CN106023298A - Point cloud rigid registration method based on local Poisson curved surface reconstruction - Google Patents
Point cloud rigid registration method based on local Poisson curved surface reconstruction Download PDFInfo
- Publication number
- CN106023298A CN106023298A CN201610456738.8A CN201610456738A CN106023298A CN 106023298 A CN106023298 A CN 106023298A CN 201610456738 A CN201610456738 A CN 201610456738A CN 106023298 A CN106023298 A CN 106023298A
- Authority
- CN
- China
- Prior art keywords
- point
- registration
- curved surface
- poisson
- sampling point
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 78
- 230000009466 transformation Effects 0.000 claims abstract description 15
- 238000005070 sampling Methods 0.000 claims description 54
- 238000013459 approach Methods 0.000 claims description 9
- 230000000007 visual effect Effects 0.000 claims description 8
- HUTDUHSNJYTCAR-UHFFFAOYSA-N ancymidol Chemical compound C1=CC(OC)=CC=C1C(O)(C=1C=NC=NC=1)C1CC1 HUTDUHSNJYTCAR-UHFFFAOYSA-N 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 4
- 238000000926 separation method Methods 0.000 claims description 4
- 230000007547 defect Effects 0.000 claims description 3
- 238000010845 search algorithm Methods 0.000 claims 1
- 238000007667 floating Methods 0.000 abstract description 7
- 238000013461 design Methods 0.000 abstract description 4
- 238000004519 manufacturing process Methods 0.000 abstract 1
- 230000008569 process Effects 0.000 description 14
- 238000010586 diagram Methods 0.000 description 8
- 230000000694 effects Effects 0.000 description 5
- 235000008331 Pinus X rigitaeda Nutrition 0.000 description 3
- 235000011613 Pinus brutia Nutrition 0.000 description 3
- 241000018646 Pinus brutia Species 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000006978 adaptation Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000003203 everyday effect Effects 0.000 description 2
- 230000000452 restraining effect Effects 0.000 description 2
- 239000007787 solid Substances 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 241000283973 Oryctolagus cuniculus Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000013256 coordination polymer Substances 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000003780 insertion Methods 0.000 description 1
- 230000037431 insertion Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 230000013011 mating Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 238000012372 quality testing Methods 0.000 description 1
- 238000007634 remodeling Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T15/00—3D [Three Dimensional] image rendering
- G06T15/10—Geometric effects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/14—Transformations for image registration, e.g. adjusting or mapping for alignment of images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2215/00—Indexing scheme for image rendering
- G06T2215/06—Curved planar reformation of 3D line structures
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- Computer Graphics (AREA)
- Processing Or Creating Images (AREA)
Abstract
The invention discloses a point cloud rigid registration method based on local sample Poisson curved surface reconstruction to find a corresponding point, and belongs to the field of digital design and manufacturing. The method is characterized by, in crude registration, mutually selecting feature point pairs from a floating point cloud and a fixed point cloud, and establishing a Poisson curved surface based on a neighborhood point set of feature points of the fixed point cloud; establishing a KD tree of the curved surface; searching a closest point of a sample point of the floating point cloud in the KD tree, and serving the closest point as a reference point; serving the closest point from the sample point to a reference point ring domain surface patch as the corresponding point, and establishing a measure function based on a corresponding point pair and calculating transformation parameters through an SVD method; and on the basis of crude registration, in precise registration, obtaining feature point pairs in a self-adaptive manner based on a public domain, and establishing error metric by utilizing the minimum distance from the point to the Poisson curved surface, so that transformation parameters can be calculated, and registration precision is further improved. In crude registration, higher registration precision can be achieved, and the precise registration can quickly converge to the global optimum and has higher robustness.
Description
Technical field
The present invention provides some cloud Rigid Registration method based on local Poisson curve reestablishing, can be used for surface in kind sampling
The registration of data various visual angles cloud data, belongs to digitized design platform field.
Background technology
In fields such as reverse-engineering, computer graphics, quality testings, need to obtain three-dimensional point cloud number from surface in kind
According to.Due to object surface block and scanning device measurement scope limit, the laser measuring equipment of current main flow and light
Grid projection measures equipment must could obtain, from multiple angles, subregion scanning, the point that body surface is whole to object being measured
Cloud data.But in translation-angle scanning process, the coordinate system at the cloud data place that scanning obtains is different every time, it is therefore desirable to
In cloud data under different coordinates is converted into the same coordinate system and be output as a complete surface three dimension cloud data,
I.e. three-dimensional point cloud registration.The precision of three-dimensional point cloud registration determines the precision of various process follow-up to sampled data, the most right
The precision of the post processings such as point cloud segmentation in reverse-engineering, feature identification, Curvature Estimation, normal estimation, curve reestablishing has important
Impact.
Three-dimensional point cloud registration process is generally divided into just registration and essence two stages of registration.Just registrate the stage, fix one and regard
The cloud data (fixing point cloud) at angle, other visual angle cloud data that floats (float some cloud), by coupling fixing point cloud and floating
The corresponding geometric properties of some cloud public territory calculates rigid transformation parameters, and the some cloud that floats the most at last with fixing point cloud point cloud registration and integration is
One complete cloud data;Essence is registered in iteration registration process on the basis of just registrating, until error convergence, thus further
Improve just registration accuracy, make error minimize.
