CN110111373A - A kind of artificial bee colony point cloud registration algorithm based on curvature reconnaissance and gravity treatment point - Google Patents
A kind of artificial bee colony point cloud registration algorithm based on curvature reconnaissance and gravity treatment point Download PDFInfo
- Publication number
- CN110111373A CN110111373A CN201910336186.0A CN201910336186A CN110111373A CN 110111373 A CN110111373 A CN 110111373A CN 201910336186 A CN201910336186 A CN 201910336186A CN 110111373 A CN110111373 A CN 110111373A
- Authority
- CN
- China
- Prior art keywords
- point cloud
- point
- curvature
- cloud
- registration
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The artificial bee colony point cloud registration algorithm based on curvature reconnaissance and gravity treatment point that the invention discloses a kind of, this method extract characteristic point according to curvature information, obtain the optimal mapping matrix that two panels point cloud can be made to be overlapped by improving artificial bee colony algorithm optimization object function.Corresponding points are constrained according to curvature information during swarm optimization and find range, reduce the scale for participating in calculating point cloud.Comparative experiments shows that compared with only with random selecting point method and the registration Algorithm of point of use cloud spatial coordinated information etc., this paper algorithm can effectively accelerate registration convergence rate while not reducing registration accuracy, significant to shorten the time used in point cloud registering.
Description
Technical field
The present invention relates to Computerized three-dimensional digital image processing field more particularly to a kind of people based on curvature reconnaissance and gravity treatment point
Worker bee group's point cloud registration algorithm.
Background technique
In three-dimensional imaging field, the registration technique of three-dimensional point cloud be it is vital, it is optimal European its task is to find one
It converts, under the point Cloud transform to the same coordinate system to overlap for obtaining same object from different perspectives, finally recovers
The complete pattern of testee.The registration process of three-dimensional point cloud is substantially an optimization process.It is asked solving the optimization
When topic, traditional method for registering is easy to be influenced by quantization error, point cloud noise and point cloud initial position.It is excellent based on colony intelligence
The three-dimensional point cloud registration Algorithm of change compensates for the shortcomings that registration Algorithm of the tradition based on gradient optimizing better, effectively improves
The adaptability and registration accuracy of point cloud registering process, but there are still being registrated, convergence rate is lower compared with slow and algorithm success rate to be lacked
Point.
Therefore, it is necessary to relevant arts can accelerate point cloud registering convergence rate, improve registration success rate.
Summary of the invention
Purpose of the invention is to overcome the shortcomings in the prior art, for using only colony intelligence optimization algorithm and point cloud space
When information carries out point cloud registering, optimization process is found two panels point cloud corresponding points and is taken a long time, and the slower disadvantage of convergence rate provides
A kind of artificial bee colony point cloud registration algorithm based on curvature reconnaissance and gravity treatment point.
The purpose of the present invention is what is be achieved through the following technical solutions:
A kind of artificial bee colony point cloud registration algorithm based on curvature reconnaissance and gravity treatment point, comprising the following steps:
Step 1. is directed to the point cloud that two panels to be registered partially overlaps, and a piece of is dynamic point cloud, and another is static point cloud, to dynamic
State point cloud extracts partial dot cloud by stochastical sampling and according to Point cloud curvature intelligence sample two ways, as dynamic point cloud
Collection, realizes the data compaction of point cloud registering；
Step 2. is finding the corresponding points stage according to the difference of Point cloud curvature range, excludes different curvature range in subject to registration cloud
Partial dot cloud, reduce the searching range of corresponding points, reduce algorithm calculation amount；
Step 3. uses point cloud data in Stanford point cloud data library as being registrated using artificial bee colony optimization algorithm is improved
The point cloud used is established the objective function of evaluation registration effect and is solved, and finding is completely coincident subject to registration cloud of two panels
Optimal transformation；During Optimized Iterative, dynamic point cloud is replaced every fixed number of iterations.
