CN104951767A - Three-dimensional face recognition technology based on correlation degree - Google Patents

Abstract

The invention discloses a three-dimensional face recognition technology based on the correlation degree. The three-dimensional face recognition technology comprises steps as follows: firstly, three-dimensional face geometric features are extracted, the distance between two eyes is taken as measuring basis, and each of other features is acquired according to the proportion of the corresponding body part to the distance between the eyes; three-dimensional geometric features measurement is performed, a sample is selected, three-dimensional geometric features of the sample are measured and subjected to normalization processing to generate a feature vector, then the feature vector is compared with a normalized feature vector in a standard sample database, the similarity is calculated, and three-dimensional face recognition judgment is finished. The three-dimensional face recognition technology based on the correlation degree has the benefits as follows: the recognition similarity is higher, and the recognition effect is good when the technology is applied to a three-dimensional face model.

Description

Based on the three-dimensional face recognition technology of correlation degree

Technical field

The invention belongs to technical field of face recognition, relate to the three-dimensional face recognition technology based on correlation degree.

Background technology

In recent years, along with the development gradually of science and technology, people to the requirement increasingly stringent of information security, human body identity recognizing technology also development thereupon.Because face is unique and unique, become the key of human body identification.Two-dimension human face recognition technology is widely used, but it is easily subject to the impact of the factors such as illumination, shooting angle and human face expression change, and accuracy of identification is relatively low.Current, along with the deep development of computer stereo vision technique, face dimensional Modeling Technology have also been obtained very large improvement, and the extraneous factor in two-dimension picture of can effectively avoiding of three-dimensional face model, to identifying the interference that accuracy causes, strengthens recognition accuracy greatly.

Three-dimensional face identification is the emphasis problem of relevant scholar's research.Current conventional face recognition technology mainly can be divided into several types: the method based on geometric properties, the method based on template, method based on model.Wherein, the method based on geometric properties is the most conventional.But the method need be combined with other algorithm and just can reach gratifying effect; Method based on template can be divided into again method, linear discriminant analysis method, neural net method, Dynamic link library matching process etc. based on relevant matches; Based on the method for model then primarily of based on hidden Markov model, based on active shape model and the composition such as method based on active appearance models.

Now, certain achievement is also achieved to the research of three-dimensional face identification.Such as, Gordon can obtain the curve distribution schematic diagram on face surface by the depth data of face, is obtained the geometric properties at each position of face by the face curvature distribution obtained; Beumier carries out light projection to face, thus obtains the three-dimensional data of face, is gathered by the intensity profile figure at the contour curve of face and other positions of face, regards a recognition feature in later stage as.Carry out coupling by the three-dimensional face model in this characteristic sum storehouse again and complete identification.

Analyzing three-dimensional Face geometric eigenvector recognizer finds, by when model mates in the three-dimensional geometry characteristic sum storehouse on the three-dimensional face model of acquisition, there is trueness error.So, usually adopt the method for similarity-rough set to carry out characteristic matching.But the method also has certain drawback, it cannot embody the criticality of face local feature change and the contingency of human face expression change.For above-mentioned drawback, a kind of three-dimensional face recognition algorithm calculated based on correlation degree is proposed herein.It introduces Probabilistic Decision-making, is described by the correlation degree of probability to localized variation, thus strengthens the accuracy identified.

Summary of the invention

The object of the present invention is to provide the three-dimensional face recognition technology based on correlation degree, solve existing three-dimensional face identification method and there is trueness error problem.

The technical solution adopted in the present invention is carried out according to following steps:

Step 1: the extraction of three-dimensional face geometric properties;

Distance between two is regarded as the benchmark of measurement, other features obtain according to the ratio of the spacing of itself and eyes:

(1) suppose that the distance between two is F1: the European geometric distance between endocanthion point and outer canthus point;

(2) the European geometric distance between distance F2: two endocanthion points between people's intraocular part;

(3) the European geometric distance between distance F3: two outer canthus points between human eye outside;

(4) the height F4 of nose: tip of the nose is to the European geometric distance of nose root;

The feature of face significant points angulation

(1) the angle F5 formed between the left wing of nose, nose, right wing of nose three;

(2) angulation F6 between eyespot in the interior eyespot of left eye, nose, right eye;

(3) angulation F7 between the outer eyespot of left eye, nose, the outer eyespot of right eye;

