CN107590825A - A kind of dot cloud hole method for repairing and mending based on SFM - Google Patents
A kind of dot cloud hole method for repairing and mending based on SFM Download PDFInfo
- Publication number
- CN107590825A CN107590825A CN201710637964.0A CN201710637964A CN107590825A CN 107590825 A CN107590825 A CN 107590825A CN 201710637964 A CN201710637964 A CN 201710637964A CN 107590825 A CN107590825 A CN 107590825A
- Authority
- CN
- China
- Prior art keywords
- point
- msub
- hole
- cloud
- mrow
- 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
Landscapes
- Image Processing (AREA)
- Processing Or Creating Images (AREA)
Abstract
The invention discloses a kind of dot cloud hole method for repairing and mending based on SFM, this method realizes that step is：(1) after obtaining cloud data respectively using grating project and SFM methods, the point cloud two-dimensional phase information obtained in grating project is recycled, extracts the boundary point of three-dimensional point cloud hole；(2) the cloud data collection obtained to the SFM cloud data collection obtained and grating project carries out registration；(3) auxiliary point that optical grating projection obtains point cloud Hole region is extracted on the data set that SFM is obtained；(4) auxiliary point is based on, further hole repair is realized using RBF.Inventive algorithm robust, repair efficiency accuracy is high, and can more recover the detailed information of object.
Description
Technical field：
The present invention relates to a kind of dot cloud hole method for repairing and mending based on SFM, belongs to a cloud post-processing technology field.
Background technology：
During three-dimensional measurement, due to object complex-shaped or metering system itself the defects of, three-dimensional point cloud
Data are usually present hole.Optical grating projection mensuration has the features such as speed is fast, and automaticity is high, but due to being needed during measurement
Project fill-in light, gained cloud data can due to object blocking and produced with the reason such as the reflectivity on testee surface in itself
Raw perforated.The presence of hole has a strong impact on the follow-up operation of a cloud.Therefore, the patch work of hole is indispensable.
The difficult point of hole repair is：1. how accurately to extract hole border；2. how to be obtained in perforated more
Information so that repairing point be more bonded original-shape.The extraction on hole border is the first job of hole repair, it
Accuracy directly affects the result of hole repair.Existing dot cloud hole border points extraction method is mainly entered on the basis of three-dimensional
OK, it is broadly divided into the method that boundary point is extracted based on an extracting method of cloud grid and directly on a cloud.The former calculates complexity,
The internal point that not uniform enough point cloud will easily be sampled also regards boundary point；The latter usually requires to calculate the relation of point and neighborhood point,
It is and easily affected by noise.By actual measurement it can be found that point cloud boundary point feature in two-dimensional phase bitmap is obvious, and utilizes
Two-dimensional phase needs to establish grid or calculating neighborhood information to extract hole border and eliminate in three-dimensional method, significantly improves
The efficiency of border points extraction.Therefore extraction hole border is extracted using a cloud two-dimensional phase information herein.
Existing hole repair algorithm is roughly divided into two classes:One kind is directly to utilize the neighborhood information on dot cloud hole periphery,
Continuity or curvature carry out hole repair, such as using SPL, enter row interpolation the methods of neutral net, so as to repair
Hole.Such algorithm is simple and easy, it is not necessary to any ground side information.It is but larger for Curvature varying or shape is more multiple
Miscellaneous dot cloud hole, because loss of learning is excessive, repair efficiency is unsatisfactory.It is another kind of mainly to be tied again by obtaining additional information
The structure for closing dot cloud hole periphery is repaired.For example Yann Quinsat etc. are proposed using priori CAD model and hole area
The continuity of domain neighborhood carries out hole repair as constraints.M Panchetti etc. propose the two dimension using corresponding picture
Curvature information between information and point cloud grid repairs hole.This kind of method is good for the perforated repair efficiency of complexity
It is good, but it is registering between the whether foundation of priori CAD model, or two-dimensional pixel and three-dimensional point cloud, it can all cause to calculate
Method overall complexity is higher.
The content of the invention
Technical problem：Less for hole repair process Hole area information, complicated hole repair effect is not enough managed
The problem of thinking, the technology of point cloud information, the present invention are obtained with reference to fusion SFM more ripe in three-dimensional reconstruction and laser scanning
Provide a kind of dot cloud hole method for repairing and mending based on SFM.The present invention is by obtaining SFM cloud data come complementary apertures cavity portion
Information, instruct hole repair, these side informations can effectively help to recover the information of dot cloud hole, improve hole repair
As a result accuracy, and recover more information.The boundary point of the two-dimensional phase information extraction point cloud of point cloud is utilized again, is reduced
The complexity of border point location.
