CN100460813C - Three-D connection rod curve matching rate detection method - Google Patents

Abstract

A check method of three-dimension spatial connecting rod curvature match degree comprises following steps that standardizing an object connecting rod curvature data, referring to the standard curvature, using wringing number and distance of sub section couple to divide all sub section couples into a plurality of equivalent classes, calculating out the rigid conversion between the sub section couples of different curvatures in each equivalent class, to find maximum match sub section couple to realize local match of curvature, cluster analyzing all rigid conversions, using the maximum cluster as the best match of connecting rod curvature, to obtain maximum rigid conversion, match degree and match error, to display match effect. The invention can effectively resolve the three-dimension spatial connecting rod curvature match problem.

Description

The detection method of three-D connection rod curve matching rate

Technical field

The present invention relates to the detection method in a kind of mechanical designing technique field, particularly a kind of detection method of three-D connection rod curve matching rate.

Background technology

Many problems that nature exists can abstractly be the problem of geometric locus, and these geometric locuses can be bidimensional, three-dimensional or higher-dimension more, such as the identification of object in the computer vision, need with object abstract be the outline line of bidimensional or three-dimensional; The property analysis of protein in the molecular biology, need with the protein molecule chain abstract be three-dimensional space broken line; During mechanism design, often face the problem of special exercise track how to realize that the user proposes, promptly how to make the coupler-point curve of designed mechanism, more approach the trajectory path that the user needs.In addition, all relate to the problem analysis of curve at three-dimensional model of robot trajectory planning, control system, three-dimensional magnetic field and flow field, brain and heart etc.Analyzing these curves just need compare the similarity of curve.Described similarity just is meant the matching degree between the curve.

Be specific to the coupler-point curve problem, in Creative Mechanism Design, usually running into needs to solve the problem that realizes some special exercise track, promptly how to compare in a plurality of coupler-point curves, so that obtain more to approach the linkage assembly of actual design curve.So coupler-point curve similarity problem determination comprehensively has theory significance and practical value to design and the mechanism kinematic that solves structure innovation mechanism.

Find through literature search prior art, Chu Jinkui etc. are at " mechanical engineering journal " (1993,29 (5): " the reproducing the comprehensive of four-bar linkage coupler-point curve " of delivering 117～122) with fast fourier transform, this article proposes the planar linkage curve to be studied in frequency domain with fast fourier transform, with any plane coupler-point curve approximate representation is limited several spectrum components, differentiate the index of matching degree as different coupler-point curves with this, be convenient to compare between the coupler-point curve, realized the target that mechanism path is comprehensive.But its deficiency is that the fast fourier transform that the method adopts can only be applied to two-dimensional data, thereby can only carry out the detection of planar linkage curve matching rate, can not be applied to three-dimensional coupler-point curve.

Summary of the invention

The object of the invention is at the deficiencies in the prior art, proposes a kind of detection method of the three-D connection rod curve matching rate based on knot theory, makes it can solve puzzlement three-D connection rod curve matching problem for a long time.

The present invention is achieved by the following technical solutions, and the present invention comprises following steps:

A. data normalization: coupler-point curve data to be compared are carried out standardization,, only reflect the geometric shape of curve so that the coupler-point curve data are eliminated the influence of factors such as speed, size;

B. divide equivalence class: to handling the standardized curve that the back forms through step a, wringing number and the distance right with the son section are index, with all son sections to being divided into several equivalence classes, thereby reduce the hunting zone of following steps c;

C. son section is to a rigid body translation: in each equivalence class, try to achieve the son section that belongs to different curves between rigid body translation, right to seek optimum matching section, thus finish the part coupling of curve.

D. cluster analysis: all rigid body translations of gained among the step c are carried out cluster analysis, get the optimum matching that maximum cluster is a coupler-point curve, release best rigid transformation, matching degree and matching error, and show matching effect with this.

Coupler-point curve of the present invention represents that with the discrete point sequence each discrete point is referred to as break.Further, in step a, described standardization is meant coupler-point curve by arc length normalization, promptly total arc length of coupler-point curve to be compared all amplified in proportion or narrows down to same value L (dimensionless), and the coordinate respective change of each break forms new sequence simultaneously.Then, this sequence is divided into N part by arc length, every segment length is L/N, forms the standardization sequence that only reflects the coupler-point curve shape new, that get rid of velocity information.