Just in registration, conventional corresponding geometric properties has corresponding point, homologous pair and corresponding surface.Chua etc. are at " 3D human
Face recognition using point signature " (IEEE International Conference on
Automatic Face and Gesture Recognition, 2000. Proceedings. IEEE, 2000:233-
238) it is that each some defined feature describes son, finds in fixing point cloud by calculating the Feature Descriptor of each point in the some cloud that floats
Corresponding Feature Descriptor, the method is computationally intensive and sensitive to noise.Papazov etc. are at " An efficient
Ransac for 3d object recognition in noisy and occluded scenes " (Computer
Vision ACCV 2010. Springer Berlin Heidelberg, 2011:135-148.) use RANSAC side
Method, in two groups of some clouds find three points right, if the spacing approximately equal of each any two points of centering; would think this three
Individual point, to being corresponding point pair, calculates transformation parameter based on point to information, and the method is appropriate only for the point cloud registering of small data quantity.
Bucksch etc. are at " Localized registration of point clouds of botanic trees "
(Geoscience and Remote Sensing Letters, IEEE, 2013,10 (3): 631-635.) uses point
The mode that line is corresponding registrates, and calculates at the Eigenvector that fixing point cloud is closest by finding arbitrfary point in the some cloud that floats
Rigid transformation parameters, the method requires problem that is higher and that cannot solve local convergence to a cloud contour shape.Above method
Based on the geometric properties that whole cloud data coupling is corresponding, and then calculating transformation parameter, this type of method amount of calculation is very big, and
Distinct methods a cloud initial data is had flatness, uniformity, without the different requirements such as noise.Scholar is additionally also had to use algebraically bent
Surface model method, the principal curvatures estimation technique etc. calculate corresponding geometric properties information.In addition can use a kind of relatively easy, effective
Just method for registering man-machine interaction method manually chooses corresponding point pair, can be counted fast and efficiently by method of least square
Calculate transformation parameter, but the greatest problem of the method is that the corresponding point precision to choosing is the highest, cause just registration accuracy relatively low, very
To converging on local optimum.
The main stream approach of current essence registration have focused largely on to closest approach iteration (Iterative Closet Point,
ICP) improvement of method, the method by Besl etc. at " Method for registration of 3-D shapes "
(Robotics-DL tentative. International Society for Optics and Photonics, 1992:
586-606.) proposing, ICP elaborates the basic theories framework of essence registration: to having two groups of good initial relative position information
Cloud data, can use all of geometric graphic element (point, line, surface, body etc.) be iterated registration and seek globally optimal solution.At ICP
In, author uses some the geometric graphic element as registration, by iterative closest point calculate corresponding point to and calculate based on Quaternion Method
Rigid transformation parameters, but the method cannot solve the problem of the local convergence caused greatly because of initial positional deviation, and iteration time
Number is more.Chen etc. are at " Object modeling by registration of multiple range images "
(Robotics and Automation, 1991. Proceedings., 1991 IEEE International
Conference on. IEEE, 1991:2724-2729.) propose a kind of based on set point normal direction to impact point incisal plane
Error estimation criterion, accelerates convergence rate, but the method is inapplicable to the Model registration that Curvature varying is big, and require two
Model has abundant overlapping region." the Semantic 3d object maps for everyday such as Rusu
Manipulation in human living environments " (KI-K ü nstliche Intelligenz, 2010,
24 (4): 345-348.) propose Feature Descriptor based on FPFH and carry out Feature Points Matching, and add characteristic point criterion,
Candidate feature point for identifying is got rid of by the geological information such as curvature, normal direction, further for incongruent characteristic point
Being rejected, whole process uses KD tree to be accelerated finding, and the method achieves extraordinary registration effect, therefore the method
A cloud algorithms library PCL that increased income uses, but the method does not the most solve the drawback of depending on initial phase unduly to position, as long as
The initial phase of two point cloud models is undesirable to position, may result in registration and lost efficacy.Various improved methods are more smart by finding
True geometric graphic element improves registration accuracy and sets up the speed of more efficient error estimation rule raising convergence, but bring
Problem is to calculate to complicate, and time cost is high, and the cloud data that each method processes often exists particularity, does not possess pervasive
Property.
Summary of the invention
It is an object of the invention to improve the accuracy that in registration process, corresponding point are mated at PD, make just to registrate quickly
Restraining and improve the precision of just registration, essence registration can reduce registration error further and improve the robustness of essence registration, technical side
Case is accomplished by
A kind of some cloud Rigid Registration method based on local Poisson curve reestablishing, it is characterised in that step is followed successively by: one, sets and waits to join
The accurate surface in kind sampling point set under two different visual anglesWith, willAs the some cloud that floats,As fixing point cloud, pass through
The mode of man-machine interactively respectively fromWithIn choose subsetWith;Two, setFor empty set,Interior mutual selected part
Characteristic point, and selected characteristic point is sequentially added into;Three, setFor empty set, forInterior each sampling point,Interior mutual
Choose matched sampling point, and institute's sampling point is sequentially added into;Four, based onInterior sampling point existsIn neighborhood point
Collection, forInterior each sampling point builds Poisson curved surface, concretely comprises the following steps: each sample neighborhood of a point point set is added auxiliary magnet by (1)
Build and close point set;(2) closing point set is carried out normal estimation;(3) completely moor based on the closing point set after normal estimation
Pine curve reestablishing;(4) the local Poisson curved surface corresponding to neighborhood point set is separated from complete Poisson curved surface, with local pool
Pine curved surface is as the Poisson curved surface of sampling point;Five, withIn sampling point to the distance of corresponding Poisson curved surface as estimating,
Solve so that the rigid transformation matrix of point set registration function minimization, thus completeWithPreliminary registration, i.e. makeWithIt is fully overlapped under the same coordinate system;Six, willReset to empty set,Interior selected part sampling point as new characteristic point,
Selection rule is to ensure that institute's sampling point existsThere is complete neighborhood point set, i.e. this sampling point existInterior neighborhood point set can be equal
The even surrounding adjacent regions being distributed in this sampling point, adds institute's sampling point;Seven, willReset to empty set, forInterior is every
Individual sampling point,Inside choose the most nearest sampling point, and institute's sampling point is sequentially added into;Eight, method described in applying step four
ForInterior each sampling point builds Poisson curved surface;Nine, withThe distance conduct of interior sampling point extremely corresponding Poisson curved surface
Estimate, solve so that the rigid transformation matrix of point set registration function minimization, thus completeWithAccuracy registration, result is defeated
After going out rigid transformationWith registration error.
