CN101315665A - Identity recognition method based on three-dimensional nose shape - Google Patents

Abstract

The invention relates to an identification method, in particular to an identification method based on three-dimensional nose shape. The method comprises the steps of: 1) the preprocessing of the curved surface of a three-dimensional face; 2) the cutting and extraction of the nose area of a three-dimensional face model; 3) the matching of the nose shape on the basis of step 2); 4) the execution of the identification of the nose shape by utilizing the matching result of step 3). The method has the advantages at two aspects by utilizing the three-dimensional nose shape to carry out identification: firstly, in the aspect of acquisition, the 3D acquiring technology can conveniently acquire 3D face data while the acquisition of 3D nose data is more convenient than the acquisition of the whole face data; secondly, the 3D face can generate distortion due to the influence of expression while the nose area can keep relatively stable shape under different expressions, therefore, compared with the 3D face identification, the identification based on the nose shape is expected to obtain better identification performance.

Description

A kind of personal identification method based on three-dimensional nose shape

Technical field

The present invention relates to a kind of personal identification method, refer in particular to a kind of personal identification method based on three-dimensional nose shape.

Background technology

The Automatic face recognition technology is with a wide range of applications in fields such as national security, military security, public safety and home entertainings, and in the past few decades, recognition of face obtains going deep into extensive studies.Yet, still face great challenge based on the two-dimension human face recognition technology of image, under the situation of light, attitude and expression shape change, the accuracy of two-dimension human face identification also far away can not be satisfactory.

The three-dimensional face recognition technology is expected to fundamentally solve the difficult problem based on " being subjected to attitude, light and expression influence " that face identification method faced of image.Based on the prerequisite that three-dimensional data has been obtained, three-dimensional face identification is subjected to the influence of light very little.Because three-dimensional data has explicit geometric configuration, three-dimensional face identification has more and overcomes the potentiality that attitude changes.Yet expression shape change has changed the shape of three-dimensional face model, causes the plastic yield of people's face regional area.The torsional deformation of people's face will directly influence the performance of face recognition algorithms.Obtain also more complicated of whole people's face data in addition.

Summary of the invention

In order to overcome the above problems, the invention provides more simple and conveniently, and be not subjected to a kind of personal identification method of influence of human face expression based on three-dimensional nose shape.

A kind of personal identification method based on three-dimensional nose shape comprises the steps:

1) pre-service of three-dimensional face curved surface;

2) cutting of three-dimensional face model nasal area is extracted;

3) in step 2) the basis on, the coupling of nose shape;

4) utilize the matching result of step 3) to carry out the identification of nose shape.Further,

Described step 1) is handled the three-dimensional local facial features that obtains noiseless, feature reservation with original three-dimensional face data by trigonometric ratio curve reestablishing, the cutting of people's face regional area and bilateral smoothing denoising;

Described step 2) detect by the plane of symmetry and the silhouette lines extraction, and the detection of prenasale and nose basic point, three-dimensional face model is placed a unified coordinate frame, utilize prenasale, nose basic point that the nasal area cutting is extracted;

Described step 3) utilizes iteration conjugate points matching algorithm to calculate the three-dimensional nose shape of model in the three-dimensional nose shape of model to be detected and the picture library, mates the similarity of the right mean distance of closest approach between latter two three-dimensional nose shape as both;

Described step 4) is to each the model implementation step 3 in the picture library) calculate, one that chooses mean distance minimum wherein as recognition result, finishes systemic-function.

Further, described step 2) the plane of symmetry detects and adopts iteration conjugate points matching algorithm alignment master pattern and its mirror image model, and right middle axial plane is the required plane of symmetry to obtain conjugate points then.

Further, step 2) following method is adopted in the detection of prenasale and nose basic point:

p _nt＝argmax _p∈C(dist(p，l _e)) (4)

p _nb＝arg?min _p∈L(y _p) (5)

$L=\left\{p\right|p\∈C,y_p>y_{p_{\mathrm{nt}}},{\mathrm{dist}}^{\′}(p,l_e)=0\}---\left(6\right)$

Wherein, p _NtBe prenasale, p _NbBe the nose basic point, C is a silhouette lines, and connection silhouette lines C 2 line segment end to end is l _e, y _pThe y axial coordinate of expression point p, dist (p, l _e) represent that some p is to straight-line segment l _eDistance, dist ' (p, l _e) represent that some p is to straight-line segment l _eThe single order differential of distance.