(4) nose and corners of the mouth left and right edge point angulation F8;

Step 2: the measurement of three-dimensional geometry feature

1, European geometric distance is calculated

The coordinate of any two three-dimensional feature points is described as P ₁(x ₁, y ₁, z ₁), P ₂(x ₂, y ₂, z ₂), then the air line distance of the two is the European geometric distance between 2, and formula is described below:

$d=\sqrt{{(x_1-x_2)}^2+{(y_1-y_2)}^2+{(z_1-z_2)}^2};---\left(1\right)$

Namely the eigenwert of F1 ~ F3 is obtained by above formula; Suppose that the coordinate of nose is (x ₁, y ₁, z ₁), the coordinate of nose root is (x ₂, y ₂, z ₂), under nose, the coordinate of root is (x ₃, y ₃, z ₃), then the space line equation that under nose root and nose, root 2 is formed is described as:

Ay+Bz+C＝0；

In formula, A=z ₂-z ₃, B=y ₃-y ₂c=y ₂(z ₃-z ₂)-z ₂(y ₃-y ₂);

Then nose height:

$F4=\frac{|{\mathrm{Ay}}_1+{\mathrm{Bz}}_1+C|}{\sqrt{A^2+B^2}};---\left(2\right)$

Step 3: obtain angle geometric properties

The angle produced between the left wing of nose, nose, right wing of nose three is regarded as the three-dimensional geometry feature of identification:

$F11={\sin }^{-1}\left(\frac{2*S}{L_1*L_2}\right)---\left(3\right)$

In formula, L ₁for describing the distance between d and a; L ₂for describing the distance between e and a; S is for describing above-mentioned leg-of-mutton area, and d represents nose left side flap, and a represents nose, and e represents nose right side flap, and b represents the nasion, and c represents root under nose;

The angle F5 formed between the left wing of nose, nose, right wing of nose three;

$F5={\sin }^{-1}\left(\frac{2*S_2}{L_3*L_4}\right)---\left(4\right)$

In formula, L ₃for describing the distance between the left wing of nose and nose; L ₄for describing the distance between nose and the right wing of nose; S ₂for describing above-mentioned leg-of-mutton area;

Angulation F6 between eyespot in the interior eyespot of left eye, nose, right eye;

$F6={\sin }^{-1}\left(\frac{2*S_3}{L_5*L_6}\right)---\left(5\right)$

In formula, L ₅for describing the distance in left eye between eyespot and nose; L ₆for describing the distance in nose and right eye between eyespot; S ₃for describing above-mentioned leg-of-mutton area;

Angulation F7 between the outer eyespot of left eye, nose, the outer eyespot of right eye;

$F7={\sin }^{-1}\left(\frac{2*S_4}{L_7*L_8}\right)---\left(6\right)$

In formula, L ₇for describing the distance in left eye between eyespot and nose; L ₈for describing the distance between nose and the outer eyespot of right eye; S ₄for describing above-mentioned leg-of-mutton area;

Nose and corners of the mouth left and right edge point angulation F8;

$F8={\sin }^{-1}\left(\frac{2*S_5}{L_9*L_{10}}\right)---\left(7\right)$

In formula, L ₉for describing the distance between the left corners of the mouth and nose; L ₁₀for describing the distance between nose and the right corners of the mouth; S ₅for describing above-mentioned leg-of-mutton area;

Gather the volumentary geometry feature of nose

Obtain the volume of nose:

$F15=\frac{1}{3}*S*F4---\left(8\right)$

In formula, S is for describing the floorage of bottom surface quadrilateral, and F4 is for describing the height of nose; Measure bd, the distance of dc, ce, be, afterwards, be diagonal line with bc, be divided into two triangle bdc and ceb, by sine and the cosine law, obtain two leg-of-mutton areas respectively, they and be the floorage S of required bottom surface quadrilateral;

Step 4: obtain similarity by the three-dimensional geometry feature of above-mentioned acquisition

Choose sample, measure the three-dimensional geometry feature of sample, and be normalized, generating feature vector, afterwards, compares with the normalization characteristic vector in standard sample database, calculates similarity, by analysis, show that calculating formula of similarity is as follows:

