CN106023156A - Point-cloud model and CAD model registering method based on detection features - Google Patents
Point-cloud model and CAD model registering method based on detection features Download PDFInfo
- Publication number
- CN106023156A CN106023156A CN201610303972.7A CN201610303972A CN106023156A CN 106023156 A CN106023156 A CN 106023156A CN 201610303972 A CN201610303972 A CN 201610303972A CN 106023156 A CN106023156 A CN 106023156A
- Authority
- CN
- China
- Prior art keywords
- point
- detection feature
- matrix
- cloud
- point set
- 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
Classifications
-
- 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
Abstract
The invention discloses a point-cloud model and CAD model registering method based on detection features. The registering method comprises the following steps: selecting Gauss curvature abrupt change points from a point-cloud model to form an initial registering control point set, and solving a corresponding matching point set on a CAD model; according to constraint conditions, finding three matching point pairs from the two point sets; according to the matching point pairs, solving an initial translation matrix and an initial rotation matrix; segmenting the CAD model into standard graphs, and according to detection feature point arrangement rules of the standard graphs, acquiring points from the CAD model to form a detection feature point set; by use of the initial rotation matrix and the initial translation matrix, acting on all points of the point-cloud model so as to form a point-cloud point set after conversion; and for the point-cloud point set and the detection feature point set, searching for a rotation matrix and a translation matrix which accurately register with each other by use of an ICP algorithm. The registering method takes both a registering speed and registering precision into consideration, and improves robustness of accurate registering.
Description
Technical field
Components Digital detection technique field of the present invention, particularly relates to a kind of point cloud model based on detection feature and CAD
The method for registering of model.
Background technology
In recent years, along with the digital measuring equipment such as three coordinate measuring machine, laser scanner, CT measuring instrument are at industrial circle
Extensive application, digitalization test plays extremely in terms of large-scale, the dimensional measurement of complex parts and accuracy of manufacture guarantee
Close important effect.Such as in the fields such as aircraft skin profile measurement, the detection of the rocket body accuracy of manufacture, cabin maintenance conditions, number
Wordization detection all plays an important role.
Generally, the point cloud model that digital measuring equipment obtains, the certainty of measurement of itself is generally possible to meet reality and surveys
Amount demand, yet with point cloud model with cad model not in the same space coordinate system, follow-up dimensional discrepancy analysis can be made by this
Become impact, in some instances it may even be possible to cause the erroneous judgement of part qualification so that the certainty of measurement of digitized measurement equipment loses meaning.Cause
This, for improving the part accuracy of manufacture and qualification rate, it is achieved point cloud model just becomes digitalization test with the registration of cad model
One key of application.
Registration about point cloud model Yu cad model mainly has method for registering based on point set iteration and based on geometry at present
The method for registering of feature.Wherein, method for registering based on point set iteration does not consider to detect the geometric properties of part, directly utilizes survey
The point of amount gained converge close the corresponding point found on cad model and by corresponding point realize iterating point cloud model with
The registration of cad model, the most representational is that (Iterative Closest Point is called for short iterative closest point algorithm
ICP algorithm).The time complexity of this type of algorithm significantly increases with a cloud number and the increase of iterations, therefore registrates effect
Rate is relatively low.Additionally, the registration control points of this kind of algorithm integrates the subset as point cloud model, registration accuracy depends critically upon a cloud mould
, when initial relative position deviation is bigger, easily there is error hiding, thus causes algorithm to lose efficacy in type, it is impossible to realize registration.
Method for registering based on geometric properties is by carrying out the point set that can reflect local geometric features in cloud data
Sampling analysis, then using the point set sampled as registration control points, thus reduce the quantity of registration point set, thus calculate speed
Hurry up.But owing to the geometric properties identification difficulty of complicated point cloud model is very big, and the point cloud model concordance of different part is poor,
The most this method for registering can lose the key feature of reflection part workmanship, and the accuracy causing registration result is the highest, no
Stable, practicality is the strongest.
In view of above-mentioned analysis, existing method for registering fails to make full use of that cad model contains more can reflect that part is crucial
The detection feature (such as can be as the geometric properties of examination criteria) of feature, and it is limited to the very the biggest of actual measurement acquisition
Cloud data amount, existing method for registering exists substantially not enough in terms of registration accuracy and registration speed, thus needs one badly
Preferably method for registering.
Summary of the invention
The technical problem to be solved in the present invention is that method for registering amount of calculation of the prior art is excessive, registrate effect in order to overcome
The defect that rate accuracy low, registration is not high enough, be not sufficiently stable, proposes a kind of point cloud model based on detection feature and CAD
The method for registering of model.