Further, described step 2) cutting in the three-dimensional nose zone of Cai Yonging is extracted and to be based on the nose basic point and the well-determined spheroid of prenasale is finished;

Further, the thin alignment based on iteration conjugate points matching algorithm is in step 2 in the described step 3)) the basis of thick alignment on finish; The matching similarity amount that adopts as shown in the formula:

$\mathrm{Dis}(M_p,M_g)=\mathrm{RMS}({\mathrm{ICP}}_{M_g}\left(M_p\right),M_g)---\left(26\right)$

Wherein RMS () represents the closest approach mean distance, $M_p^{\′}={\mathrm{ICP}}_{M_g}\left(M_p\right)$ Expression uses iteration conjugate points matching algorithm with M _pTo M _gAlignment obtains M _p'.

Below the invention will be further described:

A kind of personal identification method based on three-dimensional nose shape, its step is as follows:

1) pre-service of three-dimensional face curved surface: at first original three-dimensional face data are carried out the trigonometric ratio curve reestablishing, then the regional area of people's face is cut apart, with the method for bilateral filtering the people's face regional area that is partitioned into is carried out the denoising that feature keeps at last;

2) cutting of three-dimensional face model nasal area:

Determine human face posture three-dimensional model to be placed unified coordinate frame by the plane of symmetry and two unique points (prenasale and nose basic point) that detect three-dimensional model, realized thick alignment, for matching stage provides reasonable initial alignment position;

What the plane of symmetry detected employing is iteration conjugate points matching algorithm alignment master pattern and its mirror image model, asks the method for the middle axial plane of corresponding point again.

Following method is adopted in the detection of prenasale and nose basic point:

p _nt＝argmax _p∈C(dist(p，l _e)) (4)

p _nb＝arg?min _p∈L(y _p) (5)

$L=\left\{p\right|p\∈C,y_p>y_{p_{\mathrm{nt}}},{\mathrm{dist}}^{\′}(p,l_e)=0\}---\left(6\right)$

Wherein, p _NtBe prenasale, p _NbBe the nose basic point, C is a silhouette lines, and connection silhouette lines C 2 line segment end to end is l _e, y _pThe y axial coordinate of expression point p, dist (p, l _e) represent that some p is to straight-line segment l _eDistance, dist ' (p, l _e) represent that some p is to straight-line segment l _eHuge from the single order differential.

By detected prenasale and nose basic point, the nasal area cutting of three-dimensional face model is extracted.Utilize prenasale, nose basic point and the plane of symmetry that the three-dimensional nose shape that extracts has been placed in the unified coordinate frame, thereby realized the thick alignment of three-dimensional nose model;

3) coupling: utilize the three-dimensional nose shape of model in the three-dimensional nose shape of iteration conjugate points algorithmic match model to be detected and the storehouse, calculate the similarity of the right mean distance of closest approach between latter two three-dimensional nose shape of coupling as both;

The matching similarity amount that step 3) adopts as shown in the formula:

$\mathrm{Dis}(M_p,M_g)=\mathrm{RMS}({\mathrm{ICP}}_{M_g}\left(M_p\right),M_g)---\left(26\right)$

Wherein RMS () represents the closest approach mean distance, $M_p^{\′}={\mathrm{ICP}}_{M_g}\left(M_p\right)$ Expression uses iteration conjugate points matching algorithm with M _pTo M _gAlignment obtains M _p'.

4) identification: each model among the gallery is applied the calculating of step 3), and one that chooses mean distance minimum wherein as recognition result, to finish systemic-function.

Beneficial effect of the present invention is: nose is subjected to expression influence very little as the more stable organ of face's shape.Utilize three-dimensional nose shape to carry out the advantage that identification mainly contains two aspects, at first, aspect obtaining, 3D obtains technology can obtain 3D people's face data easily, and to obtain whole people's face data more convenient and obtain 3D nose data ratio; The second, therefore 3D people's face distortion because the influence of being expressed one's feelings can be distorted, and nasal area is the shape that can keep relative stability under the difference expression, with 3D recognition of face comparison, is expected to obtain the better recognition performance based on the identification of nose shape.

