CN202887210U - Multispectral face recognition system - Google Patents

Abstract

The utility model relates to a multispectral face recognition system. It is characterized in that the multispectral face recognition system includes a multispectral imaging system, a color camera, a face recognition module, a data storage module, a central control module and a spectrograph. The multispectral imaging system outputs data of a shot face image to the face recognition module. The face recognition module recognizes information of a standard face database according to a data storage module and outputs a recognition result. The central control module controls the image shooting of the multispectral imaging system and the recognition of the face recognition module. The multispectral imaging system includes an objective lens, a liquid crystal adjustable filtering sheet and a CCD camera. The liquid crystal adjustable filtering sheet is arranged in front of a CCD lens of the CCD camera. The objective lens is arranged on the front end of the liquid crystal adjustable filtering sheet. According to the utility model, the extraction of multiple feathers of the face image enables the between-class distance more distinct in a recognition process and enhances the separability of a recognition algorithm, so that the improvement of the recognition effect is facilitated.

Description

A kind of multispectral face identification system

Technical field

The present invention relates to the living things feature recognition field, be specifically related to a kind of multispectral face identification system.

Background technology

A kind of as biometrics identification technology, recognition of face has directly, close friend, nature, high acceptable characteristics, and the user is without any mental handicape, and image acquisition is convenient; In addition, we can also do to the result of recognition of face further analysis, obtain extra abundant informations such as sex, expression, age, so this technology has obtained widely research and used.Current, face recognition technology mainly is applied to the aspects such as video monitoring, entrance control, criminal investigation and case detection, certificate verification, and face recognition technology also has huge application prospect in fields such as medical science, file administration, video conferences in addition.

At present, face identification system and correlation technique mainly are based on the recognition of face of common RGB coloured image, and this also is the most familiar recognition method of people.Ideally, different people's faces can be distinguished by the human face detection and recognition system of a robust under unconfined condition, and simultaneously its target far away of also can adjusting the distance is distinguished.But the change of the factors such as hair that human face expression, cosmetic, glasses, face are scattered all can make people's face produce difference, and the external factor such as radiometric response of illumination, camera perspective, camera also can cause the huge variation of the institute's facial image that obtains generation.Based on the face recognition technology of traditional imaging system owing to only utilized the space geometry feature of object of observation, it is very responsive to change the uncertainty of bringing for various conditions, only have in the limited situation of externally factor and internal factor and could obtain satisfied effect, and recognition performance can sharply descend under uncontrolled environment, and therefore the defective that is difficult to overcome is arranged.

Summary of the invention

The technical matters that solves

For fear of the deficiencies in the prior art part, the present invention proposes a kind of multispectral face identification system, multispectral image be by multi-spectral imager from visible light near infrared tens in addition hundreds of continuous narrow wave band in the image that obtains, the information of different-waveband has represented the radiation/reflection case of object being observed in different spectrum frequency ranges, has clear and definite physical significance.The biological characteristic of multi-channel spectral imaging, especially multispectral imaging is surveyed can on the basis that obtains traditional biosome space characteristics information, obtain the spectral signature information of biosome simultaneously.Correlative study shows that the spectral characteristic of skin can reflect the uniqueness of skin color, and different Person's skins has different spectral characteristics, and this just provides possibility for the object that utilizes the concrete identity of spectral information recognition and verification.Multispectral imaging is applied to recognition of face, may detect the spectral information that the traditional optical image can't detect, not only can improve the stability of recognizer, the impersonation that the means such as more can avoid making up, copy cause is conducive to the reliability of Integral lifting face identification system.

Technical scheme

A kind of multispectral face identification system is characterized in that comprising multi-optical spectrum imaging system, color camera, face recognition module, data memory module, Central Control Module and spectrometer; Multi-optical spectrum imaging system exports the facial image data of taking to face recognition module, and face recognition module is identified according to the information of the standard faces database in the data memory module, and the result that then will identify exports; The image capture of Central Control Module control multi-optical spectrum imaging system and the identification of face recognition module; Described multi-optical spectrum imaging system comprises object lens 6, liquid crystal adjustable optical filter 3 and CCD camera 5; Be provided with liquid crystal adjustable optical filter 3 before the CCD of CCD camera 5 camera lens, the front end of liquid crystal adjustable optical filter 3 is provided with object lens 6.