Further, " during Optimized Iterative, the point that registration uses is replaced every fixed number of iterations in step 3
Cloud ", the specific implementation steps are as follows:
Variable needed for step 301. initializes optimization algorithm, including population population, maximum number of iterations and search equation are crucial
Parameter；And dynamic point cloud is subjected to rotation translation by the angle and translational movement randomly selected, later with equidistant reconnaissance
Method selects sampling point set；
The position of step 302. initialization Search of Individual；
Change values are still below threshold value then by the reconnaissance mode according to curvature before and after optimal solution after 10 generation of step 303. subsequent iteration
Gravity treatment point；
Vi,j=Xbest,j+φi,j(Xr1,j-Xr2,j) (1)
Vi,j=Xr1,j+α(Xbest,j-Xr1,j)+β(Xr1,j-Xr2,j) (2)
Wherein, XbestFor the optimum individual of current group；φi,jIt is the random number between [- 1,1]；R1 ≠ r2 ≠ i is from set
Randomly selected integer in { 1,2 ... NP }；A randomly selected integer in j ∈ { 1,2 ... D }；α is between [0, A]
A random number, β rand*B, here B be a mean value be μ, standard deviation be δ Gaussian Profile number；
If step 304. rand < cr, using formula (1), a new candidate solution otherwise is generated using formula (2)；And compare update
The position of Search of Individual i, if the search of all individuals finishes and thens follow the steps 305, otherwise return step 303；
Step 305. calculates select probability P using formula (3)i；
Wherein, fitiIt indicates the target function value of i-th of particle, shares NP particle；
Step 306. is assumed to share Foodnumber particle, if rand < Pi, step 307 is executed, otherwise i=i+1, if i
=Foodnumber, i=1；Repeat step 306；
Change values are still below threshold value then by random selecting point mode gravity treatment point before and after step 307. subsequent iteration 70 generation optimal solution；
If step 308. rand < cr, using formula (1), a new candidate solution otherwise is generated using formula (4)；And compare update
The position of Search of Individual i, searching times are equal to Search of Individual number and then follow the steps 309, otherwise return step 306；
Vi,j=Xr1,j+α(Xbest,j-Xr1,j)+β(Xr1,j-Xbest,j) (4)
If step 309. reaches maximum limited number of times limit, a solution replacement X is randomly generated using formula (5)i；
Vi,j=Xi,j+φi,j(Xi,j-Xk,j) (5)
In step 310. every 300 generation of iteration, is forced to update globally optimal solution；
Step 311. judges gravity treatment point threshold value；
If step 312. reaches restriction maximum number of iterations or target function value lower than target value, current optimum individual is brought into
It calculates final error value and exports；Otherwise return step 303.
Further, step 1 is specific as follows:
Step 101. carries out stochastical sampling to dynamic point cloud and extracts sampled point；
Step 102. carries out curvature range division according to Point cloud curvature information, to dynamic point cloud；
Step 103. carries out curvature range division according to Point cloud curvature information, to static point cloud；
Step 104. extracts sampled point in dynamic point cloud according to Point cloud curvature information；
Step 105. is adopted using sampled point obtained through stochastical sampling and according to the sampled point that curvature information is extracted with the ratio of 2:1 altogether
Collect 300 points, collectively constitutes subject to registration cloud and participate in subsequent calculating.
Compared with prior art, the beneficial effects brought by the technical solution of the present invention are as follows: the present invention is calculated based on bee colony optimization
The curvature information that a cloud is introduced in the point cloud registration method of method completes the registration of three-dimensional point cloud, is not reducing registration accuracy
Under the premise of accelerate algorithm the convergence speed, improve point cloud registering success rate.
Detailed description of the invention
Fig. 1 is a cloud region division schematic diagram.
Fig. 2 is sampling point acquisition schematic diagram.
Specific embodiment
The present invention is described in further detail below in conjunction with the drawings and specific embodiments.It should be appreciated that tool described herein
Body embodiment is only used to explain the present invention, is not intended to limit the present invention.
Artificial bee colony point cloud registration algorithm provided in this embodiment based on curvature reconnaissance and gravity treatment point, comprising:
Step 100, the point cloud to partially overlap for two panels to be registered, by stochastical sampling and according to Point cloud curvature intelligence sample
Two ways extracts partial dot cloud, realizes the data compaction of point cloud registering；
Step 200, the corresponding points stage is being found according to the difference of Point cloud curvature range, exclude subject to registration cloud mean curvature range phase
The biggish partial dot cloud of difference, reduces the searching range of corresponding points, reduces algorithm calculation amount；
Step 300, using artificial bee colony optimization algorithm is improved, objective function is solved, finding can make two panels subject to registration
The optimal transformation that point cloud is completely coincident.During Optimized Iterative, according to certain condition, replaces and match every certain the number of iterations
The point cloud that standard uses.