The present invention solves above-mentioned technical problem by following technical proposals:
The invention provides the method for registering of a kind of point cloud model based on detection feature and cad model, its feature is,
Comprise the following steps:
Step one, from point cloud model institute a little choose Gaussian curvature catastrophe point;
Step 2, the cloud data of Gaussian curvature catastrophe point to choose are constituted initial registration and are controlled point set, use ICP to calculate
Method tries to achieve coupling point set corresponding with this initial registration control point collection on cad model;
The constraints that step 3, basis are preset, controls point set from this initial registration and this match point is concentrated and found out at least
Three pairs of matching double points, every pair of matching double points includes an initial registration control point and a match point, and constraints includes every pair of coupling
Point to the spacing with center of gravity difference absolute value less than preset center of gravity spacing threshold, every pair of matching double points point method arrow
Dot product result is less than the dot product product threshold value preset, and these at least three pairs of matching double points are non-coplanar;
Step 4, obtaining initial translation matrix and initial rotation vector according to these at least three pairs of matching double points, this is initially put down
Moving matrix to act on the center of gravity at this at least three initial registration control point and obtain the center of gravity of this at least three match point, this initially revolves
It is equidirectional vector with the match point matched that torque battle array acts on the point obtained by initial registration control point;
Step 5, cad model is divided into test pattern, test pattern include line segment, plane, circle, ball, cylinder, circular cone,
Multiple in free form surface;
Step 6, detection feature according to each test pattern are layouted rule, adopt from the cad model of step 5 segmentation
Point, and constituted detection feature point set with the point adopted;
Step 7, utilize this initial rotation vector and this initial translation matrix act on point cloud model institute a little, with structure
Cloud point collection is put after becoming conversion;
Step 8, for conversion after put cloud point collection and detection feature point set, use ICP algorithm find meet preset termination
The spin matrix of accuracy registration between cloud point collection and detection feature point set and translation matrix is put after the conversion of condition.
It is preferred that step 8 includes:
The spin matrix R of accuracy registration is setkWith translation matrix TkInitial value R0And T0It is respectively unit matrix and 0 square
Battle array, the initial value arranging iteration count k is 1, and performs following iterative process:
Put after finding conversion in cloud point collection { S ' } that { point set that Q} is nearest obtains accurate matching double points with detection feature point set
Collection precise_match={ (gi, g 'i)|gi∈ { Q}, g 'i∈ { S ' }, i=1,2 ..., m}, wherein precise_match is
Accurately matching double points collection, (gi, gi') representing that the point that accurate matching double points is concentrated is right, m is the some logarithm that accurate matching double points is concentrated
Amount;
Utilize formulaWithCalculate accurate matching double points concentration and be contained in point after conversion respectively
Cloud point collection S ' } and { center of gravity μ of two point sets of Q} and μ ', utilize formula e to detect feature point seti=gi-μ and e 'i=g 'i-μ′
Make these two point sets to translate relative to the coincidence of respective center of gravity, wherein eiAnd ei' it is respectively some giAnd gi' knot after translation
Really;
Utilize formulaCloud point collection { S ' } and detection feature point set is put { after Q} translates after calculating conversion
Correlation matrix
Four-dimensional symmetrical matrix is constructed by each element in correlation matrix H
Four-dimensional symmetrical matrix according to structure is calculated spin matrix RkWith translation matrix Tk, wherein, by solving the four-dimension
Symmetrical matrix H '4×4Eigenvalue and characteristic vector, obtain Maximum characteristic root characteristic of correspondence vector for rotate quaternary number θ=
[θ1, θ2, θ3, θ4]T, then according to formulaMeter
Calculation obtains spin matrix Rk, according to formula Tk=μ '-Rkμ obtains translation matrix Tk;
According to formulaCalculate
Matching error amount Error of cloud point collection and detection feature point set, s in formula is put after the conversion of this and the previous time iterationiAnd qiFor
Put the corresponding point in cloud point collection and detection feature point set after conversion, then judge that whether matching error amount is less than the margin of error preset
Threshold value, if then terminating iterative process the spin matrix R of accuracy registration tried to achieve by current iterationkWith translation matrix TkAs
Iteration result, if otherwise making iteration count add 1, and performs above-mentioned iterative process again.
It is preferred that step one includes:
Calculate point cloud model Gaussian curvature a little;
Calculate the meansigma methods of institute's Gaussian curvature a little, and screen and obtain Gaussian curvature and be not less than the some conduct of this meansigma methods
Gaussian curvature catastrophe point.
It is preferred that step 3 includes:
Calculate this initial registration and control point set and the center of gravity of this coupling point set, and calculate this initial registration control point centrostigma
And the spacing of the center of gravity of this match point centrostigma and the most affiliated point set;
The method using least square fitting curved surface, calculates this initial registration control point centrostigma and this match point is concentrated
Respective some method of point is vowed;
Find out three to meeting the point of this constraints to as matching double points.
It is preferred that detection feature is layouted, rule includes following one or more:
It is that line segment is divided into multiple isometric sub-line section for the detection feature of line segment rule of layouting, and spaced
A point is adopted respectively in sub-line section etc. the sub-line section of quantity;
It is in the maximum rectangular envelope of plane, take uniform multiple of approximation for the detection feature of plane rule of layouting
Point;
Detection feature rule of layouting for circle is, circle is waited the isometric circular arc of point multistage, and takes one at every section of circular arc
Point;
It is to be multiple triangular plate by free form surface subdivision for the detection feature of free form surface rule of layouting, takes each three
The center of gravity point of gusset plate;
For the detection feature of ball rule of layouting be, for radius be r ball be b by distance plane cut and formed
Sphere, if sampling site number amounts to L, then sampling site is evenly distributed on quantity is lcIndividual parallel and equidistant ball cross-sectional periphery
On, wherein according to formulaCalculate ball number of cross sections l of equi-spaced apartc, according to formulaCalculate
Sampling site number l in ball cross-sectional peripheryp, wherein quantity is lpPoint uniform in ball cross-sectional periphery;
It is to be the cylinder of r for a height of h, radius, if sampling site number amounts to L for the detection feature of cylinder rule of layouting
Individual, then sampling site is evenly distributed on lcOn individual parallel and equidistant cylindrical cross-section circumference, wherein according to formula
Calculate cross section quantity l of equi-spaced apartc, according to formulaCalculate the sampling site number l in each cross-sectional circumferentialp, its
Middle quantity is lpPoint uniform in cross-sectional circumferential.