Technical scheme：To achieve the above object, the technical solution adopted by the present invention is：
A kind of dot cloud hole method for repairing and mending based on SFM, comprises the following steps：
Step 1：Dot cloud hole border obtains：Extracted a little using the point cloud two-dimensional phase information obtained in optical grating projection
Cloud hole border；The boundary point of a cloud is extracted first；Then the external boundary point of point cloud is removed；
Step 2：Registration is carried out to two amplitude point clouds：Slightly matched somebody with somebody first with the RANSAC algorithms based on coplanar four point set
Standard, then carry out accuracy registration using improved ICP algorithm；
Step 3：Extract auxiliary point：After registration is completed, using the dot cloud hole border obtained in step 1, SFM is extracted
The point cloud P that method obtainssIn the point cloud P that grating project obtainswThe auxiliary point IP in Hole region；
Step 4：Dot cloud hole region is filled up using RBF：Hole boundary point is obtained to equatorial projection first
Point cloud boundary characteristics polygon is taken, enters row interpolation followed by the feature of boundary characteristic polygon, obtains initial repairing point set FP
={ fpi, i=1,2,3 ... n }, initial repairing point set is then adjusted according to RBF so as to obtain final repairing point set；
Preferably：In the step 1 using two-dimensional phase information to extract dot cloud hole border the step of it is as follows：
Step 1.1：Obtain the two-dimentional boundary point of point cloud：The size of two-dimensional phase bitmap is M*N, and p [m] [n] is mutually phase diagram
On certain point, the collection of the boundary point for the cloud that sets up an office is combined into B, then when p [m] [n] meets formula, p [m] [n] ∈ B；
WhereinIt is the threshold value of phase hit；
Step 1.2：The outer boundary point of extraction point cloud：On two-dimensional phase image, the feature of exterior contour point is every in B
The initial and final non-zero points of a line；If the collection of the point on outer wheels profile is combined into S, when p [m] [n] meets formula, p [m]
[n]∈S；
Wherein m=1,2 ... M, so the boundary point Hp={ p [m] [n] | p [m] [n] ∈ B-S } of hole；
Step 1.3：Calculate the three-dimensional coordinate of dot cloud hole boundary point：Coordinate [the X of known cloud datac Yc Zc] with phase
Corresponding points p [m] [n]=θ on bitmap has following corresponding relation,
Wherein, θ is phase value；a1,a2,a3,a4,a5,a6,a7,a8For system parameter；ρ is scale factor；AcImaged for 3*3
Parameter matrix inside machine；And a1,a2,a3,a4,a5,a6,a7,a8AcIt can be obtained by the demarcation of system；Therefore hole is substituted into
The two-dimensional phase coordinate of boundary point, corresponding hole border three-dimensional coordinate point H=can be calculated by two above formula
{hi, i=1,2,3...n }；
Preferably：The step of the step 2 point cloud registration is as follows：
Step 2.1：Initial registration：In initial registration, in order to increase registering efficiency, first in PwAnd PsIn be respectively divided
The obvious region of feature corresponding to going out；In PwIn characteristic area, arbitrarily not conllinear 4 points are randomly selected using RANSAC methods,
And in PsIn find corresponding point, corresponding transformation matrix M is calculated；After r iteration, choose and cause two panels point cloud
The minimum transformation matrix M of matching error.Using M, initial matching result is obtained；
Step 2.2：Accuracy registration：In accuracy registration, firstly for PwIn every bit p, in PsIn search out Euclidean away from
From four nearest points, and the center of gravity for the minimum neighborhood that this four points are formed is obtained, form point pair；Then least square is utilized
Method calculating parameter R, t；Final result is applied to PsIn, obtain the point after registration and converge Ps'；
Preferably：Screening auxiliary point IP method is in the step 3：For the point cloud P after registration in the step 2s'
In every bit, calculate it whether the hole boundary point H={ h extracted in the step 1i, i=1,2,3...n } bounding box
In, if then thinkingOtherwise it is not；The point for adding auxiliary point converges and is combined into NP={ Pω∪IP}；
Preferably：The step of filling up perforated using RBF in the step 4 be：
Step 4.1：Extract hole characteristic polygon：The dot cloud hole point H=for being extracted step 1 with least square method
{hi, i=1,2,3...n } two dimensional surface acquisition projection matrix PM is projected to, form Porous Characteristic polygon；
Step 4.2：Obtain thick reparation point：The length of side of hole characteristic polygon is obtained, takes the average d of the length of side to exist as step-length
Row interpolation is entered so as to obtain the point of the repairing on two dimensional surface in perforated；Interpolation point is mapped using the PM obtained in step 4.2
Three-dimensional planar is returned, obtains thick reparation point FP={ fpi, i=1,2,3 ... n }；
Step 4.3：Repair hole：Conjunction NP is converged to the point that information supplement point is with the addition of in step 3 first and establishes kd trees, so
Afterwards to arbitrary point fp in FPiK neighborhood point is chosen, space curved surface is established using RBF functions, finally according to gradient condition
So that fpiAdjusted to curved surface to meet a cloud surface curvature requirement；Each of which point fpiCorresponding space curved surface equation is
F (x, y, z)=p0+p1x+p2y+p3z+β
Wherein β is error term, and
The unknown number ω of space curved surface equationj,p0,p1,p2,p3It can be obtained by solution formula QA=B；
Wherein A=[ω1 ω2 … ωn p0 p1 p2 p3]TB=[δ1 δ2 … δn 0 0 0 0]T,δ in experimentiFor error amount, δ1=δ2=...=δn=0.1, N={ Nix, Niy,
Niz, i=1,2 ..., 6 }, Dij=| Ni-Nj|3/2I=1,2 ..., 6, j=1,2 ... 6,
δ in experimentiFor error amount, δ1=δ2=...=δn=0.1, N={ Nix, Niy, Niz, i=1,2..., 6 }, Dij=|
Ni-Nj|3/2I=1,2 ..., 6, j=1,2 ... 6.
