CN205610781U - Adaptive control device that facial image gathered - Google Patents

Abstract

The utility model discloses an adaptive control device that facial image gathered, which comprises a support, be equipped with the vertical guide rail of two piece at least vertical settings on the frame, be equipped with vertical transmission shaft on the frame, the cover is equipped with along the vertical gliding vertical slider of vertical guide rail on the vertical transmission shaft, be equipped with transverse guide between the vertical slider, the last transverse transmission shaft that is equipped with of transverse guide, the last cover of transverse transmission shaft is equipped with and carries on the platform along transverse guide lateral sliding's camera, the camera carries on the bench and is equipped with the rotation motor, rotate the vertical setting of motor shaft, be equipped with along with pivot pivoted camera on rotating motor shaft, the mode of rotatory camera is automatic about and follows people's face adaptive control camera through horizontal migration, vertical migration, can reduce artifical error of correcting, has guaranteed better that the photo taking out satisfies the requirement of certificate photograph, has improved user's comfort level.

Description

A kind of adaptive controller of man face image acquiring

[technical field]

This utility model relates to the adaptive controller of a kind of man face image acquiring.

[background technology]

The self photographing equipment of present stage is broadly divided into two big types, the first kind is that the camera installation in equipment only moves up and down, face is moved to center upper and lower in screen, via people, camera installation is carried out left and right again adjust and gather photo when photo meets the requirement of certificate photo, such Machine automated degree is the highest, still need manual operation, and service efficiency height compared with the efficiency of photo studio does not goes out how many；Also having a class machine is that camera installation is fixed on a position, and the stool that camerist is seated moves up and down, and gathers photo when photo meets the requirement of certificate photo.Such Machine automated degree is of a relatively high, but has certain scope, there will be certain inadaptability for the people being too obese, old people and child.The self photographing equipment of both types is required for artificially adjusting and faces photographic head, but this mode that is manually adjusted is inaccurate cannot accurately to ensure that photo meets the requirement of certificate photo.

[utility model content]

This utility model overcomes the deficiency of above-mentioned technology, provide the adaptive controller of a kind of man face image acquiring, by moving horizontally, vertically move and automatically following face Self Adaptive Control photographic head by the way of left rotation and right rotation camera photographic head, the error of manual correction can be reduced, photo out captured by better assuring that meets the requirement of certificate photo, improves comfort level and the satisfaction of user.

For achieving the above object, this utility model have employed following technical proposal:

A kind of adaptive controller of man face image acquiring, include support 100, described support 100 is provided with at least two vertical guide being vertically arranged 1, described support 100 is provided with vertical power transmission shaft 2, the vertical slide block 3 that vertically guide rail 1 is vertically slidable it is arranged with on described vertical power transmission shaft 2, it is provided with cross slide way 4 between described vertical slide block 3, described cross slide way 4 is provided with horizontal power transmission shaft 5, it is arranged with the camera that transversely guide rail 4 slides laterally on described horizontal power transmission shaft 5 and carries platform 6, described camera carries platform 6 and is provided with rotary electric machine 7, the rotating shaft of described rotary electric machine 7 is vertically arranged, the rotating shaft of described rotary electric machine 7 is provided with the camera with axis of rotation.

Described vertical power transmission shaft 2 is provided with vertical power transmission shaft screw thread, described vertical power transmission shaft 2 one end is provided with the first motor 8, described vertical power transmission shaft 2 is with the first motor 8 axis of rotation, and described vertical slide block 3 socket place is provided with the vertical slide block screw thread rotated with vertical power transmission shaft 2 and make vertical slide block 3 vertically move with vertical power transmission shaft threaded engagement.

Described horizontal power transmission shaft 5 is provided with horizontal power transmission shaft screw thread, described horizontal power transmission shaft 5 one end is provided with the second motor 9, described horizontal power transmission shaft 5 is with the second motor 9 axis of rotation, and described camera carries platform 6 socket place and is provided with the camera rotated with horizontal power transmission shaft 5 with horizontal power transmission shaft threaded engagement and make camera carry platform 6 transverse shifting and carries platform screw thread.

It is additionally provided with controller 10 on described support 100, described controller 10 connects first motor the 8, second motor 9, be arranged at the first encoder 11 of vertical power transmission shaft 2 other end, be arranged at the second encoder 12, rotary electric machine 7 and the camera of horizontal power transmission shaft 4 other end, described first motor 8 is connected with the first encoder 11, and described second motor 9 is connected with the second encoder 12.