For realizing goal of the invention, described some cloud Rigid Registration method based on local Poisson curve reestablishing, its feature exists
In: in step 4, to sample neighborhood of a point point set add auxiliary magnet structure close point set, concretely comprise the following steps: (1) set sampling point as ,Neighborhood point set be, willProjection, to plane, is designated as;(2) known by boundary characteristic recognition methods
Do not go outBoundary point set;(3) according to the corresponding relation of projection, obtainBoundary point set;(4) based on,Between corresponding boundary point, insert discrete point, closing point set can be built.
For realizing goal of the invention, described some cloud Rigid Registration method based on local Poisson curve reestablishing, its feature exists
In: in step 4, by building dynamically spatial-data index, the Poisson curved surface corresponding to neighborhood point set is separated, concrete steps
For: (1) set constructed complete Guan Bi Poisson curved surface as, set up based on Poisson curved surfaceSpatial index KD tree with half of
The index structure that structure combines;(2) arest neighbors in the index of the sampling point in inquiry neighborhood point set;(3) net after separation is set
Lattice curved surface is, topology information based on Half-edge Structure, an annulus and the dough sheet information thereof of arest neighbors are stored in;(4) output
Locally Poisson curved surface, it is ensured thatKeep the topological integrity of original data point surface information.
For realizing goal of the invention, described some cloud Rigid Registration method based on local Poisson curve reestablishing, its feature exists
In: in step 5, withIn sampling point to the minimum distance of Poisson curved surface as estimating, concretely comprise the following steps: (1) is based on locally
Poisson curved surfaceSet up grid index structure KD tree;(2) right, searched by the K-NN search of KD treeExtremely's
Grid vertex recently;(3) topological neighborhood information based on index leaf node storage obtainsAn annulus tri patch;(4) profit
Calculate with point-to-plane distance formulaClosest approach to annulus tri patch, willAsCorresponding point;(5) calculateWith's
Distance, using this distance as measure value.
For realizing goal of the invention, described some cloud Rigid Registration method based on local Poisson curve reestablishing, its feature exists
In: in step 6,Interior selected part sampling point is as new characteristic point, concretely comprising the following steps of selection rule: (1) calculates point
CollectionSpan;(2) based onNeighborhood point set, calculateSpan;(2) setFor selecting ginseng
Number, if, thenThere is the defects such as hole, rejectOtherwise, willAs characteristic point.
For realizing goal of the invention, described some cloud Rigid Registration method based on local Poisson curve reestablishing, its feature exists
In: in the step (1) of step 6 selected characteristic point rule, calculate point setSpan, concretely comprise the following steps: a) from table in kind
Surface sample point setIn obtain at randomIndividual sampling point, forms point set;B) setFor empty set, forIn each sampling point, calculate
It arrivesIt is nearest for middle-rangeThe distance average of individual sampling point is also added into setIn;C) willThe average of middle all elements is made
For span。
For realizing goal of the invention, described some cloud Rigid Registration method based on local Poisson curve reestablishing, its feature exists
In, in step 9, use minimum range based on sampling point Poisson curved surface to carry out error estimation, withTo local Poisson
Curved surfaceMinimum range as error, concretely comprise the following steps: (1),, wherein;(2) calculateArriveClosest approach;(3) calculateEuclidean distance;(4);(5), repeat step (2)-(4)
Until;(6);(8) return error。
The present invention compared with prior art, has the advantage that
(1) carry out boundary characteristic identification based on fractional sample data and build closing point set, improve in Poisson process of reconstruction
The robustness of littleization yardstick indicator function;
(2) just registrate manually selected characteristic point, use a searching strategy for the corresponding point arriving Poisson curved surface, at the beginning of improve
Registration accuracy;
(3) essence registration can solve the problem causing registration to lose efficacy because first registration position deviation is big, improves the robust of registration
Property, it is to avoid registration is absorbed in local optimum in an iterative process;
(4) radial error estimation criterion based on a Poisson curved surface, can significantly improve the precision of registration, decrease registration
The number of times of iteration convergence.
Accompanying drawing explanation
Fig. 1 is the flow chart of present invention point based on local Poisson curve reestablishing cloud Rigid Registration method;
Fig. 2 is the schematic diagram that fractional sample span is estimated;
Fig. 3 is the organigram closing point set;
Fig. 4 is the schematic diagram of Poisson curve reestablishing and separation thereof;
Fig. 5 is based on a schematic diagram for Poisson curved surface closest approach coupling;
Fig. 6 is that registration error analyzes schematic diagram;
Fig. 7 is the mathematical model figure in embodiment;
Fig. 8 is to use the present invention and Geomagic Studio to carry out just to registrate effect contrast figure;
Fig. 9 is to use the present invention and Geomagic Studio to carry out just registration accuracy comparison diagram;
What Figure 10 was feature point number on essence registration accuracy affects schematic diagram;
Figure 11 is to use the present invention and ICP method (Besl P J, McKay N D. Method for registration of
3-D shapes[C]. Robotics-DL tentative. International Society for Optics and
Photonics, 1992:586-606.), improve ICP method (Rusu R B. Semantic 3d object maps for
everyday manipulation in human living environments[J]. KI-Künstliche
Intelligenz, 2010,24 (4): 345-348.) carry out essence registration comparison diagram;
Figure 12 is three visual angle registration design sketchs of Hood model;
Figure 13 is six visual angle registration design sketchs of Bunny model.
