric space based on MapReduce | |
| CN106570166A (en) | Video retrieval method and apparatus based on multiple partial sensitive hash tables | |
| CN111222581A (en) | Cloud storage gesture recognition system and method based on improved classification algorithm | |
| CN110162812B (en) | Target sample generation method based on infrared simulation | |
| CN107123138A (en) | Based on vanilla R points to rejecting tactful point cloud registration algorithm | |
| CN108287940B (en) | Method and device for arranging sampling points, electronic equipment and storage medium | |
| CN103606189B (en) | A kind of track base system of selection towards non-rigid three-dimensional reconstruction | |
| CN112215893B (en) | Method, device and equipment for determining target two-dimensional center coordinate point and ranging system | |
| CN107301618B (en) | Based on the GPU basis matrix accelerated parallel and homography matrix estimation method and system | |
| CN111507430B (en) | Feature coding method, device, equipment and medium based on matrix multiplication | |
| CN110399285A (en) | A kind of test case based on weak analysis of variance about subtracts method |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170714 |