Step 400, rotation translation is carried out to dynamic point cloud using obtained optimal transformation, two panels is observed by software meshlab
Point cloud is overlapped effect.
In the present embodiment, step 100 method includes:
Step 101, stochastical sampling is carried out to dynamic point cloud and extracts sampled point；
Step 102, according to Point cloud curvature information, curvature range division is carried out to dynamic point cloud；
Step 103, according to Point cloud curvature information, curvature range division is carried out to static point cloud；
Step 104, sampled point is extracted in dynamic point cloud according to curvature information；
Step 105, it is total to according to certain proportion using sampled point obtained through stochastical sampling and according to the sampled point that curvature information is extracted
Subsequent calculating is participated in subject to registration cloud of composition.
Further, step 102 includes:
For each of dynamic point cloud P point, respective Gaussian curvature is calculated using the method based on the surface MLS
kGaussianWith average curvature kMean.And thus obtain principal curvatures
Wherein, k1、k2It is minimum, maximum principal curvatures.According to the definition of Chen et al., we indicate every bit with following formula
Curvature information
Wherein, point p represents any point in point cloud, k1(p)、k2(p) the main song of minimum of point p is respectively represented with 0 < S (p) < 1
Rate, the curvature information of maximum principal curvatures and the point.Dynamic point cloud can be expressed as P={ p at this timei,Si, i=1,2 ..., Np。
Each point in point cloud has the curvature information S of oneself, S is equally divided into from 0 to 1 n sections, the curvature letter of point each in this way
Breath just belongs to one section therein.As Fig. 2 puts cloud dividing condition such as Fig. 1 as n=3.It is different according to the affiliated range of curvature information,
Dynamic point cloud P can be divided into n cloud subset Pk, k=1,2 ..., n.Point cloud subset PkIn only meet formula (28) comprising curvature information S
Point:
Further, step 103 includes:
For each of static point cloud Q point, respective Gaussian curvature is calculated using the method based on the surface MLS
kGaussianWith average curvature kMean.And thus obtain principal curvatures
Wherein, k1、k2It is minimum, maximum principal curvatures.According to the definition of Chen et al., we indicate every bit with following formula
Curvature information]
Wherein, point p represents any point in point cloud, k1(p)、k2(p) the main song of minimum of point p is respectively represented with 0 < S (p) < 1
Rate, the curvature information of maximum principal curvatures and the point.Static point cloud can be expressed as Q={ q at this timen,Sn, j=1,2 ..., Nq。
Each point in point cloud has the curvature information S of oneself, S is equally divided into from 0 to 1 n sections, the curvature letter of point each in this way
Breath just belongs to one section therein.Different according to the affiliated range of curvature information, static point cloud Q can be divided into n cloud subset Qw, w
=1,2 ..., n.Point cloud subset QwIn only meet the point of formula (32) comprising curvature information S.
Further, step 104 are as follows:
When sampling using curvature information to a cloud, a number k is randomly selected into n from 1, if the son of dynamic point cloud P
Collect PkIt is not sky, and the corresponding subset Q in static point cloud QwIt also is not sky, then from PkIn randomly select a little as extraction
Characteristic point.This step is recycled until obtaining the curvature feature point of fixed number.Assuming that will be according to Point cloud curvature information extraction
NfeatureA characteristic point, then data reduction process is as shown in Fig. 2, PsampledIt indicates to converge conjunction according to the point that curvature information is extracted.
In order to avoid being registrated failure caused by the factors such as local fit in registration process, then with the mode of stochastical sampling in a cloud P
Middle extraction NrandomA sampled point, two parts point cloud combine the son point cloud constituted eventually for registration.