Detailed description of the invention
Below in conjunction with the accompanying drawings and example the invention will be further described.
Fig. 1 is the flow chart of present invention Rigid Registration method based on local Poisson curve reestablishing, at the some Yun Hegu that floats
The PD of fixed point cloud selects initial corresponding point pair, recalculates corresponding point, Jin Erxiu according to a Poisson SURFACES MATCHING rule
Just initial corresponding point pair, calculate rigid transformation parameters and registration error based on revised point to information.In initial corresponding point pair
Selection course in, just registration is by choosing alternately, and essence registration is chosen by public territory self adaptation.Initial corresponding point are to choosing
Afterwards, using point to neighborhood point as the fractional sample data of registration, local sample data is extracted boundary characteristic build pool
Pine curved surface, calculating sample point, to the closest approach of Poisson curved surface, revises initial corresponding point pair.Based on revised corresponding point to letter
Breath, sets up the measure function of least square and estimates rigid transformation parameters, and to floating, some cloud is made rigid transformation and calculates registration by mistake
Difference, iterated transform process is until restraining.
A cloud is randomly selected fractional sample, calculates the span of each fractional sample data, and then estimate whole some cloud
Average span.In the fractional sample data shown in Fig. 2, some cloud span isd 1,d 2,……,d 7Arithmetic average etc. each distance
Value.
By to fractional sample project after extract the boundary point that fractional sample projects with it respectively, correspondence boundary point it
Between insert discrete point, and then build close point set, as it is shown on figure 3, close point set build concretely comprise the following steps: (1)If,
Discrete point setH;(2) to fractional sampleProjection is to plane point set, projector distance is5 times;(3) based on limit
Boundary's feature recognition algorithms identifiesBorder setAndBoundary point set;(4) based on,Meter
Calculate the distance between the corresponding boundary point of point;(5) average span according to a cloud calculates the number of insertion discrete point
;(6) according to the vector property of vector, calculate the coordinate of discrete point and discrete point is stored inH;(7), repetition step (5)-
(6) until;(8) traversal,In all corresponding boundary points, repeat step (4)-(7) obtain discrete point setH。
The purpose building closing point set is to improve the quality of Poisson curve reestablishing, from Fig. 4 Poisson curve reestablishing and separation thereof
Schematic diagram in can be seen that rebuild Poisson curved surface is uniform, fine and close, the defect such as cavity, deformity does not occur, can be met this
Bright registration needs, and retains the topological structure integrity of mask data, can be that registration searching corresponding point offer is joined more accurately
Examination point.
The matched rule of sampling point Poisson curved surface can be set up, with sample point based on the local Poisson curved surface builtTo local
Poisson curved surfaceClosest approachFor corresponding point, as shown in Figure 5.
For the result of relative error result Yu actual registration, after figure end partial enlargement registration result as shown in Figure 6, figure
Middle light color solid dot represents floating point cloud characteristic pointNeighborhood point set, dark solid dot represents fixing point cloud feature
PointNeighborhood point set, curve represents the curved surface of reconstruction.As can be seen from the figureWith Poisson curved surfaceNot
Being completely superposed, reason is that Poisson curve reestablishing is rightMost preferably approaching of set up scaling function, and Poisson's equation is
The differential form of elliptical equation, so curved surfaceThe deformation of elliposoidal can occur.Curved surface deformation introduces Poisson reconstruction error, this
Time error is estimated to need to considerPosition pair,Impact.If,Represent respectively,In all
Point arrivesThe average of minimum range, in the result of actual registration,Position exist two kinds of situation: Yi ZhongshiIt is positioned at,Between (Fig. 6-a), the error now calculated is then
(1)
Another kind of situation is exactly,It is positioned atThe same side (shown in Fig. 6-b), the error that now calculates is actual is
(2)
Therefore it is right to need the when of calculating errorPosition effectively judge, normal direction auxiliary law can be used to determine: right,If,,It is a pair corresponding point pair,For?In nearest grid vertex,,Point
It is not,The normal direction at place, then judge the factorFor:
If, then belong to the first situation, if, then the second situation is belonged to.According to above-mentioned analysis, by calculatingArriveMinimum range inaccurate as error, Poisson curve reestablishing can cause reconstruction error, therefore carry out error estimation time
Time needs to considerImpact, willSubstituting in formula (1) and (2) as compensating error, the result just obtaining error estimation is actual
On be that sample point arrivesMinimum range, i.e.