It is to be r for a height of h, a length of d of bus, two ends radius for the detection feature of circular cone rule of layouting1、r2And r2>
r1Circular cone, if sampling site number amounts to L, then sampling site is evenly distributed on lcIn individual parallel and equidistant circular cone cross-sectional circumferential, its
Middle according to formulaCalculate circular cone cross section quantity l of equi-spaced apartc, according to public affairs
FormulaCalculate the sampling site number l in each circular cone cross-sectional circumferentialp, wherein quantity is lp
Point uniform in circular cone cross-sectional circumferential.
It is preferred that this method for registering also includes:
Step 9, the spin matrix of accuracy registration step 8 tried to achieve and translation matrix act on point cloud model, to produce
Raw registration result.
On the basis of meeting common sense in the field, above-mentioned each optimum condition, can combination in any, obtain each preferable reality of the present invention
Example.
The most progressive effect of the present invention is:
The point cloud model based on detection feature of the present invention and the method for registering of cad model, by the initial registration stage
With the cloud data that simplifies for registration control points collection, decrease amount of calculation, and utilize distance and method to vow that constraint reduces initial registration
Space bias, make use of the metastable detection characteristic point on cad model is that guiding realizes point cloud model and CAD mould simultaneously
The accuracy registration of type, the most preferably embodying piece test process while reducing registration amount of calculation needs the geometry paid close attention to special
Levy, thus help in robustness and the accuracy of testing result improving accuracy registration, taken into account registration speed and registration accuracy.
Accompanying drawing explanation
Fig. 1 is the point cloud model based on the detection feature method for registering with cad model of a preferred embodiment of the present invention
Flow chart.
Detailed description of the invention
Provide present pre-ferred embodiments below in conjunction with the accompanying drawings, to describe technical scheme in detail, but not because of
This limits the present invention among described scope of embodiments.
In the present invention, relate to the expression for point, center of gravity, method arrow etc. calculating and occurring in the part of correlation formula,
It is interpreted as vector representation, relates to the calculating that the calculating of point, center of gravity, method arrow etc. is interpreted as carrying out for vector, unless otherwise
Explicitly stated or obviously relate to scalar or the calculating for scalar.
With reference to shown in Fig. 1, the point cloud model based on detection feature of a preferred embodiment of the present invention and the registration of cad model
Method, the point cloud model being directed to can be the some cloud mould being measured for any parts, product or device
Type, the cad model also including such as Element Design being directed to.Hereinafter the method for registering of this preferred embodiment is carried out example
Property ground explanation.
First, screen according to the data that point cloud model is comprised by Gaussian curvature, from point cloud model institute a little
Choose Gaussian curvature catastrophe point.Such as, the Gaussian curvature catastrophe point in following steps screening point set point cloud model can be used, at point
The point set of cloud model or perhaps data set { in S}, calculate each some SiGaussian curvature KSi, wherein i={1,2 ... N}, N are
Point set { the total quantity of the point that S} is comprised.IfThen put SiRetain, otherwise delete.Use and above-mentioned steps class
As any point set of choosing { S} has the point of of a relatively high Gaussian curvature, or any chooses point set { Gaussian curvature in S}
More than the point of a certain specific preset value, it is the most all feasible.
Then, the cloud data of the Gaussian curvature catastrophe point to choose is constituted initial registration and is controlled point set, uses ICP algorithm
Try to achieve coupling point set corresponding with initial registration control point collection on cad model.
According to default constraints, control point set from initial registration and match point is concentrated and found out at least three pairs of match points
Right, every pair of matching double points includes an initial registration control point and a match point, constraints include every pair of matching double points with weight
The absolute value of the difference of the spacing of the heart is less than the center of gravity spacing threshold preset, and the dot product result that the some method of every pair of matching double points is vowed is little
In default dot product product threshold value, and at least three pairs of matching double points are non-coplanar.
Specifically, in the present embodiment, control point set from initial registration and match point concentrated and found out three pairs of matching double points,
And find out three pairs of matching double points and use following steps: first, calculate initial registration and control point set and the center of gravity of coupling point set, and
Calculate the spacing of initial registration control point centrostigma and match point centrostigma and the center of gravity of the most affiliated point set;Secondly, use
The method of least square fitting curved surface, calculates initial registration control point centrostigma and respective some method of match point centrostigma is vowed;
Find out three to meeting the point of constraints to as matching double points.
As above the process finding out three pairs of matching double points is concentrated from initial registration control point set and match point, for example, can
To be realized by calculating process in detail below.
First, formula is utilizedWithCalculate two point sets (i.e. initial registration control point set P} and
Coupling point set { P ' }) center of gravity O and O ', then control point set { P} midpoint piAnd coupling point set { P, } midpoint p 'iTo respective center of gravity
Distance be respectively di=| | pi-O | | and d 'i=| | p 'i-O′||.It should be noted that i here travels through 1 to N, and N is each
Point concentrates the sum of the point comprised.In the following description, i typically can be regarded as traveling through all elements in involved set,
Except as otherwise noted.