$S(I_1,I_2)=\frac{P\left(\mathrm{\Δ}\right|{\mathrm{\Ω}}_i)P\left({\mathrm{\Ω}}_i\right)}{P\left(\mathrm{\Δ}\right|{\mathrm{\Ω}}_i)P\left({\mathrm{\Ω}}_i\right)+P\left(\mathrm{\Δ}\right|{\mathrm{\Ω}}_E)P\left({\mathrm{\Ω}}_E\right)}---\left(9\right)$

In formula, use I respectively ₁with I ₂shape of face image feature amount is described, P (Ω _e) for describing certain condition of work; Ω is belonged to by the feature difference △ of computer memory _ior belong to Ω _eprobability, complete the judgement of three-dimensional face identification, if P (Ω _i| △) >P (Ω _e| △) or S (I ₁, I ₂) >1/2, then think and identify successfully.

The invention has the beneficial effects as follows that the similarity of identification is higher, effective in the identification that the inventive method is applied to human face three-dimensional model.

Accompanying drawing explanation

Fig. 1 is eyes measurement point schematic diagram;

Fig. 2 is nose measurement point schematic diagram;

Fig. 3 is mouth measurement point schematic diagram;

Fig. 4 is nose measurement point schematic diagram in kind.

Embodiment

Below in conjunction with embodiment, the present invention is described in detail.

Face 3-D data collection: the VIVI910 laser scanner that the present invention produces by Minolta company scans, carries out record to the three-dimensional geometric information of face and texture information.Wherein, three-dimensional geometric information carries out record with the form of cylinder coordinate, and texture information carries out record in typical RGB information mode.The pre-service of human face data and segmentation: when carrying out data acquisition, owing to being subject to the interference of the factors such as angle, will cause three-dimensional model to start a leak, so need carry out pre-service to the initial three-dimensional geometrical face model obtained through scanner.In addition the step of most critical need split it, thus obtain complete three-dimensional face region.Due to the Boundary Detection of three-dimensional face effectively cannot be realized, so, in order to complete the full segmentation of three-dimensional face head portrait, the complete data texturing on two-dimension human face picture need be utilized.Two-dimension human face image gathers profile, obtains facial contour border and ear border.All the other information can be determined in three-dimensional data, then the 3-D view after shoulder is deleted.After completing border collection, effectively can realize the segmentation to three-dimensional face region.

The concrete three-dimensional face recognition technology method step of the present invention is as follows:

Step 1: the extraction of three-dimensional face geometric properties

1, the distance between two is regarded as the benchmark of measurement, other features can obtain according to ratio of the spacing of itself and eyes.

(1) suppose that the distance between two is F1: the European geometric distance between endocanthion point and outer canthus point.

(2) the European geometric distance between distance F2: two endocanthion points between people's intraocular part.

(3) the European geometric distance between distance F3: two outer canthus points between human eye outside.

(4) the height F4 of nose: tip of the nose is to the European geometric distance of nose root.

2, the feature of face significant points angulation

(1) the angle F5 formed between the left wing of nose, nose, right wing of nose three.

(2) angulation F6 between eyespot in the interior eyespot of left eye, nose, right eye.

(3) angulation F7 between the outer eyespot of left eye, nose, the outer eyespot of right eye.

(4) nose and corners of the mouth left and right edge point angulation F8.

3, the feature of region of interest volume

The volume F9 of emphasis to face nose calculates.

The particular location of all unique points as shown in Figure 1 to Figure 3.

Step 2: the measurement of three-dimensional geometry feature

After determining above-mentioned three-dimensional face geometric properties, it completes the Measurement accuracy to it, thus evaluates its accuracy for geometric properties identification.

1, European geometric distance is calculated

The coordinate of any two three-dimensional feature points can be described as P ₁(x ₁, y ₁, z ₁), P ₂(x ₂, y ₂, z ₂), then the air line distance of the two is the European geometric distance between 2, and formula is described below:

$d=\sqrt{{(x_1-x_2)}^2+{(y_1-y_2)}^2+{(z_1-z_2)}^2}---\left(1\right)$

The eigenwert of F1 ~ F3 can be obtained by above formula.Wherein, calculate different to the geometric properties of F4.Suppose that the coordinate of nose is (x ₁, y ₁, z ₁), the coordinate of nose root is (x ₂, y ₂, z ₂), under nose, the coordinate of root is (x ₃, y ₃, z ₃), then the space line equation that under nose root and nose, root 2 is formed can be described as:

Ay+Bz+C＝0；

In formula, A=z ₂-z ₃, B=y ₃-y ₂c=y ₂(z ₃-z ₂)-z ₂(y ₃-y ₂).