The final purpose using error criterion is to judge that registration is the most effective according to error amount, but error size and cloud data
Itself has much relations, if some cloud span is big, it is rough that Poisson rebuilds curved surface so that registration error strains greatly mutually;If some cloud across
Spending little, some cloud is the most intensive, and Poisson rebuilds rear curved surface relative smooth, and registration error can diminish.Right for eliminating cloud data itself
The impact that registration error judges, the present invention defines registration efficiency factor, by this factor and error, some cloud spanFoundation is joined
The most effective Rule of judgment of standard: if, then judge just to registrate inefficacy;If, then judge that essence registration lost efficacy, wherein's
Computing formula is。
Testing with models such as the Hood model shown in Fig. 7, Rabbit and analyze, wherein Hood model passes through CPC light
Spatial digitizer gathers, and some cloud span is 3;Bunny model uses the cloud data that Stanford University gathers, and puts cloud span and is
0.001。
Embodiment one: in registration at the beginning of Hood, Bunny, the present invention intersection in two width view not common regions is joined
Standard is the roundest and the most smooth, although Geomagic Studio registration does not has the process of iteration convergence, the shortest, but Geomagic
Studio registration occurs in that bigger deviation occurring in that is substantially misaligned, as shown in Figure 8.The first method for registering of the present invention passes throughCarry out mating corresponding point, during finding corresponding point,Can suitably take little (), expand Poisson surface mesh
The number on summit, increases searching scope, reaches more preferable registration accuracy;IfObtain excessive, can be because of two point cloud model initial positions
Deviation is crossed conference and is caused erroneous judgement.Owing to just registration two point cloud model initial positions are unknown, prevent from registration factor value be improper causing
Registration lost efficacy, and can add during iterated transform later based on completing the cloud data after converting for the first timeCarry out dynamically
Regulation.Although the characteristic point of primary election and corresponding point thereof exist bigger error, but during iteration, this error can gradually contract
Little.
Embodiment two: in first registration process, Hood model only needs 5 iteration can restrain, for relative complex
Bunny model, wanting of registration convergence is slow, takes around 10 times and could restrain, as shown in Figure 9.By calculating registration efficiency factor
Judge the effectiveness of just registration, for Hood model,, for Bunny model,, the first of two models is joined
Accurate efficiency factor is no more than 0.5, and therefore just registration has reached Expected Results.According toValue size can be seen that, for letter relatively
The relative error that Hood model for list, Geomagic Studio and the method for the invention calculate is than Bunny model
Greatly, therefore registration at the beginning of complex model can be obtained more more preferable effect than Geomagic Studio by the method for registering of the present invention.Due to
The characteristic point that just registration process selects is insufficient so that it is precision cannot meet the follow-up studies such as cloud data normal estimation, reconstruction
Requirement, for improving further the precision of just registration, by calculating the public territory self adaptation of two point cloud models after first registration
Increase feature is counted, and realization is smart to be registrated.
Embodiment three: on the impact of error precision and then determine two model essence registrations for comparing different characteristic point number
Two models after just registrating in Fig. 7 are carried out essence registration test under different characteristic point number by best features point number.From figure
10 it can be seen that when feature point number is 20 when, the precision for registration that increases of feature point number affects very little,
Registration process about reaches convergence at about 10 times.But the further increase that feature is counted can cause number of times that Poisson rebuilds,
The time of inquiry corresponding point increases, and in order to take into account efficiency and the precision of registration, the smart registration features point number of the present invention takes 20.
Embodiment four: use method for registering of the present invention, ICP method, improve ICP method in Fig. 8 just Hood after registration,
Bunny model carries out essence registration contrast test, not Tongfang after iterations after statistics essence registration convergence and every time iteration
Registration error value acquired by method is as shown in figure 11.Compared to a Point matching criterion, the point of employing is permissible to the matching criterior in face
Convergence faster, as can be seen from Figure 11, it is the slowest that original I CP method restrains, and the error after convergence is maximum, and the present invention is registered in
Can restrain for about 10 times.Although improving ICP method can converge to less error, but the two differing in registration accuracy
And little, but improve ICP method and need iteration just can restrain for 20 times, therefore the method for the invention is improving registration accuracy
Under premise, can more quickly restrain.
Embodiment five: use the method for the invention to use 20 characteristic points completely to registrate on Hood, Bunny model
Process, Hood model totally 3 views, with the 1st view for fixing cloud data, two other is joined for floating cloud data
Standard, complete registration effect (wherein a-c is the different views cloud data at 3 visual angles of Hood) as shown in figure 12.
Embodiment six: Bunny model all has public intersecting area due to 2-5 view and the 1st view, and the 6th regards
Figure does not has public intersecting area, so taking following registration strategies to obtain complete Bunny point cloud model: with the 1st with the 1st
View is fixing cloud data, and the 2nd, 3,4,5 views are floating point cloud model, carries out with the 1st view successively just registrating, essence
Registration, obtains new point cloud model, then useRegistrate with at the beginning of the 6th view is carried out, essence registration finally gives complete cloud data,
Effect (wherein a-f is the different views cloud data at 6 visual angles of Bunny) as shown in figure 13.
The above, be only the preferred embodiments of the present invention, is not the restriction that the present invention makees other forms, any
Those skilled in the art are changed possibly also with the technology contents of the disclosure above or are modified as the equivalence changed on an equal basis
Embodiment.But every without departing from technical solution of the present invention content, according to the technical spirit of the present invention, above example is made
Any simple modification, equivalent variations and remodeling, still fall within the protection content of technical solution of the present invention.