Beneficial effect：A kind of hole repair algorithm based on SFM provided by the invention, compared with prior art, have following
Beneficial effect：
In addition to obtaining information from hole border, the method that information is obtained in the new point cloud obtained from SFM is added,
Improve the precision of hole repairing；The method for finding hole border in three dimensions make it that algorithm is complicated, and the present invention uses again
Using the method on two-dimensional phase information searching hole border, the efficiency of algorithm is improved.
Brief description of the drawings
Fig. 1 is entire flow of the present invention.
Fig. 2 is the point cloud model figure of smooth surface ceramic cup.Reflection in smooth surface ceramic cup due to object in itself as seen from the figure
Rate is too high, and hole is generated in middle position.
Fig. 3 is the point cloud model figure of face.In faceform, eyes and brow region generate because reflectivity is relatively low
Hole, in nasal area because blocking for itself also generates hole.
Fig. 4 represents the smooth surface ceramics cyphelia cloud schematic diagram after hole repair.Point cloud chart after repairing can be more
It is bonded initial data.
Fig. 5 is the detail view that the perforated of smooth surface ceramic cup is repaired, wherein newly-increased hole repair point totally 404.SFM
The cloud data of acquisition provides 96 auxiliary points.These data provide when RBF functions calculate the surface equation of corresponding points
Crucial information so that hole repair result is more accurate.
Fig. 6 is the point cloud chart of the face after hole repair.Result is repaired to react exactly after three-dimensional reconstruction
The shape of object in itself, and recovered the details of more horn of plenty.
What Fig. 7 specifically illustrated nasal portion fills up details.For increasingly complex model such as faceform, based on SFM
Dot cloud hole method for repairing and mending can also effectively recover hole information.
Embodiment
Below in conjunction with the accompanying drawings and specific embodiment, the present invention is furture elucidated, it should be understood that these examples are merely to illustrate this
Invention rather than limitation the scope of the present invention, after the present invention has been read, those skilled in the art are various to the present invention's
The modification of the equivalent form of value falls within the application appended claims limited range.
With reference to the accompanying drawings, more detailed description is done to specific embodiments of the present invention.Programming realization tools selection
Visual Studio2013 are tested respectively to ceramic cup and face.
The specific implementation step of the present invention is as follows：
Step 1：Dot cloud hole border obtains
Step 1.1：Obtain the two-dimentional boundary point of point cloud
The size of two-dimensional phase bitmap is M*N, and p [m] [n] is the certain point on phase diagram, the set of the boundary point for the cloud that sets up an office
For B, then when p [m] [n] meets formula, p [m] [n] ∈ B；
WhereinIt is the threshold value of phase hit；
Step 1.2：The outer boundary point of extraction point cloud
On two-dimensional phase image, the feature of exterior contour point is the initial and final non-zero points in B per a line；It is if outside
The collection of point on contour line is combined into S, when p [m] [n] meets formula, p [m] [n] ∈ S；
Wherein m=1,2 ... M, so the boundary point Hp={ p [m] [n] | p [m] [n] ∈ B-S } of hole；
Step 1.3：Calculate the three-dimensional coordinate of dot cloud hole boundary point
Coordinate [the X of known cloud datac Yc Zc] with corresponding points p [m] [n]=θ on phase diagram there is following corresponding pass
System,
Wherein, θ is phase value；a1,a2,a3,a4,a5,a6,a7,a8For system parameter；ρ is scale factor；AcImaged for 3*3
Parameter matrix inside machine；And a1,a2,a3,a4,a5,a6,a7,a8AcIt can be obtained by the demarcation of system；Therefore hole is substituted into
The two-dimensional phase coordinate of boundary point, corresponding hole border three-dimensional coordinate point H=can be calculated by two above formula
{hi, i=1,2,3...n }；
Step 2：Registration is carried out to two amplitude point clouds
Step 2.1：Initial registration
In initial registration, in order to increase registering efficiency, first in PwAnd PsIn be respectively divided out corresponding to feature it is obvious
Region；In PwIn characteristic area, arbitrarily not conllinear 4 points are randomly selected using RANSAC methods, and in PsIn find corresponding to