Then can obtain nose height:

$F4=\frac{|{\mathrm{Ay}}_1+{\mathrm{Bz}}_1+C|}{\sqrt{A^2+B^2}}.---\left(2\right)$

Step 3: obtain angle geometric properties

Because the angle produced between certain position of face is not subject to the interference of extraneous factor, so it can be used for identifying.Such as, the angle produced between the left wing of nose, nose, right wing of nose three is regarded as the three-dimensional geometry feature of identification:

$F11={\sin }^{-1}\left(\frac{2*S}{L_1*L_2}\right)---\left(3\right)$

In formula, L ₁for describing the distance between d and a; L ₂for describing the distance between e and a; S is for describing above-mentioned leg-of-mutton area.Can be calculated the three-dimensional geometry feature of F5 ~ F8 by the method.D represents nose left side flap, and a represents nose, and e represents nose right side flap, and b represents the nasion, and c represents root under nose.As shown in Figure 4.

The angle F5 formed between the left wing of nose, nose, right wing of nose three.

$F5={\sin }^{-1}\left(\frac{2*S_2}{L_3*L_4}\right)---\left(4\right)$

In formula, L ₃for describing the distance between the left wing of nose and nose; L ₄for describing the distance between nose and the right wing of nose; S ₂for describing above-mentioned leg-of-mutton area, F5≤F11.

Angulation F6 between eyespot in the interior eyespot of left eye, nose, right eye.

$F6={\sin }^{-1}\left(\frac{2*S_3}{L_5*L_6}\right)---\left(5\right)$

In formula, L ₅for describing the distance in left eye between eyespot and nose; L ₆for describing the distance in nose and right eye between eyespot; S ₃for describing above-mentioned leg-of-mutton area.

Angulation F7 between the outer eyespot of left eye, nose, the outer eyespot of right eye.

$F7={\sin }^{-1}\left(\frac{2*S_4}{L_7*L_8}\right)---\left(6\right)$

In formula, L ₇for describing the distance in left eye between eyespot and nose; L ₈for describing the distance between nose and the outer eyespot of right eye; S ₄for describing above-mentioned leg-of-mutton area, F7>=F6.

Nose and corners of the mouth left and right edge point angulation F8.

$F8={\sin }^{-1}\left(\frac{2*S_5}{L_9*L_{10}}\right)---\left(7\right)$

In formula, L ₉for describing the distance between the left corners of the mouth and nose; L ₁₀for describing the distance between nose and the right corners of the mouth; S ₅for describing above-mentioned leg-of-mutton area, F8>=F11>=F5.

Gather the volumentary geometry feature of nose

Obtain the volume of nose:

$F15=\frac{1}{3}*S*F4---\left(8\right)$

In formula, S is for describing the floorage of bottom surface quadrilateral, and F4 is for describing the height of nose.Measure bd, the distance of dc, ce, be, afterwards, be diagonal line with bc, be divided into two triangle bdc and ceb, by sine and the cosine law, obtain two leg-of-mutton areas respectively, they and be the floorage S of required bottom surface quadrilateral.

The face three-dimensional geometry feature for identifying can be obtained.

Step 4: obtain similarity by the three-dimensional geometry feature of above-mentioned acquisition

Choose sample, measure the three-dimensional geometry feature of sample, and be normalized, generating feature vector, afterwards, compares with the normalization characteristic vector in standard sample database, calculates similarity, by analysis, show that calculating formula of similarity is as follows:

$S(I_1,I_2)=\frac{P\left(\mathrm{\Δ}\right|{\mathrm{\Ω}}_i)P\left({\mathrm{\Ω}}_i\right)}{P\left(\mathrm{\Δ}\right|{\mathrm{\Ω}}_i)P\left({\mathrm{\Ω}}_i\right)+P\left(\mathrm{\Δ}\right|{\mathrm{\Ω}}_E)P\left({\mathrm{\Ω}}_E\right)}---\left(9\right)$

In formula, use I respectively ₁with I ₂shape of face image feature amount is described, P (Ω _e) for describing certain condition of work.The dependent probability etc. of such as testing time.Then three-dimensional identification problem just becomes the problem calculating dependent probability, namely belongs to Ω by the feature difference △ of computer memory _ior belong to Ω _eprobability, the judgement of three-dimensional face identification can be completed.If P is (Ω _i| △) >P (Ω _e| △) or S (I ₁, I ₂) >1/2, then think and identify successfully.