Beneficial effect

A kind of multispectral face identification system that the present invention proposes, compare and have following advantage: 1, multispectral imaging is applied to human face detection and recognition, compares conventional optical imagery, can add the spectral signature information of obtaining biosome, has higher recognition performance; Introduce the high precision spectrometer in the system, obtain the spectral reflectivity of people's face skin by the inverting to gradation of image value (radiance), can overcome the impact of the environmental baseline such as illumination; The skin spectrum characteristic has relative stability, abundant spectral information has not only overcome the problem of utilizing merely the space geometry feature to bring, also become possibility so that better distinguish the similar object of observation of appearance, thereby can avoid some the impersonation means in the face recognition process; The extraction of many features in the facial image, also so that in the identifying between class distance more obvious, the separability of recognizer is stronger, is conducive to improve recognition effect.

Description of drawings

Fig. 1: be the structural representation of an embodiment of the multispectral face identification system of the present invention;

Fig. 2: among the present invention based on the imaging schematic diagram of liquid crystal tunable optical filter multispectral camera;

Fig. 3: among the present invention based on the structural representation of liquid crystal tunable optical filter multispectral camera;

1-the first camera lens, 2-the second camera lens, 3-liquid crystal adjustable optical filter, 4-three-lens, 5-CCD camera, 6-object lens, 7-CCD camera lens, 8-mounting hole.

Embodiment

Now in conjunction with the embodiments, the invention will be further described for accompanying drawing:

The invention provides a kind of multispectral face identification system, as shown in Figure 1, mainly comprise image capture module, spectrometer module and host computer system.

Such as Fig. 1, the front end of system is image capture module, comprise multi-optical spectrum imaging system and common color camera two parts, and two parts is separate.Multi-optical spectrum imaging system (shown in Fig. 2,3) is based on liquid crystal adjustable optical filter (LCTF) and the development of CCD camera, utilize the visible light near infrared area array CCD detector of high detectivity to cooperate high performance liquid crystal adjustable optical filter, light-splitting device and the device of looking in the distance, to stare the mode imaging.In the front end installation process, camera and LCTF fix with the optics brassboard, make each other not inclination, rolling problem.The liquid crystal adjustable optical filter is connected with the USB mouth of background computer by control enclosure, by USB confession electricity.Color camera is the image collecting device annex, adopt traditional on the market video capture device, be independent of multispectral people's face imaging system when building, only link to each other with the foreground display device, major function provides traditional video image and shows as the client area, to improve the customer experience degree.

Host computer system is made of the microcomputer that is positioned at the backstage usually, mainly comprises multispectral face recognition module, Central Control Module and aforesaid data memory module.In addition, the multiway images transmitting device is made of the PCI video frequency collection card usually, loads in the host computer system, and links to each other with front end high-gain camera; The high precision spectrometer equipment is used for gathering environment-identification standard spectrum data, and links to each other with data memory module.

In the embodiment, the multi-optical spectrum imaging system (as shown in Figure 2) that adopts is based on liquid crystal adjustable optical filter (LCTF) and the development of CCD camera, liquid crystal adjustable optical filter (LCTF) is the multispectral liquid crystal filter of Varispec that U.S. CRI company researches and develops, can be in wavelength coverage 400 to 720nm, 650 to 1100nm, 850 to 1800nm, 1200 to plurality of optional such as 2450nm, its half-peak breadth is 20nm.The CCD camera is QImaging Retiga Exi model.Color camera is the image collecting device annex, adopts traditional on the market video capture device, adopts dimension to look the Mv-1300Uc of company camera herein.

Host computer system is made of the microcomputer that is positioned at the backstage usually, mainly comprises multispectral face recognition module, Central Control Module and aforesaid data memory module.Computer model is Lenovo Qitian M7300.In addition, the multiway images transmitting device is made of the PCI1394 video frequency collection card usually, loads in the host computer system, and links to each other with front end high-gain camera.

The premiere feature of multispectral face recognition module is the multispectral image data acquisition, the spectrum picture acquisition module is developed on the SDK basis of LCTF and CCD parts, mainly comprises the functions such as the interface querying of liquid crystal adjustable optical filter, band selection, wave band autoscan, the setting of CCD parameter, image-capture and autostore.Its development language is Vc++6.0.In addition, face recognition module also need be finished face identification functions automatically, mainly comprises following a few partial function: people's face detects (mainly finishing affirmation and the extraction of people face position in the input picture); People's face is cut apart (mainly finishing the location to people's face each several part such as eyes, lip, cheek); People's face spectroscopic data inverting (mainly finishing the facial image gray-scale value to the conversion of spectral reflectance values); Face characteristic extracts (extracting the features such as how much of people's face, spectrum); Recognition of face (compare and distinguish with the data in the standard database).