Point, corresponding transformation matrix M is calculated；After r iteration, the conversion for make it that two panels point cloud matching error is minimum is chosen
Matrix M；Using M, initial matching result is obtained；
Step 2.2：Accuracy registration
In accuracy registration, firstly for PwIn every bit p, in PsIn search out four nearest points of Euclidean distance, and
The center of gravity for the minimum neighborhood that this four points are formed is obtained, forms point pair；Then least square method calculating parameter R, t are utilized；Will
Final result is applied to PsIn, obtain the point after registration and converge Ps'；
Step 3：Extract auxiliary point
For Ps' in every bit, calculate the hole boundary point H={ h whether it extracts in step 1i, i=1,2,
3...n } in bounding box, if then thinkingOtherwise it is not；The point for adding auxiliary point converges and is combined into NP={ Pω∪
IP}；
Step 4：Dot cloud hole region is filled up using RBF
Step 4.1：Extract hole characteristic polygon
With least square method by the dot cloud hole extracted in step 1 point H={ hi, i=1,2,3...n } project to two
Dimensional plane obtains projection matrix PM, forms Porous Characteristic polygon；
Step 4.2：Obtain thick reparation point
Obtain hole characteristic polygon the length of side, take the length of side average d enter as step-length in perforated row interpolation so that
Obtain the repairing point on two dimensional surface.Interpolation point is mapped back into three-dimensional planar using the PM obtained in step 4.1, obtains thick repair
Point FP={ fpi, i=1,2,3 ... n }；
Step 4.3：Repair hole
Conjunction NP is converged to the point that information supplement point is with the addition of in step 3 first and establishes kd trees, then in FP arbitrary one
Individual point fpiK neighborhood point is chosen, space curved surface is established using RBF functions, fp is finally caused according to gradient conditioniAdjusted to curved surface
To meet point cloud surface curvature requirement；Each of which point fpiCorresponding space curved surface equation is
F (x, y, z)=p0+p1x+p2y+p3z+β
Wherein β is error term, and
The unknown number ω of space curved surface equationj,p0,p1,p2,p3It can be obtained by solution formula QA=B；
Wherein A=[ω1 ω2 … ωn p0 p1 p2 p3]TB=[δ1 δ2 … δn 0 0 0 0]T,
δ in experimentiFor error amount, δ1=δ2=...=δn=0.1, N=
{Nix, Niy, Niz, i=1,2..., 6 }, Dij=| Ni-Nj|3/2I=1,2 ..., 6, j=1,2 ... 6.
The data set that the present invention obtains SFM is added in hole repair, and using data set addition side information, is instructed
Hole repair so that the perforated after repairing can obtain the shape of more detailed information, preferably recovery hole cavity portion.
Two-dimensional phase information is added in dot cloud hole extraction again, using feature of the three-dimensional hole boundary point in two-dimensional phase, carried
Take out hole border；Reduce the complexity of hole Boundary Extraction.
Described above is only the preferred embodiment of the present invention, it should be pointed out that：For the ordinary skill people of the art
For member, under the premise without departing from the principles of the invention, some improvements and modifications can also be made, these improvements and modifications also should
It is considered as protection scope of the present invention.
Claims (5)
1. a kind of dot cloud hole method for repairing and mending based on SFM, it is characterised in that comprise the following steps：
Step 1：Dot cloud hole border obtains：A cloud hole is extracted using the point cloud two-dimensional phase information obtained in optical grating projection
Hole border.The inner and outer boundary point of a cloud is extracted first；Second step removes the external boundary point of point cloud；
Step 2：Registration is carried out to two amplitude point clouds：Rough registration is carried out first with the RANSAC algorithms based on coplanar four point set, so
Afterwards accuracy registration is carried out using improved ICP algorithm；
Step 3：Extract auxiliary point：After registration is completed, using the dot cloud hole border obtained in step 1, extract SFM methods and obtain
The point cloud P takensIn the point cloud P that grating project obtainswThe auxiliary point IP in Hole region；
Step 4：Dot cloud hole region is filled up using RBF：Hole boundary point is obtained into point to equatorial projection first
Cloud boundary characteristics polygon, then enter row interpolation using the feature of boundary characteristic polygon, obtain initial repairing point set FP=
{fpi, i=1,2,3 ... n }, initial repairing point set is then adjusted according to RBF so as to obtain final repairing point set.