The three-dimensional face system adopting the inventive method to mould can be good at the identification of the three-dimensional face figure synthesized based on binocular stereo vision and iterated interpolation algorithm.Although do not reach the recognition accuracy in two-dimentional recognition system, but under human face expression changes little state, can effective implemention identification.The present invention is directed to the drawback that traditional algorithm cannot effectively realize when very large change occurs human face expression identifying, propose a kind of three-dimensional face recognition algorithm based on criticality.The correlation degree that human face expression is changed by probability theory by above-mentioned algorithm is introduced in the calculating of similarity, thus recognizer can effectively be completed because expression changes the process of the Face geometric eigenvector difference caused.Through experimental verification, recognizer has very high practicality.

The present invention will be described to enumerate specific embodiment below:

Embodiment 1:

First, the three-dimensional face model submeter of foundation, with different expressions, calculates the geometric similarity degree of four kinds of different expressions and correlation degree similarity respectively.

Choose 500 country variants, race face carry out three-dimensional feature extract measure, the result drawn is as shown in table 1 below:

The geometric properties of table 1 standard three-dimensional face

After the three-dimensional face geometric properties obtaining standard, need to gather the three-dimensional face geometric properties under four kinds of different expressions, then according to method herein, obtain association similarity, concrete expression shape change situation is as shown in table 2:

Table 2 three-dimensional face model Extraction of Geometrical Features

From above-mentioned 500 samples, randomly draw five faces, detect, by the repetitive measurement of the geometric properties to 5 faces, carrying out weighed value adjusting according to constantly learning, determining weights as shown in table 3 below:

Table 3 face feature vector weights

As can be seen from Table 2, when the expression of face changes time, some geometric properties of organ participating in expression shape change can have greatly changed, such as, pout one's lips and rouse the mouth feature of the cheek, eye feature time dejected and nose feature etc.Traditional method based on geometric properties there will be erroneous judgement in time running into this situation, and under the state of dejected this expression, similarity is less than 0.5, identifies and does not pass through, and occurs erroneous judgement.And the association Similarity Measure based on correlation degree in this paper well solves this problem, that is, under identification all successfully situation, the similarity of method identification is herein higher.Therefore, the identification that method is herein applied to human face three-dimensional model is highly effective.

The above is only to better embodiment of the present invention, not any pro forma restriction is done to the present invention, every any simple modification done above embodiment according to technical spirit of the present invention, equivalent variations and modification, all belong in the scope of technical solution of the present invention.

Claims (1)

1., based on the three-dimensional face recognition technology of correlation degree, it is characterized in that carrying out according to following steps:

Step 1: the extraction of three-dimensional face geometric properties;

Distance between two is regarded as the benchmark of measurement, other features obtain according to the ratio of the spacing of itself and eyes:

(1) suppose that the distance between two is F1: the European geometric distance between endocanthion point and outer canthus point;

(2) the European geometric distance between distance F2: two endocanthion points between people's intraocular part;

(3) the European geometric distance between distance F3: two outer canthus points between human eye outside;

(4) the height F4 of nose: tip of the nose is to the European geometric distance of nose root;

The feature of face significant points angulation

(1) the angle F5 formed between the left wing of nose, nose, right wing of nose three;

(2) angulation F6 between eyespot in the interior eyespot of left eye, nose, right eye;

(3) angulation F7 between the outer eyespot of left eye, nose, the outer eyespot of right eye;

(4) nose and corners of the mouth left and right edge point angulation F8;

Step 2: the measurement of three-dimensional geometry feature

1, European geometric distance is calculated

The coordinate of any two three-dimensional feature points is described as P ₁(x ₁, y ₁, z ₁), P ₂(x ₂, y ₂, z ₂), then the air line distance of the two is the European geometric distance between 2, and formula is described below:

$d=\sqrt{{(x_1-x_2)}^2+{(y_1-y_2)}^2+{(z_1-z_2)}^2}---\left(1\right)$

Namely the eigenwert of F1 ~ F3 is obtained by above formula; Suppose that the coordinate of nose is (x ₁, y ₁, z ₁), the coordinate of nose root is (x ₂, y ₂, z ₂), under nose, the coordinate of root is (x ₃, y ₃, z ₃), then the space line equation that under nose root and nose, root 2 is formed is described as:

Ay+Bz+C＝0

In formula, A=z ₂-z ₃, B=y ₃-y ₂c=y ₂(z ₃-z ₂)-z ₂(y ₃-y ₂);

Then nose height:

$F4=\frac{|{\mathrm{Ay}}_1+{\mathrm{Bz}}_1+C|}{\sqrt{A^2+B^2}};---\left(2\right)$

Step 3: obtain angle geometric properties

The angle produced between the left wing of nose, nose, right wing of nose three is regarded as the three-dimensional geometry feature of identification:

$F11={\sin }^{-1}\left(\frac{2*S}{L_1*L_2}\right)---\left(3\right)$

In formula, L ₁for describing the distance between d and a; L ₂for describing the distance between e and a; S is for describing above-mentioned leg-of-mutton area, and d represents nose left side flap, and a represents nose, and e represents nose right side flap, and b represents the nasion, and c represents root under nose;

The angle F5 formed between the left wing of nose, nose, right wing of nose three;

$F5={\sin }^{-1}\left(\frac{2*S_2}{L_3*L_4}\right)---\left(4\right)$

In formula, L ₃for describing the distance between the left wing of nose and nose; L ₄for describing the distance between nose and the right wing of nose; S ₂for describing above-mentioned leg-of-mutton area;

Angulation F6 between eyespot in the interior eyespot of left eye, nose, right eye;

$F6={\sin }^{-1}\left(\frac{2*S_3}{L_5*L_6}\right)---\left(5\right)$

In formula, L ₅for describing the distance in left eye between eyespot and nose; L ₆for describing the distance in nose and right eye between eyespot; S ₃for describing above-mentioned leg-of-mutton area;

Angulation F7 between the outer eyespot of left eye, nose, the outer eyespot of right eye;

$F7={\sin }^{-1}\left(\frac{2*S_4}{L_7*L_8}\right)---\left(6\right)$

In formula, L ₇for describing the distance in left eye between eyespot and nose; L ₈for describing the distance between nose and the outer eyespot of right eye; S ₄for describing above-mentioned leg-of-mutton area;

Nose and corners of the mouth left and right edge point angulation F8

$F8={\sin }^{-1}\left(\frac{2*S_5}{L_9*L_{10}}\right)---\left(7\right)$

In formula, L ₉for describing the distance between the left corners of the mouth and nose; L ₁₀for describing the distance between nose and the right corners of the mouth; S ₅for describing above-mentioned leg-of-mutton area;

Gather the volumentary geometry feature of nose

Obtain the volume of nose:

$F15=\frac{1}{3}*S*F4---\left(8\right)$

In formula, S is for describing the floorage of bottom surface quadrilateral, and F4 is for describing the height of nose; Measure bd, the distance of dc, ce, be, afterwards, be diagonal line with bc, be divided into two triangle bdc and ceb, by sine and the cosine law, obtain two leg-of-mutton areas respectively, they and be the floorage S of required bottom surface quadrilateral;

Step 4: obtain similarity by the three-dimensional geometry feature of above-mentioned acquisition

Choose sample, measure the three-dimensional geometry feature of sample, and be normalized, generating feature vector, afterwards, compares with the normalization characteristic vector in standard sample database, calculates similarity, by analysis, show that calculating formula of similarity is as follows:

$S(I_1,I_2)=\frac{P\left(\mathrm{\Δ}\right|{\mathrm{\Ω}}_i)P\left({\mathrm{\Ω}}_i\right)}{P\left(\mathrm{\Δ}\right|{\mathrm{\Ω}}_i)P\left({\mathrm{\Ω}}_i\right)+P\left(\mathrm{\Δ}\right|{\mathrm{\Ω}}_E)P\left({\mathrm{\Ω}}_E\right)}---\left(9\right)$

In formula, use I respectively ₁with I ₂shape of face image feature amount is described, P (Ω _e) for describing certain condition of work; Ω is belonged to by the feature difference △ of computer memory _ior belong to Ω _eprobability, complete the judgement of three-dimensional face identification, if P (Ω _i| △) >P (Ω _e| △) or S (I ₁, I ₂) >1/2, then think and identify successfully.