Description of drawings

Fig. 1 is the process flow diagram of the personal identification method based on three-dimensional nose shape of the present invention;

Fig. 2 is an original three-dimensional face data trigonometric ratio curve reestablishing design sketch of the present invention;

Fig. 3 is people's face regional area cutting effect figure of the present invention;

Fig. 4 is the bilateral smoothing denoising treatment effect figure that people's face local features of the present invention keeps;

Fig. 5 is prenasale and a nose basic point detection synoptic diagram on the silhouette lines of the present invention;

Fig. 6 is that synoptic diagram is extracted in three-dimensional nose of the present invention zone cutting;

Fig. 7 is a ROC experimental result synoptic diagram of the present invention;

Fig. 8 is a CMC experimental result synoptic diagram of the present invention;

Embodiment

Below in conjunction with accompanying drawing the present invention is elaborated.

1) three-dimensional face model pre-service

The three-dimensional face model pre-service was finished by three steps, at first carried out the trigonometric ratio curve reestablishing, and next cuts people's face regional area, handled the three-dimensional local facial features that obtains noiseless, feature reservation by bilateral smoothing denoising at last.

(1). the trigonometric ratio curve reestablishing

The raw data of FRGC3D face database is the set of 3D coordinate points, because subsequent treatment requires to set up the topological relation of point set, promptly finishes the curve reestablishing process of point set, and the piecewise smooth surface reconstruction algorithm of employing HOPPE94 is finished the trigonometric ratio of point set, as Fig. 2.

(2). the cutting of people's face regional area

3D people's face data of existing 3D data acquisition facility collection comprise zones such as neck, shoulder usually.Raw data major part as the FRGC1.03D face database has comprised shoulder and a small amount of chest region, shown in Fig. 3 first row.The purpose of people's face cutting is that raw data is carried out a rough cutting to extract face area, convenient follow-up processing.The present invention adopts the cutting method based on curvature.Because curvature is the distinctive attribute of curved surface itself, it changes for the angle of people's face, and change in location can remain unchanged.Therefore pass through the analysis to the distribution of 3D data overall curvature value, the present invention has realized the automatic segmented extraction of face area, as second row of Fig. 3.

(3). the smoothing denoising that feature keeps is handled

The 3D faceform comprises noise to a certain degree and some are by gathering the point at random that produces, as Fig. 4 first row.The present invention adopts bilateral denoising filter that it is carried out the denoising of keeping characteristics, as Fig. 4 second row.

2) cutting of three-dimensional face model nasal area

Finished by three steps the attitude location, at first detects the plane of symmetry and extract silhouette lines, determines prenasale and nose basic point then, uses a rigid transformation at last and place unified coordinate frame to finish the attitude location three-dimensional face model.

(1) plane of symmetry detects and the silhouette lines extraction

The vertex set V of given people's face grid M _M={ p _i∈ R ³| 1≤i≤N}, to any one plane, can find V _MMirror image vertex set about this plane $V_M^m=\{p_i^m\&Element;{\mathrm{<figure-callout\; id="3"\; label="R"\; filenames="A20081006358600151.png"\; state="\{\{state\}\}">R</figure-callout>}}^3|1\&le;i\&le;N\},$ V _MIn any 1 p _iCorresponding mirror point be V _M ^mIn p _i ^mThe topological structure of protoplast's face grid is consistent with the mirror image grid, say on the stricti jurise, if consider the direction of patch, the summit preface of the tri patch of mirror image grid should be opposite with former grid M, can obtain unified curved surface direction this moment, because the present invention here mainly analyzes point set, therefore can ignore the influence of topologies change.

With V _M ^mTo V _MRegistration finally aligns both, supposes that the point set that obtains is $V_M^{m^{\&prime;}}=\{p_i^{m^{\&prime;}}\&Element;{\mathrm{<figure-callout\; id="3"\; label="R"\; filenames="A20081006358600151.png"\; state="\{\{state\}\}">R</figure-callout>}}^3|1\&le;i\&le;N\},$ The corresponding order of its mid point is still constant.At this moment, V _MAnd V _M ^mNew vertex set V of ' composition:

$\stackrel{\&OverBar;}{V}=V_M+V_M^{m^{\&prime;}}---\left(1\right)$

Because three-dimensional face itself is symmetry roughly, V is a set from symmetry, and the plane of symmetry of people's face grid must be crossed V _MAnd V _M ^mThe bisector of ' middle corresponding point, so the point set that accumulates in the plane of symmetry A of people's face can be represented with following formula:

$A=\left\{x\right|<x-(p_i+p_i^{m^{\&prime;}})/2,p_i-p_i^{m^{\&prime;}}>=\mathrm{0,1}\&le;i\&le;N\}---\left(2\right)$

Wherein＜, the expression two vectors dot product.