Claims (7)
1. a some cloud Rigid Registration method based on local Poisson curve reestablishing, it is characterised in that step is followed successively by:, sets and treats
Surface in kind sampling point set under two different visual angles of registrationWith, willAs the some cloud that floats,As fixing point cloud, logical
Cross the mode of man-machine interactively respectively fromWithIn choose subsetWith;Two, setFor empty set,Choose portion the most alternately
Divide characteristic point, and selected characteristic point is sequentially added into;Three, setFor empty set, forInterior each sampling point,Interior friendship
Choose matched sampling point mutually, and institute's sampling point is sequentially added into;Four, based onInterior sampling point existsIn neighborhood
Point set, forInterior each sampling point builds Poisson curved surface, concretely comprises the following steps: each sample neighborhood of a point point set is added auxiliary by (1)
Point builds closes point set;(2) closing point set is carried out normal estimation;(3) carry out completely based on the closing point set after normal estimation
Poisson curve reestablishing;(4) the local Poisson curved surface corresponding to neighborhood point set is separated from complete Poisson curved surface, with local
Poisson curved surface is as the Poisson curved surface of sampling point;Five, withThe distance of interior sampling point extremely corresponding Poisson curved surface is as survey
Degree, solves so that the rigid transformation matrix of point set registration function minimization, thus completesWithPreliminary registration, i.e. make
WithIt is fully overlapped under the same coordinate system;Six, willReset to empty set,Interior selected part sampling point is as new feature
Point, selection rule is to ensure that institute's sampling point existsThere is complete neighborhood point set, i.e. this sampling point existInterior neighborhood point set can
It is uniformly distributed in the surrounding adjacent regions of this sampling point, institute's sampling point is added;Seven, willReset to empty set, forIn
Each sampling point,Inside choose the most nearest sampling point, and institute's sampling point is sequentially added into;Eight, side described in applying step four
Method isInterior each sampling point builds Poisson curved surface;Nine, withInterior sampling point is made to the distance of corresponding Poisson curved surface
For estimating, solve so that the rigid transformation matrix of point set registration function minimization, thus completeWithAccuracy registration, result
After output rigid transformationWith registration error.
Point cloud Rigid Registration method based on local Poisson curve reestablishing the most according to claim 1, it is characterised in that:
In step 4, to sample neighborhood of a point point set add auxiliary magnet structure close point set, concretely comprise the following steps: (1) set sampling point as ,Neighborhood point set be, willProjection, to plane, is designated as;(2) identified by boundary characteristic recognizerBoundary point set;(3) according to the corresponding relation of projection, obtainBoundary point set;(4) based on,Between corresponding boundary point, insert discrete point, closing point set can be built.
Point cloud Rigid Registration method based on local Poisson curve reestablishing the most according to claim 1, it is characterised in that:
In step 4, by building dynamically spatial-data index, the Poisson curved surface corresponding to neighborhood point set is separated, concretely comprises the following steps:
(1) set constructed complete Guan Bi Poisson curved surface as, set up based on Poisson curved surfaceSpatial index KD tree and Half-edge Structure
The index structure combined;(2) arest neighbors in the index of the sampling point in inquiry neighborhood point set;(3) grid after separation is set bent
Face is, topology information based on Half-edge Structure, an annulus and the dough sheet information thereof of arest neighbors are stored in;(4) output local
Poisson curved surface, it is ensured thatKeep the topological integrity of original data point surface information.
Point cloud Rigid Registration method based on local Poisson curve reestablishing the most according to claim 1, it is characterised in that:
In step 5, withIn sampling point to the minimum distance of Poisson curved surface as estimating, concretely comprise the following steps: (1) is based on local Poisson
Curved surfaceSet up grid index structure KD tree;(2) right, searched by the K-NN search algorithm of KD treeExtremely's
Grid vertex recently;(3) topological neighborhood information based on index leaf node storage obtainsAn annulus tri patch;(4) profit
Calculate with point-to-plane distance formulaClosest approach to annulus tri patch, willAsCorresponding point;(5) calculateWith's
Distance, using this distance as measure value.
Point cloud Rigid Registration method based on local Poisson curve reestablishing the most according to claim 1, it is characterised in that:
In step 6,Interior selected part sampling point is as new characteristic point, concretely comprising the following steps of selection rule: (1) calculates point set
Span;(2) based onNeighborhood point set, calculateSpan;(2) setFor Selection parameter, as
Really, thenThere is the defects such as hole, rejectOtherwise, willAs characteristic point.