Then, use the mode of least square fitting curved surface, determine control point set { P} midpoint piMethod vow Npi, specifically
Method is: with a piFor the centre of sphere, with given r value as radius, it is thus achieved that some piNeighborhood point set nbhd (pi), recycle formulaObtain a piNeighborhood nbhd (pi) center of gravity, then by resolve Optimality equationsObtain a piLeast square fitting curved surface method vow Npi, wherein phiFor a piNeighbour
Territory nbhd (piPoint in).Therewith it is likewise possible to determine cad model correspondence point set (i.e. mating point set) { P ' } midpoint p 'iMethod
Vow
Constraint IF conditional expressionWhether it is true, its
In, ε1、ε2For the threshold value pre-set, Count is to be effectively matched enumerator a little, if above-mentioned constraints expression formula is true, then
Determine whether that select three pairs of matching double points, whether at same plane, if at same plane, remove wherein any pair coupling
Point is right, and again carries out above-mentioned calculating process, if not at same plane, performs subsequent step.If above-mentioned constraints expression formula
It is false, the most again carries out above-mentioned calculating process.
Be effectively matched a little to rear successfully selecting three couple the most in the same plane, according to its obtain initial translation matrix and
Initial rotation vector.Initial translation matrix acts on the center of gravity at least three initial registration control point and can get at least three coupling
The center of gravity of point, it is equidirectional with the match point matched that initial rotation vector acts on the point obtained by initial registration control point
Vector.Such as, can by by from initial registration controls point set and mate point set three pairs of matching double points in, will be from initially joining
The center of gravity of accurate three points controlling point set and the center of gravity coincidence of three points from coupling point set, and obtain initial registration translation square
Battle array T0, recycle formula pi=R0·pi' solve initial registration spin matrix R0, P in this formulaiAnd pi' be respectively from
Initial registration control point set and from a pair in the corresponding point of coupling point set or travel through whole 3 right.
After obtaining initial registration translation matrix and initial registration spin matrix, segmentation cad model is to obtain detection feature
Point set.Such as, cad model is divided into the multiple standard including in line segment, plane, circle, ball, cylinder, circular cone, free form surface
Figure.Detection feature according to each test pattern is layouted rule, sampling site from the cad model of segmentation, and with the some structure adopted
Become detection feature point set.It is said that in general, the more meeting of sampling site number more precisely describes detection feature, add intensive simultaneously, adopt
Point quantity can select according to actual needs.
Layout rule for the detection feature of test pattern, be exemplified below.
Can be that line segment is divided into multiple isometric sub-line section for the detection feature of line segment rule of layouting, and mutually
A point is adopted respectively in the sub-line section of the sub-line section of the quantity such as interval.
Can be in the maximum rectangular envelope of plane, to take approximation uniform many for the detection feature of plane rule of layouting
Individual point.
Detection feature rule of layouting for circle can be, circle is waited the isometric circular arc of point multistage, and takes at every section of circular arc
One point.
Can be to be multiple triangular plate by free form surface subdivision for the detection feature of free form surface rule of layouting, take every
The center of gravity point of individual triangular plate.
For the detection feature of ball rule of layouting can be, for radius be r ball be b by distance plane cut
The sphere formed, if sampling site number amounts to L, then sampling site is evenly distributed on quantity is lcIndividual parallel and equidistant ball cross section circle
Zhou Shang, wherein according to formulaCalculate ball number of cross sections l of equi-spaced apartc, according to formulaCalculate
Sampling site number l in ball cross-sectional peripheryp, wherein quantity is lpPoint uniform in ball cross-sectional periphery.
Can be to be the cylinder of r for a height of h, radius for the detection feature of cylinder rule of layouting, if sampling site number amounts to
For L, then sampling site is evenly distributed on lcOn individual parallel and equidistant cylindrical cross-section circumference, wherein according to formulaCalculate cross section quantity l of equi-spaced apartc, according to formulaCalculate in each cross-sectional circumferential
Sampling site number lp, wherein quantity is lpPoint uniform in cross-sectional circumferential.
Can be to be r for a height of h, a length of d of bus, two ends radius for the detection feature of circular cone rule of layouting1、r2And
r2> r1Circular cone, if sampling site number amounts to L, then sampling site is evenly distributed on lcIndividual parallel and equidistant circular cone cross-sectional circumferential
On, wherein according to formulaCalculate circular cone cross section quantity l of equi-spaced apartc,
According to formulaCalculate the sampling site number l in each circular cone cross-sectional circumferentialp, wherein count
Amount is lpPoint uniform in circular cone cross-sectional circumferential.
It should be understood that the calculating process related in the above-mentioned detection feature for test pattern layouts rule, be
Scalar operation process.
After the above step, by with the detection feature point set of cad model, { Q}, for guiding, uses ICP algorithm to find some cloud
Model data collection is { with the detection feature point set { point set that Q} is nearest, thus realize accurately joining of point cloud model and cad model in S}
Accurate.
The institute of point cloud model is acted on a little, with point after composition conversion first with initial rotation vector and initial translation matrix
Cloud point collection (i.e. after initial registration, point converges).Then for putting cloud point collection and detection feature point set after conversion, ICP algorithm is used to seek
Look for meet preset end condition conversion after put cloud point collection and detection feature point set between accuracy registration spin matrix and
Translation matrix.For specifically calculating process, it is illustrated below.