Further, " the reducing the searching range of corresponding points, reduce algorithm calculation amount " in step 200 includes:
Curvature range difference between corresponding points is smaller, so a point p in finding dynamic point cloud Pi, in static point cloud Q
When corresponding points, without being found in the entire scope of Q, it is only necessary to differ in lesser range and find in curvature information.It seeks
It finds corresponding points and then calculates corresponding points square distance intermediate value, target function value of the acquired results as optimization algorithm.Think
Dynamic point cloud subset PkThe curvature information at midpoint and static point cloud subset QkThe curvature information at midpoint is similar.For PkIn
Point only need to be in QkMiddle searching corresponding points.Furthermore QkThe curvature information range at midpoint is comprising PkThe curvature information range at midpoint,
It avoids accurately find corresponding points with the point of curvature information segmented edges in this way.Curvature information is the point of S, belonging to
The serial number k of point cloud subset can be obtained by formula (33).
Wherein,It indicates to be rounded downwards.It, can each subset Q to static point cloud in order to accelerate search speedkEstablish kd-
Tree data structure is indexed.
Further, step 300 are as follows:
For dynamic point cloud P={ p to be registeredi, i=1,2 ..., NpWith static point cloud Q={ qj, j=1,2 ..., Nq, ask
The optimal European transformation matrix T between two panels point cloud is solved, keeps the corresponding points of two o'clock cloud intersection completely heavy in space coordinates
It closes.The transformation matrix includes 6 undetermined parameters, respectively along three translation of axes amount Tx,Ty,TzAnd around 3 reference axis
Rotation angle α, β, γ.In the case where not considering that model size is stretched, the expression formula of transformation matrix T are as follows:
T=RXRYRZS (34)
Wherein
In the ideal situation, a point p in PiThe point T (pi) obtained after T is converted and its corresponding points q in QjBetween away from
From shoulding be 0.In fact, their distance can indicate are as follows:
di=| | T (Pi)-Qj||2 (39)
But due to the influence of the factors such as measurement error and noise, the distance between corresponding points are unable to reach ideal value 0, because
This, point cloud registering problem has translated into optimization problem: seeking an optimal transformation matrix T and makes all corresponding points to T
(pi) and qjBetween Euclidean distance it is minimum, i.e.,
In conjunction with artificial bee colony optimization algorithm, using corresponding points apart from intermediate value as objective function F (T)
Wherein, MedSE indicates every a pair of of corresponding points square distance d in subject to registration cloud of two panelsi 2Intermediate value.
Using improved artificial bee colony algorithm --- EABC algorithm, which optimizes objective function, seeks optimal solution.
The search equation difference that EABC algorithm gathering honey bee and observation bee stage use is as follows:
Vi,j=Xr1,j+α(Xbest,j-Xr1,j)+β(Xr1,j-Xbest,j) (42)
Vi,j=Xr1,j+α(Xbest,j-Xr1,j)+β(Xr1,j-Xr2,j) (43)
Using EABC algorithm optimization objective function, optimal value is solved, its step are as follows:
For i=1,2 ..., Np
Step 1: the transformation matrix T obtained using optimization process is to piRotation translation is carried out, point T (p is obtainedi)；
Step 2: p is calculated according to formula 33iK value；
Step 3: in QkMiddle searching piCorresponding points qi；
Step 4: calculating T (Pi) and qiEuclidean distance di=| | T (pi)-qi||2；
End
Step 5: by formulaTarget function value is calculated.
Due to the randomness of initial reconnaissance process, selection is concentrated there may be distribution for the point cloud during Optimized Iterative
Problem.The subject to registration cloud chosen may concentrate on some part of a cloud.It is likely to result in registration ambiguity, therefore success rate is also
Wait improve.In response to this problem, the present invention proposes a kind of gravity treatment point invention based on curvature feature point.
It is i.e. every then to update sampled point and obtained through stochastical sampling sampling of the part according to curvature information acquisition by certain algebra
Point.
In view of sampling essence have randomness, if but the number of iterations it is enough, be iterated calculating with different sampling point sets
It is enough even more by overwhelming majority point in dynamic point cloud traversal one time.The point position obtained when if encountering sampling is bad, weight
The combination of different point sets can correct ambiguity to a certain extent after reconnaissance, the correct search process for guiding swarm intelligence algorithm.