Point cloud Rigid Registration method based on local Poisson curve reestablishing the most according to claim 5, it is characterised in that:
Step (1) calculates point setSpan, concretely comprise the following steps: a) from surface in kind sampling point setIn obtain at randomIndividual sample
Point, forms point set;B) setFor empty set, forIn each sampling point, calculate it and arriveIt is nearest for middle-rangeIndividual sampling point
Distance average is also added into setIn;C) willThe average of middle all elements is as span。
Point cloud Rigid Registration method based on local Poisson curve reestablishing the most according to claim 1, it is characterised in that
In step 9, minimum range based on sampling point Poisson curved surface is used to carry out error estimation, withTo local Poisson curved surfaceMinimum range as error, concretely comprise the following steps: (1),, wherein;(2) calculateArrive's
Closest approach;(3) calculateEuclidean distance;(4);(5), repeat step (2)-(4) until;(6);(8) return error。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610456738.8A CN106023298B (en) | 2016-06-22 | 2016-06-22 | Point cloud Rigid Registration method based on local Poisson curve reestablishing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610456738.8A CN106023298B (en) | 2016-06-22 | 2016-06-22 | Point cloud Rigid Registration method based on local Poisson curve reestablishing |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106023298A true CN106023298A (en) | 2016-10-12 |
CN106023298B CN106023298B (en) | 2018-06-19 |
Family
ID=57086364
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610456738.8A Expired - Fee Related CN106023298B (en) | 2016-06-22 | 2016-06-22 | Point cloud Rigid Registration method based on local Poisson curve reestablishing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106023298B (en) |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107610228A (en) * | 2017-07-05 | 2018-01-19 | 山东理工大学 | Curved surface increment topology rebuilding method based on massive point cloud |
CN108151712A (en) * | 2017-11-13 | 2018-06-12 | 合肥阿巴赛信息科技有限公司 | A kind of human body three-dimensional modeling and measuring method and system |
CN108230377A (en) * | 2017-12-19 | 2018-06-29 | 武汉国安智能装备有限公司 | The approximating method and system of point cloud data |
CN108520550A (en) * | 2018-04-10 | 2018-09-11 | 南京信息工程大学 | Point cloud data Poisson curve reestablishing method based on noise classification and MLS |
CN109191583A (en) * | 2018-08-10 | 2019-01-11 | 广东工业大学 | One kind being based on anisotropy MLS curved surface Accurate align method |
CN109685841A (en) * | 2019-01-03 | 2019-04-26 | 上海狮迈科技有限公司 | Threedimensional model and the method for registering and system for putting cloud |
CN109887012A (en) * | 2019-01-09 | 2019-06-14 | 天津大学 | A kind of point cloud registration method of combining adaptive search point set |
WO2019161558A1 (en) * | 2018-02-26 | 2019-08-29 | Intel Corporation | Method and system of point cloud registration for image processing |
CN111986308A (en) * | 2020-07-14 | 2020-11-24 | 武汉数字化设计与制造创新中心有限公司 | Method for identifying error salient points of double-constrained surface based on point cloud normal and curvature change |
CN112506136A (en) * | 2020-12-10 | 2021-03-16 | 北方工业大学 | Positioning point set selection method based on statistical analysis of batch blade curved surface measurement data |
CN113223067A (en) * | 2021-05-08 | 2021-08-06 | 东莞市三姆森光电科技有限公司 | Online real-time registration method for three-dimensional scanning point cloud with plane reference and incomplete |
CN113297340A (en) * | 2020-02-23 | 2021-08-24 | 北京初速度科技有限公司 | Method and device for vectorization of point cloud map and conversion of vector map into point cloud map |
CN113327275A (en) * | 2021-06-18 | 2021-08-31 | 哈尔滨工业大学 | Point cloud double-view-angle fine registration method based on multi-constraint point to local curved surface projection |
CN113344992A (en) * | 2021-05-31 | 2021-09-03 | 山东大学 | Global point cloud registration method, system, storage medium and equipment |
CN114882087A (en) * | 2022-07-12 | 2022-08-09 | 武汉瀚迈科技有限公司 | Real-time registration method for three-dimensional scanning point cloud with incomplete basic primitive |
CN117011128A (en) * | 2023-08-02 | 2023-11-07 | 广州知元科技有限责任公司 | Automatic three-dimensional model registration method and equipment based on region decomposition |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8610706B2 (en) * | 2008-10-04 | 2013-12-17 | Microsoft Corporation | Parallel surface reconstruction |
CN103489218A (en) * | 2013-09-17 | 2014-01-01 | 中国科学院深圳先进技术研究院 | Point cloud data quality automatic optimizing method and system |
CN104299260A (en) * | 2014-09-10 | 2015-01-21 | 西南交通大学 | Contact network three-dimensional reconstruction method based on SIFT and LBP point cloud registration |
-
2016
- 2016-06-22 CN CN201610456738.8A patent/CN106023298B/en not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8610706B2 (en) * | 2008-10-04 | 2013-12-17 | Microsoft Corporation | Parallel surface reconstruction |
CN103489218A (en) * | 2013-09-17 | 2014-01-01 | 中国科学院深圳先进技术研究院 | Point cloud data quality automatic optimizing method and system |
CN104299260A (en) * | 2014-09-10 | 2015-01-21 | 西南交通大学 | Contact network three-dimensional reconstruction method based on SIFT and LBP point cloud registration |
Non-Patent Citations (1)
Title |
---|
贺永兴: "《基于邻域特征的点云配准算法研究》", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (27)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107610228A (en) * | 2017-07-05 | 2018-01-19 | 山东理工大学 | Curved surface increment topology rebuilding method based on massive point cloud |
CN108151712B (en) * | 2017-11-13 | 2020-08-07 | 合肥阿巴赛信息科技有限公司 | Human body three-dimensional modeling and measuring method and system |
CN108151712A (en) * | 2017-11-13 | 2018-06-12 | 合肥阿巴赛信息科技有限公司 | A kind of human body three-dimensional modeling and measuring method and system |
CN108230377B (en) * | 2017-12-19 | 2020-08-21 | 武汉国安智能装备有限公司 | Point cloud data fitting method and system |
CN108230377A (en) * | 2017-12-19 | 2018-06-29 | 武汉国安智能装备有限公司 | The approximating method and system of point cloud data |
WO2019161558A1 (en) * | 2018-02-26 | 2019-08-29 | Intel Corporation | Method and system of point cloud registration for image processing |