The beneficial effects of the utility model are:

Human face image collecting device in this utility model can automatically be followed face Self Adaptive Control and be adjusted the orientation of photographic head, realizes taken human face photo and meets the requirement of certificate photo；Design camera in apparatus structure part and carry platform, and camera can be made to carry the mechanical framework that platform moves up and down accordingly, achieve the horizontal adjustment of camera and the function of vertical adjustment, platform carries a rotary electric machine, camera is arranged in the rotating shaft of rotary electric machine, it is achieved that the left rotation and right rotation function of camera；Meanwhile, the motor of rectilinear movement is fitted into encoder and carries out feedback control, increase substantially precision；Ball-type screw-thread drive shaft leading screw and line slideway is used to alleviate noise when machine runs and driving voltage；Compare the existing mode that can only move up and down photographic head and seat can only be moved up and down, this device can reduce the error of manual correction by moving horizontally, vertically move and automatically follow by the way of left rotation and right rotation photographic head face Self Adaptive Control photographic head, photo out captured by better assuring that meets the requirement of certificate photo, improves comfort level and the satisfaction of user.

[accompanying drawing explanation]

Fig. 1 is this utility model device for mechanical structural representation；

Fig. 2 is this utility model device circuit structural representation；

Fig. 3 is this utility model flow chart；

Fig. 4 is this utility model initial pictures and facial image direct coordinate system schematic diagram.

[detailed description of the invention]

It is described in further detail with embodiment of the present utility model below in conjunction with the accompanying drawings:

As shown in Figure 1, a kind of adaptive controller of man face image acquiring, include support 100, described support 100 is provided with at least two vertical guide being vertically arranged 1, described support 100 is provided with vertical power transmission shaft 2, the vertical slide block 3 that vertically guide rail 1 is vertically slidable it is arranged with on described vertical power transmission shaft 2, it is provided with cross slide way 4 between described vertical slide block 3, described cross slide way 4 is provided with horizontal power transmission shaft 5, it is arranged with the camera that transversely guide rail 4 slides laterally on described horizontal power transmission shaft 5 and carries platform 6, described camera carries platform 6 and is provided with rotary electric machine 7, the rotating shaft of described rotary electric machine 7 is vertically arranged, the rotating shaft of described rotary electric machine 7 is provided with the camera with axis of rotation.

Wherein, described vertical power transmission shaft 2 is provided with vertical power transmission shaft screw thread, described vertical power transmission shaft 2 one end is provided with the first motor 8, described vertical power transmission shaft 2 is with the first motor 8 axis of rotation, and described vertical slide block 3 socket place is provided with the vertical slide block screw thread rotated with vertical power transmission shaft 2 and make vertical slide block 3 vertically move with vertical power transmission shaft threaded engagement.

Described horizontal power transmission shaft 5 is provided with horizontal power transmission shaft screw thread, described horizontal power transmission shaft 5 one end is provided with the second motor 9, described horizontal power transmission shaft 5 is with the second motor 9 axis of rotation, and described camera carries platform 6 socket place and is provided with the camera rotated with horizontal power transmission shaft 5 with horizontal power transmission shaft threaded engagement and make camera carry platform 6 transverse shifting and carries platform screw thread.

As shown in Figure 2, it is additionally provided with controller 10 on described support 100, described controller 10 connects first motor the 8, second motor 9, be arranged at the first encoder 11 of vertical power transmission shaft 2 other end, be arranged at the second encoder 12, rotary electric machine 7 and the camera of horizontal power transmission shaft 4 other end, described first motor 8 is connected with the first encoder 11, and described second motor 9 is connected with the second encoder 12.

Operation principle: controller 10 controls the first motor 8 and drives vertical power transmission shaft 2 to rotate, by the vertical power transmission shaft screw thread being arranged on vertical power transmission shaft 2 and the vertical slide block threaded engagement being arranged in vertical slide block 3, make vertical slide block 3 vertically slidable, it is achieved the height of camera adjusts；Controller 10 controls the second motor 9 and drives horizontal power transmission shaft 5 to rotate, platform threaded engagement is carried with the camera being arranged in camera lift-launch platform 6 by the horizontal power transmission shaft screw thread being arranged on horizontal power transmission shaft 5, make camera carry platform 6 to slide laterally, it is achieved the left and right of camera adjusts；Controller 10 controls rotary electric machine 7 and rotates, make the camera being arranged in rotary electric machine 7 rotating shaft realize left rotation and right rotation, thus reach adaptive controller and can automatically follow face Self Adaptive Control and adjust the orientation of photographic head and meet the requirement of certificate photo to realize taken human face photo.