Whenever the number of iterations reaches certain numerical value, in the gathering honey bee stage in artificial bee colony optimization algorithm, believe according to curvature
The sampled point of breath acquisition carries out part update, carries out gravity treatment point.
Whenever the number of iterations reaches certain numerical value, in the observation bee stage in artificial bee colony optimization algorithm, stochastical sampling is obtained
The sampled point arrived carries out part update, carries out gravity treatment point.
The point cloud putting cloud and not updating that two parts obtain participates in next registration optimization process jointly.
The specific implementation steps are as follows:
Step a initializes each variable, and dynamic point cloud is carried out rotation translation by the angle and translational movement randomly selected,
Sampling point set is selected with equidistant selected-point method later.
A certain number of Search of Individual are randomly generated in step b in restriction range, initialize the position of Search of Individual.
Change values are still below threshold value then by the reconnaissance mode gravity treatment according to curvature before and after optimal solution after 10 generation of step c subsequent iteration
Point.
Vi,j=Xbest,j+φi,j(Xr1,j-Xr2,j) (44)
Vi,j=Xr1,j+α(Xbest,j-Xr1,j)+β(Xr1,j-Xr2,j) (45)
If step d rand < cr, using formula (44), a new candidate solution otherwise is generated using formula (45).And compare update
The position of Search of Individual i, if the search of all individuals finishes and thens follow the steps 5, otherwise return step 3.
Step e calculates select probability P using formula (46)i。
If step f rand < Pi, step 7 is executed, otherwise i=i+1, if i=Foodnumber, i=1.Repeat step
6。
Change values are still below threshold value then by random selecting point mode gravity treatment point before and after step g subsequent iteration 70 generation optimal solution.
If step h rand < cr, using formula (45), a new candidate solution otherwise is generated using formula (47).And compare update
The position of Search of Individual i, searching times are equal to Search of Individual number and then follow the steps 9, otherwise return step 6.
Vi,j=Xr1,j+α(Xbest,j-Xr1,j)+β(Xr1,j-Xbest,j) (47)
If step i reaches maximum limited number of times limit, a solution replacement X is randomly generated using formula (48)i。
Vi,j=Xi,j+φi,j(Xi,j-Xk,j) (48)
In step j every 300 generation of iteration, is forced to update globally optimal solution.
Step k judges gravity treatment point threshold value.
If step l reaches restriction maximum number of iterations or adaptive value lower than target value, brings current optimum individual into and calculate most
Whole error amount simultaneously exports.Otherwise return step 3.
The present invention is not limited to embodiments described above.Description and explanation sheet is intended to the description of specific embodiment above
The technical solution of invention, the above mentioned embodiment is only schematical, is not restrictive.The present invention is not being departed from
In the case of objective and scope of the claimed protection, those skilled in the art can also make under the inspiration of the present invention
The specific transformation of many forms, within these are all belonged to the scope of protection of the present invention.
Claims (3)
1. a kind of artificial bee colony point cloud registration algorithm based on curvature reconnaissance and gravity treatment point, which comprises the following steps:
Step 1. is directed to the point cloud that two panels to be registered partially overlaps, and a piece of is dynamic point cloud, and another is static point cloud, to dynamic
State point cloud extracts partial dot cloud by stochastical sampling and according to Point cloud curvature intelligence sample two ways, as dynamic point cloud
Collection, realizes the data compaction of point cloud registering；
Step 2. is finding the corresponding points stage according to the difference of Point cloud curvature range, excludes different curvature range in subject to registration cloud
Partial dot cloud, reduce the searching range of corresponding points, reduce algorithm calculation amount；
Step 3. uses point cloud data in Stanford point cloud data library as being registrated using artificial bee colony optimization algorithm is improved
The point cloud used is established the objective function of evaluation registration effect and is solved, and finding is completely coincident subject to registration cloud of two panels
Optimal transformation；During Optimized Iterative, dynamic point cloud is replaced every fixed number of iterations.