The present invention adopts iteration conjugate points matching algorithm alignment V _MAnd V _M ^m, the iteration conjugate points matching algorithm three-dimensional model that can align effectively, but guarantee the initial position that two models to be alignd of its convergent requirement have gross alignment.Therefore, when calculating the mirror image grid of protoplast's face grid M, need select symmetrical plane carefully.If the symmetrical plane of initial selected itself just near the plane of symmetry of people's face grid, the mirror image point set V that looks like to obtain with this level crossing then _M ^mAnd V _MJust have initial alignment position preferably.

The basic configuration of observer's face grid, finder's face grid of the present invention is that an above-below direction is longer, and left and right sides span is placed in the middle, and the curved surface that front and back thickness is less is similar to the shape of semielliptical.Therefore, the present invention is to the point set V of people's face grid _MPivot analysis (PCA) is done in distribution, can obtain an equalization point p and three principal directions (proper vector) v ₁, v ₂And v ₃, corresponding respectively three eigenwerts of ordering from big to small of these three proper vectors, the relation of pressing the eigen vector of PCA, v ₁Direction is the direction of point set divergence maximum, v ₂Secondly, v ₃Minimum, its variance are three character pair values.Thus, it is as follows that the present invention can select initial symmetrical plane:

Mirror＝{x|＜x-p，v ₂＞＝0} (3)

This initial plane of mirror symmetry satisfies near the requirement the plane of symmetry of primitive man's face.

Because V _MAnd V _M ^m' N is arranged to symmetric points, each be to can determining the symmetrical plane of people's face grid, the present invention is with each plane of least square fitting, the symmetrical plane A that is optimized at last, silhouette lines ask for the friendship that only needs calculating symmetrical plane A and original mesh M.

(2) prenasale and nose basic point determines

Prenasale p _NtWith nose basic point p _NbAll on silhouette lines C, suppose to connect silhouette lines C end to end 2 line segment be l _e, as shown in Figure 5, can draw following two hypothesis to the observation of a large amount of face characteristics:

A) prenasale p _NtBe to be positioned on the silhouette lines C, apart from line segment l _ePoint farthest;

B) nose basic point p _NbBe to be positioned on the silhouette lines C, along p _NtIn having a few on the silhouette lines upwards with line segment l _eFirst apart from minimum point.

Based on two top hypothesis, can list following nose and nose base detection method:

p _nt＝argmax _p∈C(dist(p，l _e)) (4)

p _nb＝argmin _p∈L(y _p) (5)

$L=\left\{p\right|p\&Element;C,y_p>y_{p_{\mathrm{nt}}},{\mathrm{dist}}^{\&prime;}(p,l_e)=0\}---\left(6\right)$

Wherein,

The expression point is to the distance of straight-line segment, y _pThe y axial coordinate of expression point p,

The expression point is to the single order differential of straight-line segment distance.

(3) unify coordinate frame

The present invention has obtained the plane of symmetry direction d of people's face grid _sWith two unique point p _NtAnd p _Nb, can be placed into people's face grid in the unified coordinate system definite to finish attitude by these three features.Order:

v _x＝d _s (7)

$v_y=\frac{p_{\mathrm{nb}}-p_{\mathrm{nt}}}{\left|\right|p_{\mathrm{nb}}-p_{\mathrm{nt}}\left|\right|}---\left(8\right)$

v _z＝v _x×v _y (9)

With p _NtBe initial point, v _x, v _y, v _zBe respectively x, y, three coordinate axis of z can be determined a new coordinate frame, all three-dimensional face models can transform in this coordinate frame.

By detected prenasale p _NtWith nose basic point p _Nb, the nasal area cutting of three-dimensional face model is extracted.With p _NtAs the spheroid center, with p _NbConstitute spheroid E as spheroid major axis one end points: $\frac{x^2}{a^2}+\frac{y^2}{{\mathrm{<figure-callout\; id="2"\; label="b"\; filenames="A20081006358600151.png"\; state="\{\{state\}\}">b</figure-callout>}}^2}+\frac{z^2}{c^2}=1,$ (being nasal area) segmented extraction in the spheroid is gone out, as Fig. 6.Because the three-dimensional nose shape that extracts has been placed in the unified coordinate frame, the three-dimensional nose model has a reasonable initial position through thick alignment.