In step b, described son section is meant, behind step a, on the same coupler-point curve by 2k+1 the formed sub-curve of break continuously; Son section is to being meant, behind step a, and two son sections of same coupler-point curve.Son section to distance be meant the Euclidean distance of two breaks in the middle of two son sections, i.e. distance between k break.The right wringing number of son section can be drawn by following formula:

$\mathrm{Wr}=\frac{1}{4\mathrm{\π}}\underset{i}{\mathrm{\Σ}}\underset{j}{\mathrm{\Σ}}{\mathrm{sg}}_{\mathrm{ij}}{\mathrm{gS}}_{\mathrm{ij}}$

In the formula:

Wr is the wringing number;

Sg _Ij=sign[(A _I+1-A _i) * (A _J+1-A _j) g (A _i-A _j)], A wherein _i, A _jBe the break on the son section.

S _IjBe A _iA _I+1, A _jA _J+1The tetragonal area of formed sphere, its meaning can be referring to " mathematics of fake " (Jiang Baiju work, Hunan education publishing house, 1991).

Described in the step b with all son sections to being divided into several equivalence classes, specifically may further comprise the steps:

B1. calculate on two coupler-point curves to be compared right wringing numbers of all son sections and son section to distance;

B2. find the minimum and maximum value in the wringing number, and be divided into N interval, by each son section to each son section of big young pathbreaker of wringing number to putting into different interval pairing N son section pair sets;

B3. remove and do not comprise the right null set of any son section;

B4. in the son section pair set of each non-NULL, the maximum and the minimum value that find the son section to adjust the distance, and be divided into N interval, each son section of big young pathbreaker of adjusting the distance by each son section is to putting into different interval pairing N son section pair sets, and the removing null set, the set of Xing Chenging is smaller or equal to N * N like this.

So far, with son section to wringing number and distance respectively approximately equal be relation of equivalence, be divided into the equivalence class of number smaller or equal to N * N.

In step c, described rigid body translation is meant rotation and translation, adopt the unit quaternion method to realize, can be referring to document (Berthold K.P.Horn.Closed-form solution of absoluteorientation using unit quaternions.Journal of the Optical Society ofAmerica A, Vol.4, page 629, April 1987), described optimum matching is meant matching error RMSD (Root Mean-Square Deviation, mean square deviation) minimum, RMSD has definition in above-mentioned document.Each son section is to there being the individual point of 2 (2k+1).Matching error has definition in above-mentioned document.

In steps d, the rigid body translation matrix comprises 3 independent variables of rotation matrix and 3 independent variables of translation vector, because the order of magnitude of 6 variablees may differ bigger, so be that object adopts k-means (k-mean cluster method) method cluster analysis to rigid body translation with the rotational component earlier, on this basis, be that object adopts the k-means method to carry out cluster analysis to rigid body translation with the translational component to each cluster again; Described maximum cluster is meant the maximum cluster of the pairing different break number M of rigid body translation in cluster; The matching degree of curve is: $\mathrm{Sim}=\frac{M}{N}\×100\%;$ The matching error expression formula is: $\mathrm{RMSD}=\sqrt{\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{\left|\right|a_i^{\′}-b_i\left|\right|}^2/N},$ Wherein,

The coordinate that the coupling break of expression article one curve shines upon by rigid body translation.b _iExpression and break Coupling break coordinate on the pairing second curve, the implication of N are total number of break after the coupler-point curve standardization as previously mentioned.

Benefit of the present invention is, utilize the wringing number that can reflect the space of curves state in the knot theory, the son section of curve is divided, thereby reduced the search volume, solved the matching problem of space connecting-rod curve effectively, its time and space complexity can be controlled in O (n ²).The present invention can make in the Mechanism Optimization design, searching by two-dimensional plane of existing coupler-point curve collection of illustrative plates expands three dimensions to, thereby when designing in space mechanism, in the time of need satisfying the special exercise track, can seek the highest curve of matching degree by searching the coupler-point curve database, to finish the design of mechanism parameter, this can accelerate the design rate of space mechanism greatly, alleviates workload.In addition, the present invention also can be applicable to the identification of body outline in the computer vision, the comparison of protein steric structure similarity in the molecular biology, even the matching problem of any three-dimensional space curve.