2. the dot cloud hole method for repairing and mending according to claim 1 based on SFM, it is characterised in that in step 1, utilize two dimension
The step of phase information is to extract dot cloud hole border is as follows：
Step 1.1：Obtain the two-dimentional boundary point of point cloud：The size of two-dimensional phase bitmap is M*N, and p [m] [n] is certain on phase diagram
A bit, the collection of the boundary point for the cloud that sets up an office is combined into B, then when p [m] [n] meets formula, p [m] [n] ∈ B,
<mrow>
<mrow>
<mo>|</mo>
<mrow>
<mi>p</mi>
<mo>&lsqb;</mo>
<mi>m</mi>
<mo>&rsqb;</mo>
<mo>&lsqb;</mo>
<mi>n</mi>
<mo>&rsqb;</mo>
<mo>-</mo>
<mi>p</mi>
<mo>&lsqb;</mo>
<mi>m</mi>
<mo>&PlusMinus;</mo>
<mn>1</mn>
<mo>&rsqb;</mo>
<mo>&lsqb;</mo>
<mi>n</mi>
<mo>&PlusMinus;</mo>
<mn>1</mn>
<mo>&rsqb;</mo>
</mrow>
<mo>|</mo>
</mrow>
<mo>></mo>
<mo>&part;</mo>
</mrow>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mi>p</mi>
<mo>&lsqb;</mo>
<mi>m</mi>
<mo>&rsqb;</mo>
<mo>&lsqb;</mo>
<mi>n</mi>
<mo>&rsqb;</mo>
<mo>&NotEqual;</mo>
<mn>0</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mstyle>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mn>8</mn>
</munderover>
</mstyle>
<mi>p</mi>
<mo>&lsqb;</mo>
<mi>m</mi>
<mo>&PlusMinus;</mo>
<mn>1</mn>
<mo>&rsqb;</mo>
<mo>&lsqb;</mo>
<mi>n</mi>
<mo>&PlusMinus;</mo>
<mn>1</mn>
<mo>&rsqb;</mo>
<mo>&NotEqual;</mo>
<mn>0</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>&Pi;</mi>
<mi>p</mi>
<mo>&lsqb;</mo>
<mi>m</mi>
<mo>&PlusMinus;</mo>
<mn>1</mn>
<mo>&rsqb;</mo>
<mo>&lsqb;</mo>
<mi>n</mi>
<mo>&PlusMinus;</mo>
<mn>1</mn>
<mo>&rsqb;</mo>
<mo>=</mo>
<mn>0</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
WhereinIt is the threshold value of phase hit；
Step 1.2：The outer boundary point of extraction point cloud：On two-dimensional phase image, the feature of exterior contour point is per a line in B
Initial and final non-zero points, if the collection of the point on outer wheels profile is combined into S, when p [m] [n] meets formula, p [m] [n] ∈
S,
<mrow>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>0</mn>
</mrow>
<mrow>
<mi>n</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</munderover>
<mi>p</mi>
<mo>&lsqb;</mo>
<mi>m</mi>
<mo>&rsqb;</mo>
<mo>&lsqb;</mo>
<mi>j</mi>
<mo>&rsqb;</mo>
<mo>=</mo>
<mn>0</mn>
<mo>|</mo>
<mo>|</mo>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mi>n</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</munderover>
<mi>p</mi>
<mo>&lsqb;</mo>
<mi>m</mi>
<mo>&rsqb;</mo>
<mo>&lsqb;</mo>
<mi>j</mi>
<mo>&rsqb;</mo>
<mo>=</mo>
<mn>0</mn>
</mrow>
Wherein m=1,2 ... M, so the boundary point Hp={ p [m] [n] | p [m] [n] ∈ B-S } of hole；
Step 1.3：Calculate the three-dimensional coordinate of dot cloud hole boundary point：Coordinate [the X of known cloud datac Yc Zc]TWith phase diagram
On corresponding points p [m] [n]=θ have following corresponding relation,
<mrow>
<mi>&theta;</mi>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>a</mi>
<mn>1</mn>
</msub>
<msub>
<mi>X</mi>
<mi>c</mi>
</msub>
<mo>+</mo>
<msub>
<mi>a</mi>
<mn>2</mn>
</msub>
<msub>
<mi>Y</mi>
<mi>c</mi>
</msub>
<mo>+</mo>
<msub>
<mi>a</mi>
<mn>3</mn>
</msub>
<msub>
<mi>Z</mi>
<mi>c</mi>
</msub>
<mo>+</mo>
<msub>
<mi>a</mi>
<mn>4</mn>
</msub>
</mrow>
<mrow>
<msub>
<mi>a</mi>
<mn>5</mn>
</msub>
<msub>
<mi>X</mi>
<mi>c</mi>
</msub>
<mo>+</mo>
<msub>
<mi>a</mi>
<mn>6</mn>
</msub>
<msub>
<mi>Y</mi>
<mi>c</mi>
</msub>
<mo>+</mo>
<msub>
<mi>a</mi>
<mn>7</mn>
</msub>
<msub>
<mi>Z</mi>
<mi>c</mi>
</msub>
<mo>+</mo>
<msub>
<mi>a</mi>
<mn>8</mn>
</msub>
</mrow>
</mfrac>
</mrow>
<mrow>
<mi>&rho;</mi>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mi>m</mi>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>n</mi>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>=</mo>
<msub>
<mi>A</mi>
<mi>c</mi>
</msub>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<msub>
<mi>X</mi>
<mi>c</mi>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>Y</mi>
<mi>c</mi>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>Z</mi>
<mi>c</mi>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
Wherein, θ is phase value；a1,a2,a3,a4,a5,a6,a7,a8For system parameter；ρ is scale factor；AcFor in 3*3 video cameras
Portion's parameter matrix；And a1,a2,a3,a4,a5,a6,a7,a8AcIt can be obtained by the demarcation of system, therefore substitute into hole border
The two-dimensional phase coordinate of point, corresponding hole border three-dimensional coordinate point H={ h can be calculated by two above formulai, i=
1,2,3...n}。
3. the dot cloud hole method for repairing and mending according to claim 1 based on SFM, it is characterised in that in step 2, described cloud
The step of registration is as follows：