The spin matrix R of accuracy registration is setkWith translation matrix TkInitial value R0And T0It is respectively unit matrix and 0 square
Battle array, the initial value arranging iteration count k is 1, and performs following iterative process.
First, put after finding conversion in cloud point collection { S ' } that { point set that Q} is nearest is accurately mated with detection feature point set
Point is to collection precise_match={ (gi, g 'i)|gi∈ { Q}, g 'i∈ { S ' }, i=1,2 ..., m}, wherein precise_
Match is accurate matching double points collection, (gi, gi') representing that the point that accurate matching double points is concentrated is right, m is that accurate matching double points is concentrated
Point is to quantity;
Utilize formulaWithCalculate accurate matching double points concentration and be contained in point after conversion respectively
Cloud point collection S ' } and { center of gravity μ of two point sets of Q} and μ ', utilize formula e to detect feature point seti=gi-μ and e 'i=g 'i-μ′
Make two point sets to translate relative to the coincidence of respective center of gravity, wherein eiAnd ei' it is respectively some giAnd gi' result after translation;
Utilize formulaCloud point collection { S ' } and detection feature point set is put { after Q} translates after calculating conversion
Correlation matrix
Four-dimensional symmetrical matrix is constructed by each element in correlation matrix H
Four-dimensional symmetrical matrix according to structure is calculated spin matrix RkWith translation matrix Tk, wherein, by solving the four-dimension
Symmetrical matrix H '4×4Eigenvalue and characteristic vector, obtain Maximum characteristic root characteristic of correspondence vector for rotate quaternary number θ=
[θ1, θ2, θ3, θ4]T, then according to formulaMeter
Calculation obtains spin matrix Rk, according to formula Tk=μ '-Rkμ obtains translation matrix Tk;
According to formulaCalculate
Matching error amount Error of cloud point collection and detection feature point set, s in formula is put after the conversion of this and the previous time iterationiAnd qiFor
Put the corresponding point in cloud point collection and detection feature point set after conversion, then judge that whether matching error amount is less than the margin of error preset
Threshold value, if then terminating iterative process the spin matrix R of accuracy registration tried to achieve by current iterationkWith translation matrix TkAs
Iteration result, if otherwise making iteration count add 1, and performs above-mentioned iterative process again.
The spin matrix R of accuracy registration is tried to achieve in iterative computation as abovekWith translation matrix TkAfter, by the rotation of accuracy registration
Torque battle array and translation matrix act on point cloud model, so that it may produce accurate registration result.
Although the foregoing describing the detailed description of the invention of the present invention, it will be appreciated by those of skill in the art that these
Being merely illustrative of, protection scope of the present invention is defined by the appended claims.Those skilled in the art is not carrying on the back
On the premise of the principle and essence of the present invention, these embodiments can be made various changes or modifications, but these change
Protection scope of the present invention is each fallen within amendment.
Claims (6)
1. a point cloud model based on detection feature and the method for registering of cad model, it is characterised in that comprise the following steps:
Step one, from point cloud model institute a little choose Gaussian curvature catastrophe point;
Step 2, the cloud data of Gaussian curvature catastrophe point to choose are constituted initial registration and are controlled point set, use ICP algorithm to ask
Obtain coupling point set corresponding with this initial registration control point collection on cad model;
The constraints that step 3, basis are preset, finds out at least three from this initial registration control point set and this match point concentration right
Matching double points, every pair of matching double points includes an initial registration control point and a match point, and constraints includes every pair of matching double points
The spacing with center of gravity difference absolute value less than preset center of gravity spacing threshold, every pair of matching double points point method arrow dot product
Result is less than the dot product product threshold value preset, and these at least three pairs of matching double points are non-coplanar;
Step 4, obtain initial translation matrix and initial rotation vector according to these at least three pairs of matching double points, this initial translation square
Battle array acts on the center of gravity at this at least three initial registration control point and obtains the center of gravity of this at least three match point, this initial rotation square
It is equidirectional vector with the match point matched that battle array acts on the point obtained by initial registration control point;
Step 5, cad model being divided into test pattern, test pattern includes line segment, plane, circle, ball, cylinder, circular cone, freedom
Multiple in curved surface;
Step 6, detection feature according to each test pattern are layouted rule, sampling site from the cad model of step 5 segmentation, and
Detection feature point set is constituted with the point adopted;
Step 7, utilize this initial rotation vector and this initial translation matrix act on point cloud model institute a little, to constitute change
Cloud point collection is put after changing;
Step 8, for conversion after put cloud point collection and detection feature point set, use ICP algorithm find meet preset end condition
Conversion after put the spin matrix of accuracy registration between cloud point collection and detection feature point set and translation matrix.
