CN101819687A - Face recognition student attendance device and method - Google Patents

Abstract

The invention discloses a face recognition student attendance device and a face recognition student attendance method, and belongs to the technical field of image recognition. The device comprises a hand-held terminal, a front-end processor, a database server and a camera, wherein the hand-held terminal is respectively connected with the camera and the database server; the camera is connected with the front-end processor; and the front-end processor is connected with the database server. According to different sizes of classrooms, different cameral moving control strategies are applied, active coordination of a person to be checked is not needed, and normal order of classes is not influenced in real sense, so the efficiency is improved, and substitute reply is avoided.

Description

A kind of face recognition student attendance device and method

Technical field

The invention belongs to the image recognition technology field, particularly a kind of face recognition student attendance device and method.

Background technology

Usually the work attendance mode comprises that attendance recorder, people are modes such as work attendance, at present, the college student adoptable attendance recorder of work attendance of giving a lesson comprises card attendance recorder and finger-print type attendance recorder, all there is following shortcoming in this dual mode: this dual mode all needs the people for cooperating with on one's own initiative, cause the manual operation mistake easily, wherein the card of card attendance recorder is lost easily, and the finger-print type attendance recorder occurs easily lining up and uses the situation of attendance recorder, delay the class period, influence efficient; There is following problem in the mode that the teacher calls the roll: the student gives a lesson at hundreds of people's lyceum, occurs the situation that attendance recorder is used in queuing easily, delay the class period and be difficult to and adhere to, and the stream of people that goes to school and leaves school is bigger, delays the class period, influences efficient.

Summary of the invention

For solving the deficiency of above work attendance mode, the present invention proposes a kind of face recognition student attendance device and method, utilizes this Work attendance device and Work attendance method, raises the efficiency to reach, and reduces for answering, and finishes the purpose of work attendance automatically.

Technical scheme of the present invention is achieved in that face recognition student attendance device includes: handheld terminal, video camera, front-end processor and database server, wherein, handheld terminal is connected with database server with front-end processor respectively, video camera is connected with front-end processor, and front-end processor is connected with database server;

A kind of face recognition student attendance method of the present invention comprises that step is as follows:

Step 1: handheld terminal sends to front-end processor with curriculum information and work attendance request;

Step 2: front-end processor stores the curriculum information of receiving in the buffer memory into after receiving the work attendance request, and starts video camera;

Step 3: judge whether it is lyceum, then carry out following steps in this way: video camera starts the back and cruises by the k that a sets in advance presetting bit, take since the 1st presetting bit, stop ts and capture image in each presetting bit, enter next presetting bit after depositing in picture in the buffer memory, repeat the work of first presetting bit, after k presetting bit of traversal, finish presetting bit and cruise; And send shutdown signal to video camera; Execution in step 5;

Step 4: judge whether it is little classroom, then carry out following steps in this way:

1) opens video camera, start the presetting bit cruise function;

2) video camera is taken since the 1st presetting bit, stops ts and captures image in each presetting bit, and image is deposited in the buffering;

3) determine the position of people's face in picture: utilize complexion model with the picture binaryzation, area of skin color and non-area of skin color are separated; Calculate through denoising and morphology, obtain the colour of skin and be communicated with the district;

4) judge whether the picture edge exists incomplete people's face, if there is incomplete people's face, then execution in step 5), otherwise execution in step 6);

Wherein, judge that the rule of the imperfect people's face of picture marginal existence is:, judge that then the picture coboundary contains incomplete people's face if be communicated with the district through the picture coboundary and without left hand edge; If being communicated with the district has passed through the picture right hand edge and without lower limb, has judged that then the picture right hand edge contains incomplete people's face;

5) determine that video camera horizontally rotates adjustment amount α and pitching adjustment amount β, adjust the shooting angle of video camera, gather complete facial image;

Concrete grammar is as follows: at first calculate each colour of skin connection district, edge area in the picture, calculate with the pixel number that covers; Keep the area of skin color of the maximum of passing through coboundary and pass through the area of skin color of the maximum of right hand edge, calculate their barycenter respectively, and these two coordinate records are got off;

Center-of-mass coordinate is based on the image pixel coordinate system, and image pixel coordinate and image physical coordinates are changed:

If barycenter mark be (u, v), unit is a pixel, the image physical coordinates be (x, y), unit be millimeter, initial point is the intersection point of the lens axis and the plane of delineation,

(u, v) the plane and (x, y) pass on plane is:

$x=\frac{u-u_0}{c},$ $y=\frac{v-v_0}{c}---\left(1\right)$

Be expressed in matrix as:

$\left[\begin{array}{c}x\\ \mathrm{<figure-callout\; id="1"\; label="y"\; filenames="HSA00000085835200031.png,HSA00000085835200041.png"\; state="\{\{state\}\}">y</figure-callout>}\\ 1\end{array}\right]=\frac{1}{\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="HSA00000085835200031.png,HSA00000085835200041.png"\; state="\{\{state\}\}">c</figure-callout>}}\left[\begin{array}{ccc}1& 0& -u_0\\ 0& 1& -v_0\\ 0& 0& 1\end{array}\right]\left[\begin{array}{c}u\\ v\\ 1\end{array}\right]---\left(2\right)$

Wherein c is 1 millimeter pixel value that image contains;

Video camera adopts the perspective projection model, and process is as follows:

If people's face center is that (dx, dy), then the adjustment amount that horizontally rotates of video camera is α=∠ O to the skew of picture centre on the present frame ₁OP ', pitching adjustment amount are β=∠ POP ', and according to triangle relation, α and β should satisfy

$\tan \mathrm{\&alpha;}=\frac{\mathrm{dx}}{f},$ $\tan \mathrm{\&beta;}=\frac{\mathrm{dy}}{f}---\left(3\right)$

Wherein, O is the camera optics center; Ray OO ₁Be camera optical axis; Line segment OO ₁Length be focus of camera f; P is the projection of target on the picture plane;

6) video camera is taken pictures in new presetting bit, and picture is kept in the buffer memory; Execution in step 3);

Step 5: front-end processor is conducted oneself one by one to the picture that collects, and face detects and face recognition algorithms is calculated; The accessing database server compares with picture in the database server, and front-end processor calculates and contains certain classmate in the video camera picture shot, and then this classmate's work attendance labelled amount is set to 1;

Step 6: use the handheld terminal accessing database, obtain staff list absent from duty, the teacher checks personnel absent from duty by this list, if any erroneous judgement, revises the work attendance mark of personnel absent from duty in the database by manual methods, further guarantees the work attendance accuracy.

The described lyceum of step 3 is meant the classroom that the above seat of hundred people is fixing, and the described little classroom of step 4 is meant less than unfixed classroom, hundred people's seat.When installing, equipment manually selects to be provided with by actual conditions.

Advantage of the present invention: the present invention varies in size according to the classroom, uses different video cameras and moves control strategy, has really realized not needing to be cooperated with on one's own initiative by the fieldworker, does not influence the order of normally going to school and leaving school, and then has improved efficient, avoids for the generation of answering phenomenon.

Description of drawings

Fig. 1 is a face recognition student attendance device position view of the present invention;

Fig. 2 is a face recognition student attendance device database server Attendance Sheet attributed graph of the present invention;

Fig. 3 is a face recognition student attendance device front-end processor structured flowchart of the present invention;

Fig. 4 is face recognition student Work attendance method overall flow figure of the present invention;

Fig. 5 is a face recognition student attendance device front-end processor workflow diagram of the present invention;

Fig. 6 is face recognition student Work attendance method lyceum strategic process figure of the present invention;

Fig. 7 is face recognition student Work attendance method primary school teacher chamber strategic process figure of the present invention;

Whether Fig. 8 exists people's face computing method process flow diagram for face recognition student Work attendance method of the present invention edge;

Fig. 9 is communicated with the cell relation exemplary plot for face recognition student Work attendance method of the present invention presetting bit with the edge colour of skin;

Figure 10 is that face recognition student Work attendance method image pixel coordinate system of the present invention and image physical coordinates are transformational relation figure;

Figure 11 calculates synoptic diagram for face recognition student Work attendance method video camera adjustment amount of the present invention;

Figure 12 adjusts front and back shooting area comparison diagram for face recognition student Work attendance method video camera of the present invention.

Embodiment

Below in conjunction with drawings and Examples the present invention is described in further detail.

Fig. 1 is a face recognition student attendance device.Wherein, handheld terminal is carried by the teacher, and video camera and front-end processor are placed on the blackboard top position, and database server is placed on point of presence.