Step 2.1：Initial registration：In initial registration, in order to increase registering efficiency, first in PwAnd PsIn be respectively divided out pair
The obvious region of feature answered；In PwIn characteristic area, randomly select at any not conllinear 4 points using RANSAC methods, and
PsIn find corresponding point, corresponding transformation matrix M is calculated；After r iteration, choose and cause two panels point cloud matching
The minimum transformation matrix M of error；Using M, initial matching result is obtained；
Step 2.2：Accuracy registration：In accuracy registration, firstly for PwIn every bit p, in PsIn search out Euclidean distance most
Four near points, and the center of gravity for the minimum neighborhood that this four points are formed is obtained, form point pair；Then least square method meter is utilized
Calculate parameter R, t；Final result is applied to PsIn, obtain the point after registration and converge Ps'。
4. the dot cloud hole method for repairing and mending based on SFM according to claim 1 or 2 or 3, it is characterised in that sieved in step 3
The method for selecting auxiliary point IP is：For the point cloud P after the registration that is obtained in step 2s' in every bit, whether calculate it in step
The rapid 1 hole boundary point H={ h extractedi, i=1,2,3...n } bounding box in, if then thinkingOtherwise it is not；
The point for adding auxiliary point converges and is combined into NP={ Pω∪IP}。
5. the dot cloud hole method for repairing and mending based on SFM according to claim 1 or 2 or 3, it is characterised in that sharp in step 4
The step of filling up perforated with RBF be：
Step 4.1：Extract hole characteristic polygon：With least square method by the dot cloud hole extracted in step 1 point H=
{hi, i=1,2,3...n } two dimensional surface acquisition projection matrix PM is projected to, form Porous Characteristic polygon；
Step 4.2：Obtain thick reparation point：The length of side of hole characteristic polygon is first obtained, takes the average d of the length of side as step-length in hole
Row interpolation is entered so as to obtain the point of the repairing on two dimensional surface in hole region；Interpolation point is mapped back using the PM obtained in step 4.1
Three-dimensional planar, obtain thick reparation point FP={ fpi, i=1,2,3 ... n }；
Step 4.3：Repair hole：Conjunction NP is converged to the point that information supplement point is with the addition of in step 3 first and establishes kd trees, it is then right
Arbitrary point fp in FPiK neighborhood point is chosen, space curved surface is established using RBF functions, is finally caused according to gradient condition
fpiAdjusted to curved surface to meet a cloud surface curvature requirement；Each of which point fpiCorresponding space curved surface equation is
F (x, y, z)=p0+p1x+p2y+p3z+β
Wherein β is error term, and
The unknown number ω of space curved surface equationj,p0,p1,p2,p3It can be obtained by solution formula QA=B；
Wherein A=[ω1 ω2 … ωn p0 p1 p2 p3]TB=[δ1 δ2 … δn 0 0 0 0]T,δ in experimentiFor error amount, δ1=δ2=...=δn=0.1, N={ Nix, Niy,
Niz, i=1,2 ..., 6 }, Dij=| Ni-Nj|3/2I=1,2 ..., 6, j=1,2 ... 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710637964.0A CN107590825B (en) | 2017-07-31 | 2017-07-31 | Point cloud hole repairing method based on SFM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710637964.0A CN107590825B (en) | 2017-07-31 | 2017-07-31 | Point cloud hole repairing method based on SFM |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107590825A true CN107590825A (en) | 2018-01-16 |
| CN107590825B CN107590825B (en) | 2020-10-02 |
Family
ID=61042042
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710637964.0A Active CN107590825B (en) | 2017-07-31 | 2017-07-31 | Point cloud hole repairing method based on SFM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107590825B (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108399609A (en) * | 2018-03-06 | 2018-08-14 | 北京因时机器人科技有限公司 | A kind of method for repairing and mending of three dimensional point cloud, device and robot |
| CN109147040A (en) * | 2018-08-28 | 2019-01-04 | 浙江大学 | Human body dot cloud hole method for repairing and mending based on template |
| CN109242828A (en) * | 2018-08-13 | 2019-01-18 | 浙江大学 | 3D printing product 3 D defects detection method based on optical grating projection multistep phase shift method |
| CN109859114A (en) * | 2018-12-27 | 2019-06-07 | 北京大学 | Three-dimensional point cloud restorative procedure based on local flatness and non-local similitude |
| CN110349257A (en) * | 2019-07-16 | 2019-10-18 | 四川大学 | A kind of binocular measurement missing point cloud interpolating method based on the mapping of phase puppet |