When system uses, enter guarded region when having detected the visitor, the background control system automated control chart can be coordinated optical filter and high-gain camera to gather interviewee's multispectral facial image as the electricity in the acquisition module; View data passes to host computer system through the multi pass acquisition card; The back-end computer system finishes face characteristic by face recognition module and extracts, by with data memory module in standard faces image library exchange contrast finish final recognition function, the output recognition result.

Face recognition module also need be finished face identification functions automatically, mainly comprises following a few partial function: people's face detects (mainly finishing affirmation and the extraction of people face position in the input picture); People's face is cut apart (mainly finishing the location to people's face each several part such as eyes, lip, cheek); People's face spectroscopic data inverting (mainly finishing the facial image gray-scale value to the conversion of spectral reflectance values); Face characteristic extracts (extracting the features such as how much of people's face, spectrum); Recognition of face (compare and distinguish with the data in the standard database).Modular system is when functional development, and take the used operating system of background computer and the API that provides thereof as the basis, development language is without particular requirement, but needs to guarantee that each functional module is relatively independent, to improve the stability of module.The flow process of face recognition module is as follows:

Step 1: extract five zones from multispectral facial image: hair, forehead, left cheek, right cheek and lip; We intercept a fritter human face region from facial image, and represent a sample with the spectral absorption index vector that obtains, and for direct picture, have extracted five zones, are respectively: hair, forehead, left cheek, right cheek, lip.And for all the other angled images, what get then is the subset in these five zones, and these zones are in the visible range of correspondence image.

Step 2: entering between the radiation intensity Lg of detector and the reflectivity R has linear relationship: Lg=k * R+b, according to inverse model:

$k=\frac{\mathrm{N\Σ}\left(D{N}_i{R}_i\right)-\mathrm{\ΣD}{N}_i\mathrm{\Σ}{R}_i}{N{\mathrm{\Σ}{R}_i}^2-{\left(\mathrm{\Σ}{R}_i\right)}^2}$

$b=\frac{\mathrm{\Σ}\left(D{N}_i{R}_i\right)-\mathrm{k\Σ}{R_i}^2}{\mathrm{\Σ}{R}_i}$

In the following formula, DN _iBe the gray-scale value of the sample areas on the image, R _iBe the corresponding region spectral reflectivity of spectrometer gained, N is the pixel count in the sample areas, and summation is to the summation of N point;

Calculate the spectral reflectivity of multispectral facial image and the Relation Parameters k of gray scale, b by least square method;

It is regional to have little or no prestige for everyone who obtains, and the advanced person passes through and tests the linear approach inverting, obtains reflectivity data R (x, y, λ _i), spectral reflectivity vector R _t=(R _t(λ ₁), R _t(λ ₂) ... Rt (λ _Ln)) ^TObtain by following formula

$\begin{array}{cc}{R}_t\left({\mathrm{\λ}}_i\right)=\frac{1}{N}\underset{x,y}{\mathrm{\Σ}}R(x,y,{\mathrm{\λ}}_i)& i=\mathrm{1,2},...I_n\end{array}$

Wherein, R (x, y, λ _i) be λ _iWave band, the reflectivity that pixel (x, y) is located, I _nBe the wave band number, to all pixels summation in the choosing face skin square zone, t is a kind of in following five kinds of types of organizations: forehead, left cheek, right cheek, hair and lip.The method of asking of spectral reflectivity is:

The DN value of image that imaging spectrometer obtains can not represent the reflectance value of target, namely after sensor is calibrated, by sensor output is the gray-scale value of target, and the database that we set up is the reflectivity of target, be necessary the radiance value that camera obtains is converted to reflectance value, i.e. the calibration of so-called multispectral data.

Suppose between the radiation intensity Lg that enters detector and the reflectivity R linear relationship is arranged

Lg=k×R+b

Inverse model is

$k=\frac{\mathrm{N\Σ}\left(D{N}_i{R}_i\right)-\mathrm{\ΣD}{N}_i\mathrm{\Σ}{R}_i}{N{\mathrm{\Σ}{R}_i}^2-{\left(\mathrm{\Σ}{R}_i\right)}^2}$

$b=\frac{\mathrm{\Σ}\left(D{N}_i{R}_i\right)-\mathrm{k\Σ}{R_i}^2}{\mathrm{\Σ}{R}_i}$

Try to achieve unknown number k by least square method, b, and these parameters all are the constant values in this linear model.The constant value that a little binary regressions are obtained is applied to the curve of spectrum inverting at other pixel place of image, just can obtain the accurately curve of spectrum of arbitrary pixel place.