Each component function is as follows:

Handheld terminal: the information input sends the work attendance request, the accessing database server;

Video camera: images acquired information.When video camera is installed, manually configure the presetting bit and the image acquisition strategy of video camera in advance according to the big or small concrete condition in classroom;

Front-end processor: 1) the video camera presetting bit is set manually, and preserves each presetting bit focal length; 2) unlatching of control video camera, close, the horizontally rotating of position, preset grabgraf and The Cloud Terrace, luffing angle adjustment; 3) receive the information that handheld terminal sends; 4) people's face and detection algorithm calculate; 5) accessing database;

Database server: be provided with Attendance Sheet in the database server, comprise course code, teacher's numbering, time, student name, student's student number, work attendance mark in the Attendance Sheet, its middle school student's student number comprises that student information, student information comprise name, student number and photo, as shown in Figure 2.

Fig. 3 is a face recognition student attendance device front-end processor structured flowchart of the present invention, front-end processor comprises kernel processor chip, buffer memory and communication interface, the input end of kernel processor chip is connected with video camera, and the output terminal of kernel processor chip is connected with video camera, buffer memory and communication interface.

Fig. 4 and Fig. 5 are face recognition student Work attendance method overall flow, comprise the steps:

Step 1: handheld terminal sends curriculum information and work attendance request to front-end processor: the teacher uses the keypad on the handheld terminal, input course code name and class period, wherein, class period is not used concrete Hour Minute Second, adopt the course joint number to replace, for example: on March 3rd, 2010, first class can be expressed as: 2010030301;

Step 2: front-end processor stores the curriculum information of receiving in the buffer memory into after receiving the work attendance request, starts video camera;

Step 3: whether judge course at lyceum, then adopt lyceum strategy, execution in step 5 in this way; Otherwise execution in step 4;

Step 4: judge that course whether in little classroom, then adopts primary school teacher chamber strategy, execution in step 5 in this way;

Step 5: front-end processor is conducted oneself one by one to the picture that collects, and face detects and face recognition algorithms is calculated; The accessing database server compares with picture in the database server, and front-end processor calculates and contains certain classmate in the video camera picture shot, and then this classmate's work attendance labelled amount is set to 1;

Step 6: use the handheld terminal accessing database, obtain staff list absent from duty, the teacher checks personnel absent from duty by this list, if any erroneous judgement, revises the work attendance mark of personnel absent from duty in the database by manual methods, further guarantees the work attendance accuracy.

Wherein, the described lyceum strategy of step 3 as shown in Figure 6, detailed process is as follows: at first default k presetting bit, token variable i assignment is 1, and described presetting bit is set when mounted, provides figure film source preferably for detecting with face recognition algorithms for people's face, people's face can not be too much in every pictures, be that presetting bit can not be very little, present embodiment selects to support the The Cloud Terrace of 128 presetting bits, and 4 people's faces are taken in each presetting bit; Video camera stopped 4 seconds in this presetting bit, front-end processor was grabbed a frame figure in per 1 second and is kept among the flash, and after past 4s image acquisition time, i adds 1 with token variable, enter next default position, repeat the work that first presetting bit is done, by that analogy, travel through all k presetting bit, after presetting bit is cruised and is finished, be that front-end processor detects i=k+1, front-end processor sends shutdown signal to video camera, and video camera cuts out.

Primary school teacher chamber strategy described in the step 4 is shown in Fig. 7 figure～12.

At first, the characteristics in little classroom are: the seat is the common desk and chair that can move, and promptly the seat distributes and do not fix, and the position that thereupon causes people's face to occur is also unexpected.Therefore as lyceum, only take the video camera preset point to cruise, with the route method for scanning, following problem may occur: two preset point overlay areas appear crossing in people's face, promptly people's face part is on the picture that the i preset point is gathered, another part appears on the picture of j preset point collection, and the result causes not discern this people's face in i preset point and j preset point.

Therefore on the basis that presetting bit is cruised, add step S701 and step S702, as shown in Figure 7.Fig. 8 is described further step S701: step S701: utilize and judge based on the method for detecting human face of the colour of skin whether the picture edge has people's face, if have, then video camera is adjusted to the position of marginal man's face, to photograph the complete image of the people's face that is in the edge originally, concrete grammar is as follows:

Step 1: picture is done image segmentation based on the colour of skin, promptly utilize complexion model the picture binaryzation, area of skin color and non-area of skin color is separated;

Step 2: calculate through denoising and morphology, obtain the colour of skin and be communicated with the district;

Step 3: judge whether the picture edge exists imperfect people's face, and judgment rule is: the colour of skin is communicated with the district and has passed through coboundary and without left hand edge, judged that then coboundary contains incomplete people's face; The colour of skin is communicated with the district and has passed through right hand edge, and does not pass through lower limb, then is judged as right hand edge and contains incomplete people's face; Here mainly be to detect accurately and identification in order to collect more images more accurately, not relate to, therefore having ignored this colour of skin connection district may be the influence of large-area health area of skin color.