3) three-dimensional nose Model Matching

The M of given three-dimensional nose model _pAnd M _g, with iteration conjugate points algorithmic match M _pAnd M _g, both similarities are calculated as follows:

$\mathrm{Dis}(M_p,M_g)=\mathrm{RMS}({\mathrm{ICP}}_{M_g}\left(M_p\right),M_g)---\left(26\right)$

Wherein RMS () represents the closest approach mean distance, $M_p^{\&prime;}={\mathrm{ICP}}_{M_g}\left(M_p\right)$ Expression uses iteration conjugate points matching algorithm with M _pTo M _gAlignment obtains M _p'.

4) three-dimensional nose Model Identification

A given three-dimensional nose model M to be identified _p, according to the matching algorithm that proposes in the step 3) each the three-dimensional nose model in itself and the picture library being mated respectively, the model of selecting matching result distance value minimum is as final recognition result.

Experimental result

The present invention has tested the performance based on the personal identification method of three-dimensional nose shape on FRGC v1.0 storehouse.Have 276 people in this storehouse, 943 faceforms.Wherein 198 people have the image more than.Master pattern is made up of about 20000 to 110000 points.Experimental selection 233 models, comprise 100 people.ROC and CMC experimental result picture are seen Fig. 7 and Fig. 8 respectively.

Claims (6)

1. the personal identification method based on three-dimensional nose shape is characterized in that this method comprises the steps:

1) pre-service of three-dimensional face curved surface;

2) cutting of three-dimensional face model nasal area is extracted;

3) in step 2) the basis on, the coupling of nose shape;

4) utilize the matching result of step 3) to carry out the identification of nose shape.

2. the personal identification method based on three-dimensional nose shape according to claim 1 is characterized in that:

Described step 1) is handled the three-dimensional local facial features that obtains noiseless, feature reservation with original three-dimensional face data by trigonometric ratio curve reestablishing, the cutting of people's face regional area and bilateral smoothing denoising;

Described step 2) detect by the plane of symmetry and the silhouette lines extraction, and the detection of prenasale and nose basic point, three-dimensional face model is placed a unified coordinate frame, utilize prenasale, nose basic point that the nasal area cutting is extracted;

Described step 3) utilizes iteration conjugate points matching algorithm to calculate the three-dimensional nose shape of model in the three-dimensional nose shape of model to be detected and the picture library, mates the similarity of the right mean distance of closest approach between latter two three-dimensional nose shape as both;

Described step 4) is to each the model implementation step 3 in the picture library) calculate, one that chooses mean distance minimum wherein as recognition result, finishes systemic-function.

3. the personal identification method based on three-dimensional nose shape according to claim 2 is characterized in that: the plane of symmetry described step 2) detects and adopts iteration conjugate points matching algorithm alignment master pattern and its mirror image model, obtains the right middle axial plane of conjugate points then.

4. the personal identification method based on three-dimensional nose shape according to claim 2 is characterized in that: step 2) detection of prenasale and nose basic point adopts following method:

p _nt＝argmax _p∈C(dist(p，l _e)) (4)

p _nb＝argmin _p∈L(y _p) (5)

$L=\left\{p\right|p\&Element;C,y_p>y_{p_{\mathrm{nt}}},{\mathrm{dist}}^{\&prime;}(p,l_e)=0\}---\left(6\right)$

Wherein, p _NtBe prenasale, p _NbBe the nose basic point, C is a silhouette lines, and connection silhouette lines C 2 line segment end to end is l _e, y _pThe y axial coordinate of expression point p, dist (p, l _e) represent that some p is to straight-line segment l _eDistance, dist ' (p, l _e) represent that some p is to straight-line segment l _eThe single order differential of distance.

5. the personal identification method based on three-dimensional nose shape according to claim 1 and 2 is characterized in that: described step 2) cutting in the three-dimensional nose zone of Cai Yonging is extracted and to be based on the nose basic point and the well-determined spheroid of prenasale is finished.

6. the personal identification method based on three-dimensional nose shape according to claim 1 is characterized in that: described step

3) the thin alignment based on iteration conjugate points matching algorithm in is in step 2) the basis of thick alignment on finish; The matching similarity amount that adopts as shown in the formula:

$\mathrm{Dis}(M_p,M_g)=\mathrm{RMS}({\mathrm{ICP}}_{M_g}\left(M_p\right),M_g)$ (26)

Wherein RMS () represents the closest approach mean distance, ${M_p}^{\&prime;}={\mathrm{ICP}}_{M_g}\left(M_p\right)$ Expression uses iteration conjugate points matching algorithm with M _pTo M _gAlignment obtains M _p'.

Images