2. method for registering as claimed in claim 1, it is characterised in that step 8 includes:
The spin matrix R of accuracy registration is setkWith translation matrix TkInitial value R0And T0It is respectively unit matrix and 0 matrix, if
The initial value putting iteration count k is 1, and performs following iterative process:
Put after finding conversion in cloud point collection { S ' } that { point set that Q} is nearest obtains accurate matching double points collection with detection feature point set
Precise_match={ (gi, g 'i)|gi∈ { Q}, g 'i∈ { S ' }, i=1,2 ..., m}, wherein precise_match is essence
Really matching double points collection, (gi, gi') representing that the point that accurate matching double points is concentrated is right, m is that the point of accurate matching double points concentration is to quantity;
Utilize formulaWithCalculate after accurate matching double points concentration is contained in conversion respectively and put cloud point collection
S ' } and { center of gravity μ of two point sets of Q} and μ ', utilize formula e to detect feature point seti=gi-μ and e 'i=g 'i-μ ' to this two
Individual point set is made to translate relative to the coincidence of respective center of gravity, wherein eiAnd ei' it is respectively some giAnd gi' result after translation;
Utilize formulaCloud point collection { S ' } and detection feature point set { being correlated with after Q} translation is put after calculating conversion
Matrix
Four-dimensional symmetrical matrix is constructed by each element in correlation matrix H
Four-dimensional symmetrical matrix according to structure is calculated spin matrix RkWith translation matrix Tk, wherein, four-dimensional symmetrical by solving
Matrix H '4×4Eigenvalue and characteristic vector, obtain Maximum characteristic root characteristic of correspondence vector for rotate quaternary number θ=[θ1, θ2,
θ3, θ4]T, then according to formulaIt is calculated
Spin matrix Rk, according to formula Tk=μ '-Rkμ obtains translation matrix Tk;
According to formulaCalculate this
Matching error amount Error of cloud point collection and detection feature point set, s in formula is put after conversion with front an iterationiAnd qiFor conversion
Corresponding point in rear some cloud point collection and detection feature point set, then judge that whether matching error amount is less than the margin of error threshold preset
Value, if then terminating iterative process the spin matrix R of accuracy registration tried to achieve by current iterationkWith translation matrix TkAs repeatedly
For result, if otherwise making iteration count add 1, and again perform above-mentioned iterative process.
3. method for registering as claimed in claim 1, it is characterised in that step one includes:
Calculate point cloud model Gaussian curvature a little;
Calculate the meansigma methods of institute's Gaussian curvature a little, and screen and obtain Gaussian curvature and be not less than the point of this meansigma methods as Gauss
Curvature mutation point.
4. method for registering as claimed in claim 1, it is characterised in that step 3 includes:
Calculate this initial registration and control point set and the center of gravity of this coupling point set, and calculate this initial registration control point centrostigma and
The spacing of the center of gravity of this match point centrostigma and the most affiliated point set;
The method using least square fitting curved surface, calculates this initial registration control point centrostigma and this match point centrostigma is each
From some method vow;
Find out three to meeting the point of this constraints to as matching double points.
5. method for registering as claimed in claim 1, it is characterised in that detection feature rule of layouting includes following one or many
:
It is that line segment is divided into multiple isometric sub-line section for the detection feature of line segment rule of layouting, and at the number such as spaced
A point is adopted respectively in the sub-line section of the sub-line section of amount;
It is in the maximum rectangular envelope of plane, take multiple points that approximation is uniform for the detection feature of plane rule of layouting;
Detection feature rule of layouting for circle is, circle is waited the isometric circular arc of point multistage, and takes a point at every section of circular arc;
It is to be multiple triangular plate by free form surface subdivision for the detection feature of free form surface rule of layouting, takes each triangular plate
Center of gravity point;
For the detection feature of ball rule of layouting be, for radius be r ball be b by distance plane cut the ball formed
Face, if sampling site number amounts to L, then sampling site is evenly distributed on quantity is lcIn individual parallel and equidistant ball cross-sectional periphery, its
Middle according to formulaCalculate ball number of cross sections l of equi-spaced apartc, according to formulaCalculate ball cross section
Sampling site number l on circumferencep, wherein quantity is lpPoint uniform in ball cross-sectional periphery;
It is to be the cylinder of r for a height of h, radius for the detection feature of cylinder rule of layouting, if sampling site number amounts to L, then
Sampling site is evenly distributed on lcOn individual parallel and equidistant cylindrical cross-section circumference, wherein according to formulaCalculate
Cross section quantity l of equi-spaced apartc, according to formulaCalculate the sampling site number l in each cross-sectional circumferentialp, wherein count
Amount is lpPoint uniform in cross-sectional circumferential;
It is to be r for a height of h, a length of d of bus, two ends radius for the detection feature of circular cone rule of layouting1、r2And r2> r1's
Circular cone, if sampling site number amounts to L, then sampling site is evenly distributed on lcIn individual parallel and equidistant circular cone cross-sectional circumferential, wherein
According to formulaCalculate circular cone cross section quantity l of equi-spaced apartc, according to formulaCalculate the sampling site number l in each circular cone cross-sectional circumferentialp, wherein quantity is lp's
Point is uniform in circular cone cross-sectional circumferential.