2. the artificial bee colony point cloud registration algorithm according to claim 1 based on curvature reconnaissance and gravity treatment point, feature exist
In, " being to replace the point cloud that uses of registration every fixed number of iterations during Optimized Iterative " in step 3, specific implementation
Steps are as follows:
Variable needed for step 301. initializes optimization algorithm, including population population, maximum number of iterations and search equation are crucial
Parameter；And dynamic point cloud is subjected to rotation translation by the angle and translational movement randomly selected, later with equidistant reconnaissance
Method selects sampling point set；
The position of step 302. initialization Search of Individual；
Change values are still below threshold value then by the reconnaissance mode according to curvature before and after optimal solution after 10 generation of step 303. subsequent iteration
Gravity treatment point；
Vi,j=Xbest,j+φi,j(Xr1,j-Xr2,j) (1)
Vi,j=Xr1,j+α(Xbest,j-Xr1,j)+β(Xr1,j-Xr2,j) (2)
Wherein, XbestFor the optimum individual of current group；φi,jIt is the random number between [- 1,1]；R1 ≠ r2 ≠ i is from set
Randomly selected integer in { 1,2 ... NP }；A randomly selected integer in j ∈ { 1,2 ... D }；α is between [0, A]
A random number, β rand*B, here B be a mean value be μ, standard deviation be δ Gaussian Profile number；
If step 304. rand < cr, using formula (1), a new candidate solution otherwise is generated using formula (2)；And compare update
The position of Search of Individual i, if the search of all individuals finishes and thens follow the steps 305, otherwise return step 303；
Step 305. calculates select probability P using formula (3)i；
Wherein, fitiIt indicates the target function value of i-th of particle, shares NP particle；
Step 306. is assumed to share Foodnumber particle, if rand < Pi, step 307 is executed, otherwise i=i+1, if i=
Foodnumber, i=1；Repeat step 306；
Change values are still below threshold value then by random selecting point mode gravity treatment point before and after step 307. subsequent iteration 70 generation optimal solution；
If step 308. rand < cr, using formula (1), a new candidate solution otherwise is generated using formula (4)；And compare update
The position of Search of Individual i, searching times are equal to Search of Individual number and then follow the steps 309, otherwise return step 306；
Vi,j=Xr1,j+α(Xbest,j-Xr1,j)+β(Xr1,j-Xbest,j) (4)
If step 309. reaches maximum limited number of times limit, a solution replacement X is randomly generated using formula (5)i；
Vi,j=Xi,j+φi,j(Xi,j-Xk,j) (5)
In step 310. every 300 generation of iteration, is forced to update globally optimal solution；
Step 311. judges gravity treatment point threshold value；
If step 312. reaches restriction maximum number of iterations or target function value lower than target value, current optimum individual is brought into
It calculates final error value and exports；Otherwise return step 303.
3. the artificial bee colony point cloud registration algorithm according to claim 1 based on curvature reconnaissance and gravity treatment point, feature exist
In step 1 is specific as follows:
Step 101. carries out stochastical sampling to dynamic point cloud and extracts sampled point；
Step 102. carries out curvature range division according to Point cloud curvature information, to dynamic point cloud；
Step 103. carries out curvature range division according to Point cloud curvature information, to static point cloud；
Step 104. extracts sampled point in dynamic point cloud according to Point cloud curvature information；
Step 105. is adopted using sampled point obtained through stochastical sampling and according to the sampled point that curvature information is extracted with the ratio of 2:1 altogether
Collect 300 points, collectively constitutes subject to registration cloud and participate in subsequent calculating.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910336186.0A CN110111373A (en) | 2019-04-24 | 2019-04-24 | A kind of artificial bee colony point cloud registration algorithm based on curvature reconnaissance and gravity treatment point |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910336186.0A CN110111373A (en) | 2019-04-24 | 2019-04-24 | A kind of artificial bee colony point cloud registration algorithm based on curvature reconnaissance and gravity treatment point |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110111373A true CN110111373A (en) | 2019-08-09 |
Family
ID=67486644