US11423508B2 (en) | 2018-02-26 | 2022-08-23 | Intel Corporation | Method and system of point cloud registration for image processing |
CN108520550A (en) * | 2018-04-10 | 2018-09-11 | 南京信息工程大学 | Point cloud data Poisson curve reestablishing method based on noise classification and MLS |
CN109191583A (en) * | 2018-08-10 | 2019-01-11 | 广东工业大学 | One kind being based on anisotropy MLS curved surface Accurate align method |
CN109191583B (en) * | 2018-08-10 | 2022-09-23 | 广东工业大学 | Curved surface accurate alignment method based on anisotropic MLS |
CN109685841B (en) * | 2019-01-03 | 2020-09-18 | 上海狮迈科技有限公司 | Registration method and system of three-dimensional model and point cloud |
CN109685841A (en) * | 2019-01-03 | 2019-04-26 | 上海狮迈科技有限公司 | Threedimensional model and the method for registering and system for putting cloud |
CN109887012A (en) * | 2019-01-09 | 2019-06-14 | 天津大学 | A kind of point cloud registration method of combining adaptive search point set |
CN109887012B (en) * | 2019-01-09 | 2023-03-31 | 天津大学 | Point cloud registration method combined with self-adaptive search point set |
CN113297340B (en) * | 2020-02-23 | 2023-12-19 | 北京初速度科技有限公司 | Vectorization method and device for point cloud map and method and device for converting vector map into point cloud map |
CN113297340A (en) * | 2020-02-23 | 2021-08-24 | 北京初速度科技有限公司 | Method and device for vectorization of point cloud map and conversion of vector map into point cloud map |
CN111986308B (en) * | 2020-07-14 | 2023-09-01 | 武汉数字化设计与制造创新中心有限公司 | Point cloud normal and curvature change based double-constraint surface error salient point identification method |
CN111986308A (en) * | 2020-07-14 | 2020-11-24 | 武汉数字化设计与制造创新中心有限公司 | Method for identifying error salient points of double-constrained surface based on point cloud normal and curvature change |
CN112506136A (en) * | 2020-12-10 | 2021-03-16 | 北方工业大学 | Positioning point set selection method based on statistical analysis of batch blade curved surface measurement data |
CN113223067B (en) * | 2021-05-08 | 2021-11-19 | 广东三姆森科技股份有限公司 | Online registration method for incomplete three-dimensional scanning point cloud with plane reference |
CN113223067A (en) * | 2021-05-08 | 2021-08-06 | 东莞市三姆森光电科技有限公司 | Online real-time registration method for three-dimensional scanning point cloud with plane reference and incomplete |
CN113344992B (en) * | 2021-05-31 | 2022-06-28 | 山东大学 | Global point cloud registration method, system, storage medium and equipment |
CN113344992A (en) * | 2021-05-31 | 2021-09-03 | 山东大学 | Global point cloud registration method, system, storage medium and equipment |
CN113327275A (en) * | 2021-06-18 | 2021-08-31 | 哈尔滨工业大学 | Point cloud double-view-angle fine registration method based on multi-constraint point to local curved surface projection |
CN114882087A (en) * | 2022-07-12 | 2022-08-09 | 武汉瀚迈科技有限公司 | Real-time registration method for three-dimensional scanning point cloud with incomplete basic primitive |
CN117011128A (en) * | 2023-08-02 | 2023-11-07 | 广州知元科技有限责任公司 | Automatic three-dimensional model registration method and equipment based on region decomposition |
CN117011128B (en) * | 2023-08-02 | 2024-02-13 | 广州知元科技有限责任公司 | Automatic three-dimensional model registration method and equipment based on region decomposition |
Also Published As
Publication number | Publication date |
---|---|
CN106023298B (en) | 2018-06-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106023298A (en) | Point cloud rigid registration method based on local Poisson curved surface reconstruction | |
CN106600622B (en) | A kind of point cloud data segmentation method based on super voxel | |
CN104867126B (en) | Based on point to constraint and the diameter radar image method for registering for changing region of network of triangle | |
CN107093205B (en) | A kind of three-dimensional space building window detection method for reconstructing based on unmanned plane image | |
CN106485690A (en) | Cloud data based on a feature and the autoregistration fusion method of optical image | |
CN110443836A (en) | A kind of point cloud data autoegistration method and device based on plane characteristic | |
CN105139379B (en) | Based on the progressive extracting method of classified and layered airborne Lidar points cloud building top surface | |
CN110222642A (en) | A kind of planar architectural component point cloud contour extraction method based on global figure cluster | |
CN107346550B (en) | It is a kind of for the three dimensional point cloud rapid registering method with colouring information | |
Deng et al. | Noisy depth maps fusion for multiview stereo via matrix completion | |
CN103295239A (en) | Laser-point cloud data automatic registration method based on plane base images | |
Takahashi et al. | Expandable YOLO: 3D object detection from RGB-D images | |
CN106780551B (en) | A kind of Three-Dimensional Moving Targets detection method and system | |
CN105740798A (en) | Structure analysis based identification method for object in point cloud scene | |
CN111145228A (en) | Heterogeneous image registration method based on local contour point and shape feature fusion | |
CN109523595A (en) | A kind of architectural engineering straight line corner angle spacing vision measuring method | |
CN109815847A (en) | A kind of vision SLAM method based on semantic constraint | |
Wei et al. | Automatic coarse registration of point clouds using plane contour shape descriptor and topological graph voting | |
CN107610219A (en) | The thick densification method of Pixel-level point cloud that geometry clue perceives in a kind of three-dimensional scenic reconstruct | |
Zhang et al. | Robust procedural model fitting with a new geometric similarity estimator | |
CN108595373A (en) | It is a kind of without control DEM method for registering | |
CN115588178A (en) | Method for automatically extracting high-precision map elements | |
CN115063615A (en) | Repeated texture image matching method based on Delaunay triangulation | |
Ma et al. | Complex buildings orientation recognition and description based on vector reconstruction | |
He et al. | Minimum spanning tree based stereo matching using image edge and brightness information |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180619 Termination date: 20190622 |