Meanwhile, the rotational value of the second motor 9 is fed back in controller 10 by the first encoder 11 by the rotational value of the first motor 8, the second encoder 12, improves the accuracy that camera position adjusts.

As shown in Figure 3 and Figure 4, the acquisition method of human face image collecting device, its step is as follows:

A, controller 10 control camera shooting initial pictures, and be input in controller 10, set with the initial pictures upper left corner as initial point o ', it it is to the right x ' axle positive direction, it is y ' axle positive direction downwards, set up rectangular coordinate system O ' x ' y ', in initial pictures rectangular coordinate system O ' x ' y ', set up facial image rectangular coordinate system Oxy；

B, Face datection based on Adaboost algorithm, train strong classifier in detection training part, set n training sample (x₁,y₁),...,(x_i,y_i),..(x_n,y_n), y_iDesirable 0 and 1, y_i=1 represents face sample, y_i=0 represents non-face sample, initialization weights omega:

The weight of face sample

${\mathrm{\ω}}_{face}=\frac{1}{2m}---\left(1\right)$

Wherein m is the number of face sample,

The weight of non-face sample

${\mathrm{\ω}}_{nonface}=\frac{1}{2l}---\left(2\right)$

Wherein l is the number of non-face sample,

For t=1:T,

1. weight normalization

$W_{t,i}=\frac{{\mathrm{\ω}}_{t,i}}{{\Σ}_{j=1}^n{\mathrm{\ω}}_{t,j}}---\left(3\right)$

2. the Weak Classifier h of optimum is filtered out_t(x_k,p_k,f_k,θ_k), its classification error rate is:

${\mathrm{\ϵ}}_t={\min }_{f_k,p_k,{\mathrm{\θ}}_k}\underset{i}{\Σ}{W}_{t,i}|h_t(x_k,p_k,f_k,{\mathrm{\θ}}_k)-y_i|---\left(4\right)$

Wherein x_kRepresent a detection subwindow, p_kThe direction of the instruction sign of inequality, f_kIt is characterized, θ_kFor

Threshold value.

3. weights are updated

${\mathrm{\ω}}_{t+1,i}={\mathrm{\ω}}_t{\mathrm{\β}}_t^{1-e_i}={\mathrm{\ω}}_t\frac{{\mathrm{\ϵ}}_t}{1-{\mathrm{\ϵ}}_t}---\left(5\right)$

WhereinRepresent classification error and account for the ratio that classification is correct, wherein e_iDesirable 0 and 1.Work as e_iWhen=0, x_iCorrectly classified；Work as e_iWhen=1, x_iClassified by mistake,

Last strong classifier is:

Wherein

C, face eyes based on hierarchy plan and the location of mouth:

The calculating of the eigenvalue of characteristic pattern based on small echo, in characteristic pattern based on small echo, the computing formula of the n-th eigenvalue Sal (n) is:

$Sal\left(n\right)=\underset{k=1}{\overset{-j}{\Σ}}\left|{\mathrm{\ω}}_k{C}^{\left(k\right)}\left(W_{2^j}f\right(n\left)\right)\right|,0\≤n\≤2^{j}N,-{\log }_{2}N\≤j\≤-1---\left(7\right)$

Wherein weights omega_kIt is allocated to the wavelet coefficient of the different scale of maximum, is again reciprocal coefficient of standard deviation,

Weights omega_kComputing formula be:

${\mathrm{\ω}}_k=\frac{1}{{\mathrm{\σ}}_k}=\sqrt{\frac{1}{S}\underset{z=1}{\overset{S}{\Σ}}{\left(W_{2^k}f\right(z)-\frac{1}{S}\underset{z=1}{\overset{S}{\Σ}}|W_{2^k}f\left(z\right)\left|\right)}^2}---\left(8\right)$

Wherein S is the maximum wavelet coefficient quantity at k level,It is the greatest coefficient in 0≤z≤S,

The check and correction in eyes region based on principal component analysis:

Owing to eyes region is a rectangular area, it is made up of M column vector so eyes region can be regarded as, is expressed as Γ₁,Γ₂,..,Γ_M,

The meansigma methods of eyes zone sample is:

$\mathrm{\Ψ}=\frac{1}{M}\underset{n=1}{\overset{M}{\Σ}}{\mathrm{\Γ}}_n---\left(9\right)$

Difference between each eyes region and meansigma methods is:

φ_n=Γ_n-Ψ (10)

The covariance matrix in eyes region is:

$C=\frac{1}{M}\underset{n=1}{\overset{M}{\Σ}}{\mathrm{\φ}}_n{\mathrm{\φ}}_n^T={\mathrm{AA}}^T---\left(11\right)$

Wherein A=[φ₁,φ₂,...,φ_M],

The amount of calculation of the characteristic vector directly calculating Matrix C in reality is very big, can pass through matrix A A^TCharacteristic vector v_nDefinition calculate:

A^TAv_n=λ_nv_n (12)

Formula (13) both sides premultiplication matrix A can be obtained:

AA^TAv_n=A λ_nv_n (13)

From formula (14) formula, Av_nIt is the characteristic vector of Matrix C,

Thus it is possible to use characteristic vector μ_nIt is expressed as the linear combination of M sample:

${\mathrm{\μ}}_n=\underset{k=1}{\overset{M}{\Σ}}{v}_{nk}{\mathrm{\φ}}_k={\mathrm{Av}}_n---\left(14\right)$

Wherein n=1,2 ..., M,

It is expressed as Γ₁,Γ₂,..,Γ_MEyes region projection in feature space:

${\mathrm{\ω}}_n={\mathrm{\μ}}_n^T(T_n-\mathrm{\ψ})---\left(15\right)$

Wherein n=1,2 ..., M,

Weight one vector of composition of each characteristic vector in eyes region:

Ω^T={ ω₁,ω₂,...,ω_n} (16)

The difference of two squares between eyes region and reconstruct image is:

${\mathrm{\ϵ}}^2={||{\mathrm{\Γ}}_n-{\mathrm{\Γ}}_f||}^2=||{\mathrm{\Γ}}_n-(\mathrm{\ψ}+{\Σ}_{n=1}^M{\mathrm{\ω}}_n{\mathrm{\μ}}_n||---\left(17\right)$

Calculated the difference of two squares by formula (18) and just can obtain the eyes region of minimal reconstruction error.

Eyes based on eye template and the location of mouth:

Eye template is made up of three squares, and the black patch at center represents the pupil of black in human eye, is expressed as B, and this black bars left and right sides is white square, is expressed as W₁And W₂。

The definition of the similar function SIM for measuring eye template is:

$SIM=\frac{I_{W_1}+I_{W_2}}{I_B+\mathrm{\β}}---\left(18\right)$

WhereinFor region W₁The summation of intensity,For region W₂The summation of intensity, I_BFor the summation of the intensity of region B, β is a positive number avoiding denominator to be zero,

Q eye template in face sample, optimal eye position (x_best,y_best) it is defined as follows:

x_best=median (x₁,x₂,..x_Q) (19)

y_best=median (y₁,y₂,..y_Q) (20)

The calculating being partially worth in d, facial image:

Horizontal off set and the calculating of vertical offset: set a width of width of facial image window of facial image rectangular coordinate system O ' x ' y '₁, a height of height₁, a width of width of initial pictures₂, a height of height₂, the initial point of coordinate system Oxy is (face at coordinate system O ' x ' y '_x,face_y), the horizontal offset of coordinate system Oxy relative coordinate system O ' x ' y ' is Δ w, and vertical offset is Δ h, therefore minimum horizontal off set Δ level_minWith vertical offset Δ vertical_minIt is respectively as follows:

$\begin{array}{c}{\mathrm{\Δlevel}}_{\min }=\min (\frac{{\mathrm{width}}_1}{2}+{\mathrm{face}}_x-1.5{\mathrm{width}}_1,\frac{{\mathrm{width}}_1}{2}+{\mathrm{face}}_x-\frac{{\mathrm{width}}_2}{2},\\ \frac{{\mathrm{width}}_1}{2}+{\mathrm{face}}_x+\frac{{\mathrm{width}}_2}{2}-1.5{\mathrm{width}}_1)\end{array}---\left(21\right)$

$\begin{array}{c}{\mathrm{\Δvertical}}_{\min }=\min (\frac{{\mathrm{height}}_1}{2}+{\mathrm{face}}_y-1.2{\mathrm{height}}_1,\frac{{\mathrm{height}}_1}{2}+{\mathrm{face}}_y-\frac{{\mathrm{height}}_2}{2},\\ \frac{{\mathrm{height}}_1}{2}+{\mathrm{face}}_y+\frac{{\mathrm{height}}_2}{2}-1.2{\mathrm{height}}_1)\end{array}---\left(22\right)$