Description of drawings

Fig. 1 is the inventive method general flow chart.

Fig. 2 is the spatiality synoptic diagram before coupler-point curve a and the coupler-point curve b conversion.

Fig. 3 is the spatiality synoptic diagram after coupler-point curve a and the coupler-point curve b conversion.

Fig. 4 is the spatiality synoptic diagram before coupler-point curve a and the coupler-point curve c conversion.

Fig. 5 is the spatiality synoptic diagram after coupler-point curve a and the coupler-point curve c conversion.

Embodiment

Below in conjunction with accompanying drawing embodiments of the invention are elaborated: present embodiment has provided detailed embodiment and process being to implement under the prerequisite with the technical solution of the present invention, but protection scope of the present invention is not limited to following embodiment.

As shown in Figure 1, the embodiment of present embodiment is:

1) coupler-point curve is by one group of coordinate point sequence approximate representation, and the most original coordinate point sequence (be designated as SCO[n]) be to obtain by the ADAMS simulation software.Promptly at first with coupler-point curve by arc length normalization, promptly the arc length of all coupler-point curves to be compared is all amplified in proportion or narrows down to same length L (dimensionless), SCO[n] in the coordinate of each point also change.Then, with SCO[n] sequence is divided into N part by arc length, every segment length is L/N, forms new sequence SC[N], get n=500 herein, N=180.Realize by the C# language programming.

2) any coupler-point curve can be used point sequence { p ₁, p ₂..., p _NReplace, to any one long son section { p wherein for 2k+1 _I-k..., p _i..., p _I+kBe designated as

In computing machine with p _iThe whole son section of some expression, and The expression with

Sequence { the p of reversed in order _I+k..., p _i..., p _I-k.Coupler-point curve C ₁It is right to go up any son section

With

Between distance be a _iAnd a _jDistance; In like manner, to coupler-point curve C ₂It is right to go up any son section Find out the maximal value Wr of wringing number _MaxAnd Wr _Min, then each interval size is (Wr _Max-Wr _Min)/N is right with each height section

Deposit among the interval pairing ArrayList type stores space ALO N element in; In each non-NULL storage space, again according to the intersegmental distance of son will be wherein the son section to depositing among the new ArrayList type stores space AL N * N element in; Remove empty project.In AL, curve C ₁And C ₂On son section to storage respectively so that next step try to achieve the son section to rigid body translation.That is, comprise several equivalence class elements under the AL, each equivalence class element comprises two element a and b, and the experimental process section that comprises among a, the b is to respectively from curve C ₁And C ₂

3) in each equivalence class of AL, each height section of getting among the element a is right

Adopt the unit quaternion method, select all son sections to mating successively with the right inverted sequence sequence of each son section in b, promptly try to achieve rigid body translation and matching error, the son section of choosing the matching error minimum is to being

Match objects

Utilization the method, all the son sections in all equivalence classes are to all finding match objects.Coupling section with rotational component, translational component and correspondence is right

With

Be identity element, deposit among the ArrayList type stores space ALp.

4) at first be the cluster object with the rotational component, the element among the ALp is carried out cluster analysis, choose the k-means clustering method, k gets 3; In each existing cluster, be the cluster object with the translational component, carry out the k-means cluster analysis, k gets 2.

5) in each cluster, try to achieve the number and the C of the right difference coupling break of all son sections ₁And C ₂Corresponding relation on the curve between break, getting the maximum cluster of coupling number is required cluster, the break number M of this cluster, break corresponding relation and this cluster centre point are institute and ask.Similarity is $\mathrm{Sim}=\frac{M}{N}\×100\%,$ Matching error is $\mathrm{RMSD}=\sqrt{\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{\left|\right|a_i^{\′}-b_i\left|\right|}^2/N},$ Wherein Zi Mu meaning as previously mentioned.

6) in Maple software, read in the rigid body translation of curvilinear coordinates and step 5) gained, draw respectively show before the conversion with conversion after curvilinear figure.