Step 3

Ask for the spectral reflectivity of hair zones: for everyone face hair zones, the advanced person passes through and tests the linear approach inverting, according to gray scale DN _HiObtain reflectivity data: R _h(x, y, λ _i)=(DN _Hi(x, y, λ _i)-b)/k, by

Obtain spectral reflectivity vector R _h=(R _h(λ ₁), R _h(λ ₂) ... R _h(λ _Ln)) ^T

Wherein, R (x, y, λ _i) be λ _iWave band, the reflectivity that pixel (x, y) is located, I _nBe the wave band number, to all pixel summations in the choosing face hair square zone;

Ask for the spectral reflectivity of forehead region: for everyone face forehead region, the advanced person passes through and tests the linear approach inverting, according to gray scale DN _FiObtain reflectivity data: R _f(x, y, λ _i)=(DN _Fi(x, y, λ _i)-b)/k, by

Obtain spectral reflectivity vector R _f=(R _f(λ ₁), R _f(λ ₂) ... R _f(λ _Ln)) ^T

Wherein, R (x, y, λ _i) be λ _iWave band, the reflectivity that pixel (x, y) is located, I _nBe the wave band number, to all pixel summations in the choosing face forehead square zone;

Ask for the spectral reflectivity in left cheek zone: for the left cheek of everyone face zone, the advanced person passes through and tests the linear approach inverting, according to gray scale DN _LciObtain reflectivity data R _Lc(x, y, λ _i)=(DN _Lci(x, y, λ _i)-b)/k, by Obtain spectral reflectivity vector R _Lc=(R _Lc(λ ₁), R _Lc(λ ₂) ... R _Lc(λ _Ln)) ^T

Wherein, R (x, y, λ _i) be λ _iWave band, the reflectivity that pixel (x, y) is located, I _nBe the wave band number, to all pixel summations in the left cheek square of the choosing face zone;

Ask for the spectral reflectivity in right cheek zone: for the right cheek of everyone face zone, the advanced person passes through and tests the linear approach inverting, according to gray scale DN _{R c}Obtain reflectivity data

By

Obtain spectral reflectivity vector R _Rc=(R _Rc(λ ₁), R _Rc(λ ₂) ... R _Rc(λ _Ln)) ^T

Wherein, R (x, y, λ _i) be λ _iWave band, the reflectivity that pixel (x, y) is located, I _nBe the wave band number, to all pixel summations in the right cheek square of the choosing face zone;

Ask for the spectral reflectivity of lip region: for everyone face lip region, the advanced person passes through and tests the linear approach inverting, according to gray scale DN _LiObtain reflectivity data R _l(x, y, λ _i)=(DN _Li(x, y, λ _i)-b)/k, by

Obtain spectral reflectivity vector R _l=(R _l(λ ₁), R _l(λ ₂) ..R _l(λ _Ln)) ^TWherein, R (x, y, λ _i) be λ _iWave band, the reflectivity that pixel (x, y) is located, I _nBe the wave band number, to all pixel summations in the choosing face lip square zone;

Step 4: the spectrum to regional adopts respectively the envelope null method to analyze spectrum.If curve of spectrum array is R (i), i=0,1,2 ... k-1, the wavelength array is W (i), i=0,1,2...k-1, concrete steps are as follows:

Step a:i=0 brings R (i), W (i) into the envelope node listing;

Step b: the envelope node of looking for novelty, if i=k-1 finishes; Otherwise, make j=i+1, continue circulation;

Step c: check the intersection point of straight line (i, j) and curve of spectrum W (i), if j=k-1 finishes, R (i), W (i) are joined in the envelope node table, otherwise:

1）m=j+1

2) if m=j-1 finishes inspection, j is the node on the envelope, and R (i), W (i) are joined in the envelope node table, and i=j forwards step b to;

3) ask the intersection point r1 (m) of straight line (i, j) and curve of spectrum W (i);

4) if R (m)＞r1 (m, then j is not the point on the envelope, j=j+1 forwards step c to; If

R (m)＜r1 (m); Then straight line (i, j) has at most an intersection point with curve of spectrum W (i), and m=m+1 forwards 2 to);

Steps d: after obtaining the envelope node table, adjacent node is connected successively with straight-line segment, obtains functional value H (i) i=0 of the point on the corresponding broken line of W (i), 1,2 ... k-1, thereby obtain this curve of spectrum, envelope, H (i)＞R (i) is obviously arranged;

Step e: the curve of spectrum is carried out envelope eliminate: R _u(i)=and R (i)/H (i), i=0,1,2,3 ... k-1, thereby the curve of spectrum R after obtaining envelope and eliminating _u(i);

Calculate the parameters such as centre wavelength M, absorption width D, absorption degree of depth W, symmetry S and absorption area A of the curve of spectrum with envelope removal method, specific algorithm is as follows: all maximum points of at first finding out the reflectance spectrum curve, with envelope they are connected successively, calculate that reflectivity obtains " ratio reflectivity " with corresponding envelope luminance factor value on each wavelength location, each minimum point of ratio reflectance curve is characteristic absorption peak.Obtain after the characteristic absorption peak, the characteristic parameter of each absorption peak is also just determined thereupon.

Step 5: SAI is calculated respectively in each zone,

$\mathrm{SAI}=\frac{d{R}_{S1}+(1-d)R_{S2}}{R_M},$

Wherein: centre wavelength M is wavelength location corresponding to minimum point on the reflectance curve; R _S1, λ _S1For absorbing reflectivity and the wavelength location of left point of shoulder S1 on the reflectance curve; R _M, λ _MReflectivity and wavelength location for absorption point M on the reflectance curve; R _S2, λ _S2For absorbing reflectivity and the wavelength location of right point of shoulder S2 on the reflectance curve; Absorb right point of shoulder on the reflectance curve and be absorption width W=λ with the wavelength difference that absorbs left point of shoulder _S2-λ _S1Reflectance curve polishing wax absorption depth D=| 1-R _M|;

Step 6: calculate t people from place of the types of organization face sample i in above-mentioned 5 zones to the distance of sample j, define by mahalanobis distance: ${D_t}^{\′}(i,j)={({\stackrel{\‾}{\mathrm{SAI}}}_t\left(i\right)-{\stackrel{\‾}{\mathrm{SAI}}}_{t(}j))}^T{\∑}_t^{-1}({\stackrel{\‾}{\mathrm{SAI}}}_t\left(i\right)-{\stackrel{\‾}{\mathrm{SAI}}}_t\left(j\right))$

People's face sample i to the distance of sample j is

D(i,j)=ω _fD _f(i,j)+ω _lcD _lc(i,j)+ω _rcD _rc(i,j)+ω _hD _h(i,j)+ω _lD _l(i,j)

Each weights of following formula ω is visible or invisiblely get 1 or 0 in image according to corresponding zone;

Step 7: for each sample, ∑ t is corresponding vector

Covariance matrix, dimension is I _nEach t of types of organization of whole database gets a covariance matrix ∑ t.We are similar to ∑ t by the diagonal matrix Lt that variance corresponding to each wave band generates.

Step 8: suppose to have C class sample, for training sample j, the similar T that is defined as with it in the test sample book storehouse _jWe calculate first the distance B (i, j) of j each image i in the Sample Storehouse, if D is (T _j, be the central minimum of C kind distance j), think that then sample j is correctly validated.

Before multispectral face identification system uses, need to be in data memory module the Criterion face database, mainly be included in grade personnel's standard multispectral face characteristic storehouse and personal information, pre-identification information etc., wherein personal information should comprise name, sex, age etc., to help inquiry and the monitoring of recognition result.In addition, after recognition system is built and finished, also need use the spectroscopic data under several typical scenes of high precision spectrometer collection on-gauge plate in system's environment of living in, comprise the conditions such as intense light irradiation, low light level photograph and indoor illumination, and deposit data memory module in.The standard spectrum data are used for inverting and the correction of multispectral people's face curve of spectrum, the impact that makes identifying not changed by other environmental baselines such as illumination.It should be noted that institute's normal data of depositing all must regular update in the database.

When system uses, enter guarded region when having detected the visitor, the background control system automated control chart can be coordinated optical filter and high-gain camera to gather interviewee's multispectral facial image as the electricity in the acquisition module; View data passes to host computer system through the multi pass acquisition card; The back-end computer system finishes face characteristic by face recognition module and extracts, by with data memory module in standard faces image library exchange contrast finish final recognition function, the output recognition result.

Claims (1)

1. a multispectral face identification system is characterized in that comprising multi-optical spectrum imaging system, color camera, face recognition module, data memory module, Central Control Module and spectrometer; Multi-optical spectrum imaging system exports the facial image data of taking to face recognition module, and face recognition module is identified according to the information of the standard faces database in the data memory module, and the result that then will identify exports; The image capture of Central Control Module control multi-optical spectrum imaging system and the identification of face recognition module; Described multi-optical spectrum imaging system comprises object lens (6), liquid crystal adjustable optical filter (3) and CCD camera (5); Be provided with liquid crystal adjustable optical filter (3) before the CCD camera lens of CCD camera (5), the front end of liquid crystal adjustable optical filter (3) is provided with object lens (6).

Images