The calculating of side deflection angle:

If left eye centre coordinate is (x₁,y₁), right eye centre coordinate is (x₂,y₂), the centre coordinate of mouth is (x₃,y₃), the horizontal range of two eyes is l, and vertical dimension is d, during the deflection of facial image side, puts horizontal line and be respectively l with the distance of point in the eyes of left and right two under nose₁And l₂, putting horizontal line under nose with the distance of point in mouth is l₃。

The plane deflection angle α of facial image:

$\mathrm{\α}=arctan\frac{d}{l}=arctan\frac{y_2-y_1}{x_2-x_1}---\left(23\right)$

Point and the differential seat angle β of left and right two ophthalmic point under nose₁:

${\mathrm{\β}}_1=\arctan \frac{d}{l_1}-\arctan \frac{d}{l_2}=\arctan \frac{y_2-y_1}{\frac{{\mathrm{width}}_1}{2}-x_1}-\arctan \frac{y_2-y_1}{x_2-\frac{{\mathrm{width}}_1}{2}}---\left(24\right)$

Point and the differential seat angle β of point in mouth under nose₂:

${\mathrm{\β}}_2=arctan\frac{d}{l_3}=arctan\frac{y_2-y_1}{\frac{{\mathrm{width}}_1}{2}-x_3}---\left(25\right)$

The side deflection angle β of facial image:

$\begin{array}{c}\mathrm{\β}=\frac{{\mathrm{\β}}_1+{\mathrm{\β}}_2}{2}\\ =\frac{\arctan \frac{y_2-y_1}{\frac{{\mathrm{width}}_1}{2}-x_1}-\arctan \frac{y_2-y_1}{x_2-\frac{{\mathrm{width}}_1}{2}}+\arctan \frac{y_2-y_1}{\frac{{\mathrm{width}}_1}{2}-x_3}}{2}\end{array};---\left(26\right)$

E, controller 10 control camera lift-launch platform 6 and move horizontally Δ level₁, vertically move Δ vertical₁, rotary electric machine 7 rotational angle β, then control camera shooting.

Claims (4)

1. null1. the adaptive controller of a man face image acquiring，It is characterized in that: include support (100)，Described support (100) is provided with at least two vertical guide being vertically arranged (1)，Described support (100) is provided with vertical power transmission shaft (2)，The vertical slide block (3) that vertically guide rail (1) is vertically slidable it is arranged with on described vertical power transmission shaft (2)，It is provided with cross slide way (4) between described vertical slide block (3)，Described cross slide way (4) is provided with horizontal power transmission shaft (5)，It is arranged with the camera that transversely guide rail (4) slides laterally on described horizontal power transmission shaft (5) and carries platform (6)，Described camera carries platform (6) and is provided with rotary electric machine (7)，Described rotary electric machine (7) rotating shaft is vertically arranged，Described rotary electric machine (7) rotating shaft is provided with the camera with axis of rotation.
2. The adaptive controller of a kind of man face image acquiring the most according to claim 1, it is characterized in that: described vertical power transmission shaft (2) is provided with vertical power transmission shaft screw thread, described vertical power transmission shaft (2) one end is provided with the first motor (8), described vertical power transmission shaft (2) is with the first motor (8) axis of rotation, and described vertical slide block (3) socket place is provided with the vertical slide block screw thread rotated with vertical power transmission shaft (2) and make vertical slide block (3) vertically move with vertical power transmission shaft threaded engagement.
3. The adaptive controller of a kind of man face image acquiring the most according to claim 1, it is characterized in that: described horizontal power transmission shaft (5) is provided with horizontal power transmission shaft screw thread, described horizontal power transmission shaft (5) one end is provided with the second motor (9), described horizontal power transmission shaft (5) is provided with the camera rotated with horizontal power transmission shaft (5) with horizontal power transmission shaft threaded engagement and make camera carry platform (6) transverse shifting carries platform screw thread with the second motor (9) axis of rotation, described camera lift-launch platform (6) socket place.
4. The adaptive controller of a kind of man face image acquiring the most according to claim 1, it is characterized in that: on described support (100), be additionally provided with controller (10), described controller (10) connects the first motor (8), second motor (9), it is arranged at first encoder (11) of vertical power transmission shaft (2) other end, it is arranged at second encoder (12) of horizontal power transmission shaft (4) other end, rotary electric machine (7) and camera, described first motor (8) is connected with the first encoder (11), described second motor (9) is connected with the second encoder (12).