The people's face that is in lower limb in this preset point must be in the coboundary of adjacent preset point, the people's face that is in left hand edge in this preset point must be in the right hand edge of adjacent preset point, therefore, in order to avoid repeating work as far as possible, do not consider left hand edge and lower limb, in Fig. 9, provide presetting bit and be communicated with the relation of distinguishing with the edge colour of skin, suppose current the 5th presetting bit that be in, picture shot has been determined several colours of skin district through Face Detection, the position in colour of skin district is shown in ellipse among the figure, the presetting bit that numeral in the ellipse is made a response to this marginal man's face, presetting bit 5 positions, only the colour of skin district to right hand edge and coboundary makes a decision, and the colour of skin district of left hand edge and lower limb also can not be omitted, and can make the scanning reaction to it by other presetting bits;

Step 4: calculate each edge colour of skin area, i.e. pixel number of Fu Gaiing, the area of skin color (if any) of the area of skin color (if any) of the maximum of reservation process coboundary and the maximum of process right hand edge, calculate their barycenter respectively, these two coordinate records are got off, these two coordinates might be same, enter video camera adjustment amount computation process S702 step.

Figure 10～Figure 12 is described further step S702: the area of skin color that the photocentre of video camera is moved to the edge, be in the people's at edge complete facial image with collection, calculate barycenter among the step S701 and be based on the image pixel coordinate system, horizontally rotate adjustment amount α and pitching adjustment amount β in order to calculate video camera, two coordinate systems need be done conversion, as shown in figure 10, the image pixel coordinate system is made as (u, v), unit is a pixel, and the image physical coordinates is (x, y), unit is a millimeter, and initial point is the intersection point of the lens axis and the plane of delineation

(u, v) the plane and (x, y) pass on plane is:

$x=\frac{u-u_0}{c},$ $y=\frac{v-v_0}{c}---\left(1\right)$

Be expressed in matrix as:

$\left[\begin{array}{c}x\\ \mathrm{<figure-callout\; id="1"\; label="y"\; filenames="HSA00000085835200031.png,HSA00000085835200041.png"\; state="\{\{state\}\}">y</figure-callout>}\\ 1\end{array}\right]=\frac{1}{\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="HSA00000085835200031.png,HSA00000085835200041.png"\; state="\{\{state\}\}">c</figure-callout>}}\left[\begin{array}{ccc}1& 0& -{\mathrm{<figure-callout\; id="0"\; label="u"\; filenames="HSA00000085835200031.png,HSA00000085835200061.png"\; state="\{\{state\}\}">u</figure-callout>}}_0\\ 0& 1& -{\mathrm{<figure-callout\; id="0"\; label="v"\; filenames="HSA00000085835200031.png,HSA00000085835200061.png"\; state="\{\{state\}\}">v</figure-callout>}}_0\\ 0& 0& 1\end{array}\right]\left[\begin{array}{c}u\\ v\\ 1\end{array}\right]---\left(2\right)$

Wherein c is 1 millimeter pixel value that image contains.

Camera model adopts the Pin-hole model that extensively adopts in the computer vision, i.e. perspective projection model, and it is equivalent to the thin lens imaging physically, as shown in figure 11:

If people's face center is that (dx, dy), then the adjustment amount that horizontally rotates of video camera is α=∠ O to the skew of picture centre on the present frame ₁OP ', pitching adjustment amount are β=∠ POP ', and according to triangle relation, α and β should satisfy

$\tan \mathrm{\&alpha;}=\frac{\mathrm{dx}}{f},$ $\tan \mathrm{\&beta;}=\frac{\mathrm{dy}}{f},---\left(3\right)$

Wherein, O is the camera optics center; Ray OO ₁Be camera optical axis; Line segment OO ₁Length be focus of camera f; P is the projection of target on the picture plane; Camera Positioning is converted into to the problem of people's face, with camera optical axis OO ₁Move to ray OP position, be about to initial point O ₁Move to the problem that P is ordered.

The inventive method is introduced feedback mechanism to the calculating of f, and promptly according to the results modification f of last time to the camera angle adjustment, this feedback mechanism can make the rotation of platform more steady.

Figure 12 has provided video camera and has adjusted relatively synoptic diagram of front and back shooting area.