| CN110415331A (en) * | 2019-06-14 | 2019-11-05 | 中国电建集团昆明勘测设计研究院有限公司 | Method for rapidly detecting, judging and repairing contour or hole based on point cloud data |
| CN112102491A (en) * | 2020-08-12 | 2020-12-18 | 西安交通大学 | Skin damage surface three-dimensional reconstruction method based on surface structured light |
| CN112215763A (en) * | 2020-09-01 | 2021-01-12 | 青岛秀山移动测量有限公司 | Planar internal missing point cloud repairing method based on image data assistance |
| CN112529891A (en) * | 2020-12-21 | 2021-03-19 | 深圳辰视智能科技有限公司 | Hollow hole identification and contour detection method and device based on point cloud and storage medium |
| CN113610726A (en) * | 2021-08-05 | 2021-11-05 | 华南农业大学 | Improved cubic B-spline interpolation repairing method for region limitation |
| CN114092469A (en) * | 2021-12-02 | 2022-02-25 | 四川大学 | Method and device for determining repair area of blade and readable storage medium |
| CN114581328A (en) * | 2022-03-03 | 2022-06-03 | 中冶赛迪重庆信息技术有限公司 | Material pile point cloud data interpolation method, device, equipment and medium |
| US11900639B2 (en) | 2018-01-19 | 2024-02-13 | Interdigital Vc Holdings, Inc. | Processing a point cloud |
| CN118037601A (en) * | 2024-04-07 | 2024-05-14 | 法奥意威(苏州)机器人系统有限公司 | Point cloud filling method and electronic equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102663823A (en) * | 2012-04-01 | 2012-09-12 | 北京农业信息技术研究中心 | Plant organ mesh surface hole filling method and system thereof |
| US20150036870A1 (en) * | 2013-07-30 | 2015-02-05 | The Boeing Company | Automated graph local constellation (glc) method of correspondence search for registration of 2-d and 3-d data |
| CN104361632A (en) * | 2014-11-03 | 2015-02-18 | 北京航空航天大学 | Triangular mesh hole-filling method based on Hermite radial basis function |
| CN104504663A (en) * | 2014-12-29 | 2015-04-08 | 佛山市诺威科技有限公司 | Iterative repair method for holes of false tooth triangular mesh model |
| CN106097436A (en) * | 2016-06-12 | 2016-11-09 | 广西大学 | A kind of three-dimensional rebuilding method of large scene object |
-
2017
- 2017-07-31 CN CN201710637964.0A patent/CN107590825B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102663823A (en) * | 2012-04-01 | 2012-09-12 | 北京农业信息技术研究中心 | Plant organ mesh surface hole filling method and system thereof |
| US20150036870A1 (en) * | 2013-07-30 | 2015-02-05 | The Boeing Company | Automated graph local constellation (glc) method of correspondence search for registration of 2-d and 3-d data |
| CN104361632A (en) * | 2014-11-03 | 2015-02-18 | 北京航空航天大学 | Triangular mesh hole-filling method based on Hermite radial basis function |
| CN104504663A (en) * | 2014-12-29 | 2015-04-08 | 佛山市诺威科技有限公司 | Iterative repair method for holes of false tooth triangular mesh model |
| CN106097436A (en) * | 2016-06-12 | 2016-11-09 | 广西大学 | A kind of three-dimensional rebuilding method of large scene object |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11900639B2 (en) | 2018-01-19 | 2024-02-13 | Interdigital Vc Holdings, Inc. | Processing a point cloud |
| CN108399609A (en) * | 2018-03-06 | 2018-08-14 | 北京因时机器人科技有限公司 | A kind of method for repairing and mending of three dimensional point cloud, device and robot |
| CN108399609B (en) * | 2018-03-06 | 2020-07-31 | 北京因时机器人科技有限公司 | Three-dimensional point cloud data repairing method and device and robot |
| CN109242828A (en) * | 2018-08-13 | 2019-01-18 | 浙江大学 | 3D printing product 3 D defects detection method based on optical grating projection multistep phase shift method |
| CN109242828B (en) * | 2018-08-13 | 2022-04-19 | 浙江大学 | Three-dimensional defect detection method for 3D printed product based on grating projection multistep phase shift method |
| CN109147040A (en) * | 2018-08-28 | 2019-01-04 | 浙江大学 | Human body dot cloud hole method for repairing and mending based on template |
| CN109147040B (en) * | 2018-08-28 | 2020-06-30 | 浙江大学 | Template-based human point cloud hole repairing method |
| CN109859114B (en) * | 2018-12-27 | 2020-10-16 | 北京大学 | Three-dimensional point cloud repairing method based on local smoothness and non-local similarity |
| CN109859114A (en) * | 2018-12-27 | 2019-06-07 | 北京大学 | Three-dimensional point cloud restorative procedure based on local flatness and non-local similitude |
| CN110415331A (en) * | 2019-06-14 | 2019-11-05 | 中国电建集团昆明勘测设计研究院有限公司 | Method for rapidly detecting, judging and repairing contour or hole based on point cloud data |
| CN110415331B (en) * | 2019-06-14 | 2022-08-19 | 中国电建集团昆明勘测设计研究院有限公司 | Method for rapidly detecting, judging and repairing contour or hole based on point cloud data |
| CN110349257B (en) * | 2019-07-16 | 2020-02-28 | 四川大学 | Phase pseudo mapping-based binocular measurement missing point cloud interpolation method |