6. the method for registering as described in any one in claim 1-5, it is characterised in that this method for registering also includes:
Step 9, the spin matrix of accuracy registration step 8 tried to achieve and translation matrix act on point cloud model, join to produce
Quasi-result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610303972.7A CN106023156B (en) | 2016-05-10 | 2016-05-10 | The method for registering of point cloud model and CAD model based on detection feature |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610303972.7A CN106023156B (en) | 2016-05-10 | 2016-05-10 | The method for registering of point cloud model and CAD model based on detection feature |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106023156A true CN106023156A (en) | 2016-10-12 |
CN106023156B CN106023156B (en) | 2018-09-11 |
Family
ID=57099705
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610303972.7A Active CN106023156B (en) | 2016-05-10 | 2016-05-10 | The method for registering of point cloud model and CAD model based on detection feature |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106023156B (en) |
Cited By (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106846382A (en) * | 2017-01-22 | 2017-06-13 | 深圳市唯特视科技有限公司 | A kind of image registration object detection method based on histogram control point |
CN107170003A (en) * | 2017-05-08 | 2017-09-15 | 广东工业大学 | The model matching method and system of a kind of sail shape outside plate |
CN107730434A (en) * | 2017-11-09 | 2018-02-23 | 哈尔滨工业大学 | A kind of method that numerical model generation point cloud model is carefully seen by two-phase composite material |
CN108986149A (en) * | 2018-07-16 | 2018-12-11 | 武汉惟景三维科技有限公司 | A kind of point cloud Precision Registration based on adaptive threshold |
CN109270079A (en) * | 2018-11-15 | 2019-01-25 | 燕山大学 | A kind of Surface Flaw accurate detecting method based on point cloud model |
US10260870B2 (en) | 2017-09-19 | 2019-04-16 | Industrial Technology Research Institute | On-line measuring system, datum calibrating method, deviation measuring method and computer-readable medium |
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 |
CN110288636A (en) * | 2019-05-05 | 2019-09-27 | 中国矿业大学 | It is a kind of based on plane characteristic constraint LiDAR point cloud without initial value method for registering |
CN110827382A (en) * | 2019-11-11 | 2020-02-21 | 杭州都市高速公路有限公司 | Automatic inspection method for arc hinge joint structural size of assembled culvert segment |
CN111429494A (en) * | 2020-04-13 | 2020-07-17 | 中国空气动力研究与发展中心超高速空气动力研究所 | Biological vision-based point cloud high-precision automatic registration method |
CN112289416A (en) * | 2020-12-18 | 2021-01-29 | 南京佗道医疗科技有限公司 | Method for evaluating guide needle placement accuracy |
CN113155957A (en) * | 2021-03-03 | 2021-07-23 | 广东工业大学 | Ultrasonic guiding method for internal structure curved surface of large-deformation part |
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 |
CN113741337A (en) * | 2021-09-10 | 2021-12-03 | 哈尔滨工业大学 | Planning method and device for machining track of all-surface uniformly-distributed micro-pit structure of thin-wall spherical shell type micro component |
CN114111593A (en) * | 2021-11-25 | 2022-03-01 | 征图新视(江苏)科技股份有限公司 | Keyboard automatic detection equipment and imaging method thereof |
CN114274139A (en) * | 2020-09-27 | 2022-04-05 | 西门子股份公司 | Automatic spraying method, device, system and storage medium |
CN114782508A (en) * | 2022-06-21 | 2022-07-22 | 杭州三坛医疗科技有限公司 | Hip joint registration point screening method and device |
CN116342671A (en) * | 2023-05-23 | 2023-06-27 | 第六镜科技(成都)有限公司 | Point cloud and CAD model registration method, device, electronic equipment and storage medium |
CN114274139B (en) * | 2020-09-27 | 2024-04-19 | 西门子股份公司 | Automatic spraying method, device, system and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102169579A (en) * | 2011-03-31 | 2011-08-31 | 西北工业大学 | Rapid and accurate registration method of dense point cloud model |
CN102880737A (en) * | 2012-07-30 | 2013-01-16 | 北京信息科技大学 | Workpiece registering method and system based on flexible assembly |
CN103236064A (en) * | 2013-05-06 | 2013-08-07 | 东南大学 | Point cloud automatic registration method based on normal vector |
CN103995496A (en) * | 2014-04-28 | 2014-08-20 | 南京航空航天大学 | Aircraft part high-precision matching component processing method based on digital measurement |
-
2016
- 2016-05-10 CN CN201610303972.7A patent/CN106023156B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102169579A (en) * | 2011-03-31 | 2011-08-31 | 西北工业大学 | Rapid and accurate registration method of dense point cloud model |
CN102880737A (en) * | 2012-07-30 | 2013-01-16 | 北京信息科技大学 | Workpiece registering method and system based on flexible assembly |
CN103236064A (en) * | 2013-05-06 | 2013-08-07 | 东南大学 | Point cloud automatic registration method based on normal vector |
CN103995496A (en) * | 2014-04-28 | 2014-08-20 | 南京航空航天大学 | Aircraft part high-precision matching component processing method based on digital measurement |
Cited By (27)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106846382A (en) * | 2017-01-22 | 2017-06-13 | 深圳市唯特视科技有限公司 | A kind of image registration object detection method based on histogram control point |
CN107170003A (en) * | 2017-05-08 | 2017-09-15 | 广东工业大学 | The model matching method and system of a kind of sail shape outside plate |
US10260870B2 (en) | 2017-09-19 | 2019-04-16 | Industrial Technology Research Institute | On-line measuring system, datum calibrating method, deviation measuring method and computer-readable medium |
TWI659205B (en) * | 2017-09-19 | 2019-05-11 | 財團法人工業技術研究院 | On-line measuring system, coordinate calibrating method, deviation measuring method and computer-readable medium |
CN107730434B (en) * | 2017-11-09 | 2020-12-01 | 哈尔滨工业大学 | Method for generating point cloud model by two-phase composite material mesoscopic numerical model |