Claims (2)

1. face recognition student attendance device, comprise handheld terminal, front-end processor and database server, it is characterized in that: this Work attendance device also comprises video camera, handheld terminal is connected with database server with video camera respectively, video camera is connected with front-end processor, and front-end processor is connected with database server.

2. the face recognition student attendance method of the described a kind of face recognition student attendance device of claim 1 is characterized in that: may further comprise the steps:

Step 1: handheld terminal sends to front-end processor with curriculum information and work attendance request;

Step 2: front-end processor stores the curriculum information of receiving in the buffer memory into after receiving the work attendance request, and starts video camera;

Step 3: judge whether it is lyceum, then carry out following steps in this way: video camera starts the back and cruises by the k that a sets in advance presetting bit, take since the 1st presetting bit, stop ts and capture image in each presetting bit, enter next presetting bit after depositing in picture in the buffer memory, repeat the work of first presetting bit, after k presetting bit of traversal, finish presetting bit and cruise; And send shutdown signal to video camera; Execution in step 5;

Step 4: judge whether it is little classroom, then carry out following steps in this way:

1) opens video camera, start the presetting bit cruise function;

2) video camera is taken since the 1st presetting bit, stops ts and captures image in each presetting bit, and image is deposited in the buffering;

3) determine the position of people's face in picture: utilize complexion model with the picture binaryzation, area of skin color and non-area of skin color are separated; Calculate through denoising and morphology, obtain the colour of skin and be communicated with the district;

4) judge whether the picture edge exists incomplete people's face, if there is incomplete people's face, then execution in step 5), otherwise execution in step 6);

Wherein, judge that the rule of the imperfect people's face of picture marginal existence is:, judge that then the picture coboundary contains incomplete people's face if be communicated with the district through the picture coboundary and without left hand edge; If being communicated with the district has passed through the picture right hand edge and without lower limb, has judged that then the picture right hand edge contains incomplete people's face;

5) determine that video camera horizontally rotates adjustment amount α and pitching adjustment amount β, adjust the shooting angle of video camera, gather complete facial image;

Concrete grammar is as follows: at first calculate each colour of skin connection district, edge area in the picture, calculate with the pixel number that covers; Keep the area of skin color of the maximum of passing through coboundary and pass through the area of skin color of the maximum of right hand edge, calculate their barycenter respectively, and these two coordinate records are got off;

Center-of-mass coordinate is based on the image pixel coordinate system, and image pixel coordinate and image physical coordinates are changed:

If barycenter mark be (u, v), unit is a pixel, the image physical coordinates be (x, y), unit be millimeter, initial point is the intersection point of the lens axis and the plane of delineation,

(u, v) the plane and (x, y) pass on plane is:

$x=\frac{u-u_0}{c},y=\frac{v-v_0}{c}---\left(1\right)$

Be expressed in matrix as:

$\left[\begin{array}{c}x\\ y\\ 1\end{array}\right]=\frac{1}{c}\left[\begin{array}{ccc}1& 0& -u_0\\ 0& 1& -v_0\\ 0& 0& 1\end{array}\right]\left[\begin{array}{c}u\\ v\\ 1\end{array}\right]---\left(2\right)$

Wherein c is 1 millimeter pixel value that image contains;

Video camera adopts the perspective projection model, and process is as follows:

If people's face center is that (dx, dy), then the adjustment amount that horizontally rotates of video camera is α=∠ O to the skew of picture centre on the present frame ₁OP ', pitching adjustment amount are β=∠ POP ', and according to triangle relation, α and β should satisfy

$\tan \mathrm{\&alpha;}=\frac{\mathrm{dx}}{f},\tan \mathrm{\&beta;}=\frac{\mathrm{dy}}{f}---\left(3\right)$

Wherein, O is the camera optics center; Ray OO ₁Be camera optical axis; Line segment OO ₁Length be focus of camera f; P is the projection of target on the picture plane;

6) video camera is taken pictures in new presetting bit, and picture is kept in the buffer memory; Execution in step 3);

Step 5: front-end processor is conducted oneself one by one to the picture that collects, and face detects and face recognition algorithms is calculated; The accessing database server compares with picture in the database server, and front-end processor calculates and contains certain classmate in the video camera picture shot, and then this classmate's work attendance labelled amount is set to 1;

Step 6: use the handheld terminal accessing database, obtain staff list absent from duty, the teacher checks personnel absent from duty by this list, if any erroneous judgement, revises the work attendance mark of personnel absent from duty in the database by manual methods, further guarantees the work attendance accuracy.

Images