By the matching result of accompanying drawing to two groups of space connecting-rod curves of present embodiment, three space connecting-rod curves having selected the generation of space 7R mechanism mate, curve a, b, c, wherein a and b are more once, a and c are more once, by Fig. 2 and Fig. 4 as can be seen the similarity of a and b obviously greater than the similarity of a and c, and this point can be described equally with the result of present embodiment gained, Fig. 3 and Fig. 5 are respectively curve a and b, effect after the rigid body translation mapping of a and c process present embodiment gained, the similarity of a and b is 96.1%, error is 0.79, the similarity of a and c is 42.8%, and error is 2.97.This has also verified correctness of the present invention and practicality to a certain extent.Present embodiment not only can be applicable to space connecting-rod curve similarity determination problem, also extends to the coupling field of space arbitrary curve.

Claims (7)

1. the detection method of a three-D connection rod curve matching rate is characterized in that, comprises following steps:

A. data normalization: coupler-point curve data to be compared are carried out standardization;

B. divide equivalence class: to handling the standardized curve that the back forms through step a, wringing number and the distance right with the son section are index, with all son sections to being divided into several equivalence classes; Described son section is meant: behind step a, on the same coupler-point curve by 2k+1 the formed sub-curve of break continuously; Described son section is to being meant: behind step a, and two son sections of same coupler-point curve; Son section to distance be meant the Euclidean distance of two breaks in the middle of two son sections;

C. son section is to a rigid body translation: in each equivalence class, try to achieve the son section that belongs to different curves between rigid body translation, right to seek optimum matching section, thus finish the part coupling of curve;

D. cluster analysis: all rigid body translations of gained among the step c are carried out cluster analysis, get the optimum matching that maximum cluster is a coupler-point curve, release best rigid transformation, matching degree and matching error, and show matching effect with this.

2. the detection method of three-D connection rod curve matching rate according to claim 1 is characterized in that, described coupler-point curve is represented with the discrete point sequence.

3. the detection method of three-D connection rod curve matching rate according to claim 1, it is characterized in that, among the step a, described standardization is meant coupler-point curve by arc length normalization, promptly total arc length of coupler-point curve to be compared is all amplified in proportion or narrow down to same value L, L is a dimensionless, the coordinate respective change of each break forms new sequence simultaneously, then, this sequence is divided into N part by arc length, every segment length is L/N, forms the standardization sequence that only reflects the coupler-point curve shape new, that get rid of velocity information.

4. the detection method of three-D connection rod curve matching rate according to claim 1 is characterized in that, among the step b, described with all son sections to being divided into several equivalence classes, may further comprise the steps:

B1. calculate on two coupler-point curves to be compared right wringing numbers of all son sections and son section to distance;

B2. find the minimum and maximum value in the wringing number, and be divided into N interval, by each son section to each son section of big young pathbreaker of wringing number to putting into different interval pairing N son section pair sets;

B3. remove and do not comprise the right null set of any son section;

B4. in the son section pair set of each non-NULL, the maximum and the minimum value that find the son section to adjust the distance, and be divided into N interval, each son section of big young pathbreaker of adjusting the distance by each son section is to putting into different interval pairing N son section pair sets, and the removing null set, the set of Xing Chenging is smaller or equal to N * N like this;

So far, the son section to equating to be relation of equivalence respectively with wringing number and distance, is divided into the equivalence class of number smaller or equal to N * N.

5. the detection method of three-D connection rod curve matching rate according to claim 1 is characterized in that, among the step c, described rigid body translation is meant rotation and translation, adopts the unit quaternion method to realize.

6. the detection method of three-D connection rod curve matching rate according to claim 1 is characterized in that, in the steps d, described maximum cluster is meant: the maximum cluster of the pairing different break number M of rigid body translation in cluster;

Described matching degree is: $\mathrm{Sim}=\frac{M}{N}\×100\%,$ N is total number of break after the coupler-point curve standardization.

7. the detection method of three-D connection rod curve matching rate according to claim 1 is characterized in that, in the steps d, and described matching error, its expression formula is:

$\mathrm{RMSD}=\sqrt{\underset{l=1}{\overset{M}{\mathrm{\Σ}}}{\left|\right|a_l^{\′}-b_l\left|\right|}^2/N},$

Wherein,

The coordinate that the coupling break of expression article one curve shines upon by rigid body translation, b _iExpression and break Coupling break coordinate on the pairing second curve, N are total number of break after the coupler-point curve standardization, and M is the number of coupling break.

Images