CN107730434A (en) * | 2017-11-09 | 2018-02-23 | 哈尔滨工业大学 | A kind of method that numerical model generation point cloud model is carefully seen by two-phase composite material |
CN108986149A (en) * | 2018-07-16 | 2018-12-11 | 武汉惟景三维科技有限公司 | A kind of point cloud Precision Registration based on adaptive threshold |
CN109270079A (en) * | 2018-11-15 | 2019-01-25 | 燕山大学 | A kind of Surface Flaw accurate detecting method based on point cloud model |
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 |
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 |
CN110288636A (en) * | 2019-05-05 | 2019-09-27 | 中国矿业大学 | It is a kind of based on plane characteristic constraint LiDAR point cloud without initial value method for registering |
CN110827382A (en) * | 2019-11-11 | 2020-02-21 | 杭州都市高速公路有限公司 | Automatic inspection method for arc hinge joint structural size of assembled culvert segment |
CN111429494A (en) * | 2020-04-13 | 2020-07-17 | 中国空气动力研究与发展中心超高速空气动力研究所 | Biological vision-based point cloud high-precision automatic registration method |
CN111429494B (en) * | 2020-04-13 | 2023-04-07 | 中国空气动力研究与发展中心超高速空气动力研究所 | Biological vision-based point cloud high-precision automatic registration method |
CN114274139B (en) * | 2020-09-27 | 2024-04-19 | 西门子股份公司 | Automatic spraying method, device, system and storage medium |
CN114274139A (en) * | 2020-09-27 | 2022-04-05 | 西门子股份公司 | Automatic spraying method, device, system and storage medium |
CN112289416A (en) * | 2020-12-18 | 2021-01-29 | 南京佗道医疗科技有限公司 | Method for evaluating guide needle placement accuracy |
CN113155957A (en) * | 2021-03-03 | 2021-07-23 | 广东工业大学 | Ultrasonic guiding method for internal structure curved surface of large-deformation part |
CN113327275B (en) * | 2021-06-18 | 2022-11-08 | 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 |
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 |
CN113741337B (en) * | 2021-09-10 | 2023-02-03 | 哈尔滨工业大学 | Planning method and device for machining track of all-surface uniformly-distributed micro-pit structure of thin-wall spherical shell type micro component |
CN113741337A (en) * | 2021-09-10 | 2021-12-03 | 哈尔滨工业大学 | Planning method and device for machining track of all-surface uniformly-distributed micro-pit structure of thin-wall spherical shell type micro component |
CN114111593A (en) * | 2021-11-25 | 2022-03-01 | 征图新视(江苏)科技股份有限公司 | Keyboard automatic detection equipment and imaging method thereof |
CN114111593B (en) * | 2021-11-25 | 2024-02-27 | 征图新视(江苏)科技股份有限公司 | Keyboard automatic detection equipment and imaging method thereof |
CN114782508A (en) * | 2022-06-21 | 2022-07-22 | 杭州三坛医疗科技有限公司 | Hip joint registration point screening method and device |
CN116342671A (en) * | 2023-05-23 | 2023-06-27 | 第六镜科技(成都)有限公司 | Point cloud and CAD model registration method, device, electronic equipment and storage medium |
CN116342671B (en) * | 2023-05-23 | 2023-08-08 | 第六镜科技(成都)有限公司 | Point cloud and CAD model registration method, device, electronic equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN106023156B (en) | 2018-09-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106023156A (en) | Point-cloud model and CAD model registering method based on detection features | |
CN106600622B (en) | A kind of point cloud data segmentation method based on super voxel | |
CN102779345B (en) | A kind of some cloud Precision Registration based on center of gravity Euclidean distance | |
CN110335297A (en) | A kind of point cloud registration method based on feature extraction | |
CN105783786B (en) | Part chamfering measuring method and device based on structured light vision | |
CN108010116A (en) | Point cloud feature point detecting method and point cloud feature extracting method | |
CN102607459B (en) | Splicing method and splicing device of Lidar measurement data | |
CN105913489A (en) | Indoor three-dimensional scene reconstruction method employing plane characteristics | |
CN103106632B (en) | A kind of fusion method of the different accuracy three dimensional point cloud based on average drifting | |
Li et al. | A leaf segmentation and phenotypic feature extraction framework for multiview stereo plant point clouds | |
CN106597363A (en) | Pedestrian location method in indoor WLAN environment | |
CN102999937A (en) | Curved planar reconstruction method for cardiac scattered-point cloud data | |
CN113327275B (en) | Point cloud double-view-angle fine registration method based on multi-constraint point to local curved surface projection | |
CN102997875B (en) | A kind of analytical approach of Complex Helical Surface high-acruracy survey uncertainty | |
CN108986149A (en) | A kind of point cloud Precision Registration based on adaptive threshold | |
CN109902425A (en) | The tunnel cross-section extracting method of ground formula point cloud | |
CN103278126A (en) | Sphericity error assessment method for part based on minimum area | |
CN104463826A (en) | Novel point cloud parallel Softassign registering algorithm | |
CN105354850A (en) | Method for detecting size of complex curved part in three-dimensional matching manner based on electric-field properties | |
CN112446844A (en) | Point cloud feature extraction and registration fusion method | |
CN105184786A (en) | Floating-point-based triangle characteristic description method | |
US9383429B2 (en) | System and method for determining geo location of a target using locus of emitter positions (LEP) | |
CN104596464B (en) | Swing arm contour surface shape detection method based on combination of radial scanning and circular scanning | |
CN103292654B (en) | A kind of method calculating function size of cylindrical part | |
You et al. | An Eulerian method for computing the coherent ergodic partition of continuous dynamical systems |
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 |