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910336186.0A Pending CN110111373A (en) | 2019-04-24 | 2019-04-24 | A kind of artificial bee colony point cloud registration algorithm based on curvature reconnaissance and gravity treatment point |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110111373A (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104835151A (en) * | 2015-04-24 | 2015-08-12 | 南京邮电大学 | Improved artificial bee colony algorithm-based image registration method |
| CN105654063A (en) * | 2016-01-08 | 2016-06-08 | 东南大学 | Motor imagery EEG pattern recognition method based on time-frequency parameter optimization of artificial bee colony |
| CN109300149A (en) * | 2018-09-30 | 2019-02-01 | 天津商业大学 | The three-dimensional image registration method optimized based on gravity treatment point strategy and artificial bee colony |
-
2019
- 2019-04-24 CN CN201910336186.0A patent/CN110111373A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104835151A (en) * | 2015-04-24 | 2015-08-12 | 南京邮电大学 | Improved artificial bee colony algorithm-based image registration method |
| CN105654063A (en) * | 2016-01-08 | 2016-06-08 | 东南大学 | Motor imagery EEG pattern recognition method based on time-frequency parameter optimization of artificial bee colony |
| CN109300149A (en) * | 2018-09-30 | 2019-02-01 | 天津商业大学 | The three-dimensional image registration method optimized based on gravity treatment point strategy and artificial bee colony |
Non-Patent Citations (4)
| Title |
|---|
| YONGCUN CAO 等: "An improved global best guided artificial bee colony algorithm for continuous optimization problems", 《CLUSTER COMPUTING》 * |
| 付鲲 等: "基于曲率信息的人工蜂群点云配准算法", 《计算机应用研究》 * |
| 段渊: "一种新的多目标人工蜂群算法", 《系统科学与数学》 * |
| 熊小峰等: "精英区域学习的转轴人工蜂群算法", 《四川大学学报(工程科学版)》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104392426B (en) | A kind of no marks point three-dimensional point cloud method for automatically split-jointing of self adaptation | |
| CN107886529A (en) | A kind of point cloud registration method for three-dimensional reconstruction | |
| CN110490912B (en) | 3D-RGB point cloud registration method based on local gray level sequence model descriptor | |
| CN109903319B (en) | Multi-resolution-based fast iteration closest point registration algorithm | |
| CN108470358B (en) | Point cloud registration method based on second-order oscillation artificial bee colony algorithm | |
| CN109767463A (en) | A kind of three-dimensional point cloud autoegistration method | |
| CN111415379B (en) | Three-dimensional point cloud data registration method based on cuckoo optimization | |
| CN107423678A (en) | A kind of training method and face identification method of the convolutional neural networks for extracting feature | |
| CN110211129B (en) | Low-coverage point cloud registration algorithm based on region segmentation | |
| CN108932536A (en) | Human face posture method for reconstructing based on deep neural network | |
| CN104835153B (en) | Non-rigid surface's alignment schemes based on rarefaction representation | |
| CN112132752B (en) | Fine splicing method for multi-view scanning point cloud of large complex curved surface | |
| CN107730542A (en) | Cone beam computed tomography image corresponds to and method for registering | |
| CN107610219A (en) | The thick densification method of Pixel-level point cloud that geometry clue perceives in a kind of three-dimensional scenic reconstruct | |
| CN105631899A (en) | Ultrasonic image motion object tracking method based on gray-scale texture feature | |
| CN108597016A (en) | Torr-M-Estimators basis matrix robust estimation methods based on joint entropy | |
| CN117274339A (en) | Point cloud registration method based on improved ISS-3DSC characteristics combined with ICP | |
| CN110120008A (en) | A kind of artificial bee colony point cloud registration method based on local curvature information | |
| CN108062766B (en) | Three-dimensional point cloud registration method fusing color moment information | |
| CN117496205A (en) | Heterogeneous scene matching method based on ITHM-Net | |
| CN110111373A (en) | A kind of artificial bee colony point cloud registration algorithm based on curvature reconnaissance and gravity treatment point | |
| CN110415281B (en) | Loam curvature weighting-based point set rigid registration method | |
| CN104978721B (en) | Atmospheric perturbation image recovery method based on variational regularization | |
| US20220130048A1 (en) | System and method for estimating motion of target inside tissue based on surface deformation of soft tissue | |
| CN109977855A (en) | A kind of face's method for registering based on student-t distribution |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190809 |
|
| WD01 | Invention patent application deemed withdrawn after publication |