| CN110349257A (en) * | 2019-07-16 | 2019-10-18 | 四川大学 | A kind of binocular measurement missing point cloud interpolating method based on the mapping of phase puppet |
| CN112102491A (en) * | 2020-08-12 | 2020-12-18 | 西安交通大学 | Skin damage surface three-dimensional reconstruction method based on surface structured light |
| CN112102491B (en) * | 2020-08-12 | 2022-12-06 | 西安交通大学 | Skin damage surface three-dimensional reconstruction method based on surface structured light |
| CN112215763A (en) * | 2020-09-01 | 2021-01-12 | 青岛秀山移动测量有限公司 | Planar internal missing point cloud repairing method based on image data assistance |
| CN112215763B (en) * | 2020-09-01 | 2023-02-24 | 青岛秀山移动测量有限公司 | Planar internal missing point cloud repairing method based on image data assistance |
| CN112529891A (en) * | 2020-12-21 | 2021-03-19 | 深圳辰视智能科技有限公司 | Hollow hole identification and contour detection method and device based on point cloud and storage medium |
| CN112529891B (en) * | 2020-12-21 | 2024-03-19 | 深圳辰视智能科技有限公司 | Method and device for identifying hollow holes and detecting contours based on point cloud and storage medium |
| CN113610726A (en) * | 2021-08-05 | 2021-11-05 | 华南农业大学 | Improved cubic B-spline interpolation repairing method for region limitation |
| CN114092469B (en) * | 2021-12-02 | 2022-08-26 | 四川大学 | Method and device for determining repair area of blade and readable storage medium |
| CN114092469A (en) * | 2021-12-02 | 2022-02-25 | 四川大学 | Method and device for determining repair area of blade and readable storage medium |
| CN114581328A (en) * | 2022-03-03 | 2022-06-03 | 中冶赛迪重庆信息技术有限公司 | Material pile point cloud data interpolation method, device, equipment and medium |
| CN118037601A (en) * | 2024-04-07 | 2024-05-14 | 法奥意威(苏州)机器人系统有限公司 | Point cloud filling method and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107590825B (en) | 2020-10-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107590825A (en) | A kind of dot cloud hole method for repairing and mending based on SFM | |
| CN111629193B (en) | Live-action three-dimensional reconstruction method and system | |
| CN108090960B (en) | A kind of Object reconstruction method based on geometrical constraint | |
| CN103744086B (en) | A kind of high registration accuracy method of ground laser radar and close-range photogrammetry data | |
| CN110223383A (en) | A kind of plant three-dimensional reconstruction method and system based on depth map repairing | |
| CN105160702B (en) | The stereopsis dense Stereo Matching method and system aided in based on LiDAR point cloud | |
| Zhou et al. | 2.5 D building modeling with topology control | |
| CN104063894A (en) | Point cloud three-dimensional model reestablishing method and system | |
| CN104346608A (en) | Sparse depth map densing method and device | |
| Habib et al. | Generation of complex polyhedral building models by integrating stereo-aerial imagery and lidar data | |
| JP6534296B2 (en) | Three-dimensional model generation device, three-dimensional model generation method, and program | |
| CN107730519A (en) | A kind of method and system of face two dimensional image to face three-dimensional reconstruction | |
| JP4058293B2 (en) | Generation method of high-precision city model using laser scanner data and aerial photograph image, generation system of high-precision city model, and program for generation of high-precision city model | |
| CN114463521B (en) | Building target point cloud rapid generation method for air-ground image data fusion | |
| CN107862735A (en) | A kind of RGBD method for reconstructing three-dimensional scene based on structural information | |
| CN104036541A (en) | Fast three-dimensional reconstruction method in vision measurement | |
| CN115128628A (en) | Road grid map construction method based on laser SLAM and monocular vision | |
| CN108171790A (en) | A kind of Object reconstruction method based on dictionary learning | |
| CN103544732A (en) | Three-dimensional reconstruction method for lunar vehicle | |
| CN114782357A (en) | Self-adaptive segmentation system and method for transformer substation scene | |
| ITRM20130115A1 (en) | PROCEDURE AND MATCHING DEVICE FOR THE DIGITAL MODELING OF OBJECTS BY STEREOSCOPIC IMAGES | |
| CN109801223A (en) | A kind of cloud denoising and empty method for repairing and mending | |
| CN116091706B (en) | Three-dimensional reconstruction method for multi-mode remote sensing image deep learning matching | |
| CN112002007A (en) | Model obtaining method and device based on air-ground image, equipment and storage medium | |
| CN116051980B (en) | Building identification method, system, electronic equipment and medium based on oblique photography |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |