CN109635766A - The face of convolutional neural networks based on small sample is taken pictures Work attendance method and system - Google Patents
The face of convolutional neural networks based on small sample is taken pictures Work attendance method and system Download PDFInfo
- Publication number
- CN109635766A CN109635766A CN201811563353.7A CN201811563353A CN109635766A CN 109635766 A CN109635766 A CN 109635766A CN 201811563353 A CN201811563353 A CN 201811563353A CN 109635766 A CN109635766 A CN 109635766A
- Authority
- CN
- China
- Prior art keywords
- attendance
- convolutional neural
- neural networks
- human face
- personnel
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 107
- 238000000034 method Methods 0.000 title claims abstract description 58
- 239000011159 matrix material Substances 0.000 claims description 36
- 238000012549 training Methods 0.000 claims description 26
- 238000005070 sampling Methods 0.000 claims description 12
- 238000000605 extraction Methods 0.000 claims description 9
- 230000014509 gene expression Effects 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 9
- 238000012952 Resampling Methods 0.000 claims description 6
- 230000008921 facial expression Effects 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 6
- 239000000284 extract Substances 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 claims 1
- 230000006399 behavior Effects 0.000 claims 1
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 8
- 238000012360 testing method Methods 0.000 description 7
- 230000001537 neural effect Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 241000288673 Chiroptera Species 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- XXQCMVYBAALAJK-UHFFFAOYSA-N ethyl n-[4-[benzyl(2-phenylethyl)amino]-2-(2-phenylethyl)-1h-imidazo[4,5-c]pyridin-6-yl]carbamate Chemical compound N=1C=2C(N(CCC=3C=CC=CC=3)CC=3C=CC=CC=3)=NC(NC(=O)OCC)=CC=2NC=1CCC1=CC=CC=C1 XXQCMVYBAALAJK-UHFFFAOYSA-N 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011056 performance test Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/10—Office automation; Time management
- G06Q10/109—Time management, e.g. calendars, reminders, meetings or time accounting
- G06Q10/1091—Recording time for administrative or management purposes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/20—Education
- G06Q50/205—Education administration or guidance
-
- G—PHYSICS
- G07—CHECKING-DEVICES
- G07C—TIME OR ATTENDANCE REGISTERS; REGISTERING OR INDICATING THE WORKING OF MACHINES; GENERATING RANDOM NUMBERS; VOTING OR LOTTERY APPARATUS; ARRANGEMENTS, SYSTEMS OR APPARATUS FOR CHECKING NOT PROVIDED FOR ELSEWHERE
- G07C1/00—Registering, indicating or recording the time of events or elapsed time, e.g. time-recorders for work people
- G07C1/10—Registering, indicating or recording the time of events or elapsed time, e.g. time-recorders for work people together with the recording, indicating or registering of other data, e.g. of signs of identity
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Educational Administration (AREA)
- Data Mining & Analysis (AREA)
- Educational Technology (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Primary Health Care (AREA)
- Life Sciences & Earth Sciences (AREA)
- Operations Research (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Quality & Reliability (AREA)
- General Engineering & Computer Science (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
- Collating Specific Patterns (AREA)
Abstract
It takes pictures Work attendance method and system the present invention provides a kind of face of convolutional neural networks based on small sample, a kind of face of the convolutional neural networks based on small sample is taken pictures Work attendance method: having initially set up the face database to attendance personnel, and establish suitable convolutional neural networks model, then convolutional neural networks are trained, and identified using the photo that trained convolutional neural networks treat attendance personnel, obtain corresponding ID number;Finally using obtained ID number, attendance data is counted, one is generated and records the Excel table for needing attendance attendance data.The beneficial effects of the present invention are: technical solution proposed by the invention proposes that a kind of new face is taken pictures Work attendance method, devises a set of face based on convolutional neural networks and take pictures attendance checking system, it is practical for the attendance of intelligentized record student.
Description
Technical field
The present invention relates to field of image processings more particularly to a kind of face of the convolutional neural networks based on small sample to take pictures
Work attendance method and system.
Background technique
Colleges and universities' number is numerous, its management difficulty is very big, and it is one very heavy how to realize that more intelligentized management has become
The research topic wanted, the attendance rate that student attends class have reacted the attitude towards study and basic quality of the life, turn out for work total direct
Influence whether the end of term is eligible for this examination and marking of the teacher to student's usual performance, the rate of attendance of student are with the life
No can smoothly graduate has close relationship.So the record work of student attendance is most important.
In school, teacher mainly takes attendance in class realizations to the attendance record work of student by teacher, in the compression of nowadays course
And class attend class number it is more in the case where, in this way, waste time, and replace answering due to being likely to occur student
Situation, cannot be truly reflected completely student to class situation and ensure attendance accuracy and the transparency.Therefore, for student
The record of attendance and statistical work are difficult.
To solve the above-mentioned problems, simpler and intelligent record student attendance, in view of convolutional neural networks in machine
Excellent performance in study, therefore the invention proposes a kind of face attendance camera system based on convolutional neural networks, teacher
Called when needing to record attendance everybody rectify it is seated clap a group picture, multiple classes are taken a group photo input Face datection when the end of term
Face therein is cut out with extraction module, the human face photo extracted is then inputted into convolutional neural networks module again
It is identified, the statistical module for eventually passing through student attendance number automatically generates Excel Attendance Sheet, always examines in this term of statistic
Diligent number.
Summary of the invention
To solve the above-mentioned problems, it takes pictures and examines the present invention provides a kind of face of convolutional neural networks based on small sample
Diligent method and system, a kind of face of the convolutional neural networks based on small sample are taken pictures Work attendance method, are mainly comprised the steps that
S101: the human face photo library to attendance personnel is established;It is described have to attendance personnel it is multiple;In the human face photo library
Comprising all facial expression photos to attendance personnel, the mug shot of multiple different expressions is each corresponding with to attendance personnel,
And a unique ID number is each corresponding with to attendance personnel;
S102: it establishes based on the improved convolutional neural networks for adapting to face attendance of LeNet-5, and is shone using the face
Valut is trained the convolutional neural networks, obtains a trained convolutional neural networks;
S103: when attendance, using photographing device to needing all personnel of attendance to be taken pictures together, Face datection is utilized
Method the face in the photo of shooting is detected and is proposed, form a face photograph collection;
S104: it using the human face photo collection as the input of the trained convolutional neural networks, obtains needing attendance
The corresponding ID number of personnel;
S105: according to obtained ID number, counting the attendance data for the personnel for needing attendance, obtains record and turns out for work
The Excel table of data;Attendance EP (end of program).
Further, in step S102, the convolutional neural networks share 7 layers, successively are as follows: input layer, convolutional layer, Chi Hua
Layer, convolutional layer, pond layer, full articulamentum and output layer.
Further, in step S102, mug shot used in the convolutional neural networks is that pixel is 100 × 100
Gray level image, the size of C1 layers of convolution kernel is 8 × 8, and convolution kernel type is 6;F2 layers of sampling area are 3 × 3, and sampling type is
6 kinds;The size of C3 layers of convolution kernel is 8 × 8, and the type of convolution kernel is 12 kinds;F4 layers of sampling area are 3 × 3, sampling type 12
Kind;And 6 Gabor filters are provided at C1 layers, the convolution template size of this 6 Gabor filters is 8 × 8, direction point
Not are as follows: 0, π/6, π/3, pi/2,2 π/3,5 π/6;C3 layers are provided with 12 Gabor filters, the volume of this 12 Gabor filters
Product module board size is 8 × 8, and direction is respectively as follows: 0, π/12, π/6, π/4, π/3,5 π/12, pi/2,7 π/12,2 π/3,3 π/4,5
π/6、11π/12。
Further, in step S102, the activation primitive of the convolutional neural networks is ReLu function;The ReLu function
Expression formula are as follows: f (x)=max (0, x).
Further, in step S102, in the training convolutional neural networks, frequency of training x, learning rate y;Its
In, x and y are empirical value;The value that the value of x is 250, y is 0.06.
Further, it in step S102, in the training convolutional neural networks, is combined using scrambling and bootstrap
Sample expands method and solves to use the human face photo library training convolutional neural networks bring small sample problem of small sample;It utilizes
Bootstrap expands the human face photo in the human face photo library, the specific steps are as follows:
S201: being set and extract quantity as 1000, and generated 1000 numbers from 1 to 135 at random using MATLAB, this mistake
Journey, which is equivalent to, randomly selects human face photo from the human face photo library using resampling technique;
S202: the random carry out Gauss of choose 1000 human face photos is added make an uproar, at random plus make an uproar, rotation process processing;
S203: and treated that human face photo is added in the human face photo library by this 1000, and to the face
Photo library is reconstructed.
Further, in step S105, the attendance data for the personnel for needing attendance is counted, record is obtained and turns out for work
The specific method of the Excel table of data includes:
S301: after human face photo inputs convolutional neural networks, convolutional neural networks can identify the corresponding people of the human face photo
And its corresponding ID number is exported, and the ID number of output is generated to the matrix a of a 1 × N;Wherein, N is input convolutional neural networks
Picture sum, the element in a is corresponding in turn to the ID number of input picture;
S302: the full 0 matrix b that a size is 1 × M is defined;And Ergodic Matrices a, successively from 1 to M, these are counted in statistics a
The number that word occurs, and successively update the element in matrix b;M is the total number of persons for needing attendance;After traversal, obtain more
Matrix B after new, each element in matrix B are followed successively by the attendance number of student of the class number from 1 to M;
S303: according to matrix B, the Excel table for recording the attendance data of attendance personnel in need is automatically generated.
Further, a kind of face of the convolutional neural networks based on small sample is taken pictures attendance checking system, it is characterised in that: packet
It includes with lower module:
Module is established in human face photo library, for establishing the human face photo library to attendance personnel;It is described have to attendance personnel it is more
It is a;Include all facial expression photos to attendance personnel in the human face photo library, is each corresponding with multiple to attendance personnel
The mug shot of different expressions, and each a unique ID number is corresponding with to attendance personnel;
Network training module, for establishing based on the improved convolutional neural networks for adapting to face attendance of LeNet-5, and benefit
The convolutional neural networks are trained with the human face photo library, obtain a trained convolutional neural networks;
Face datection and extraction module, when being used for attendance, using photographing device to need all personnel of attendance together into
Row is taken pictures, and the face in the photo of shooting is detected and proposed using the method for Face datection, forms a human face photo
Collection;
Convolutional neural networks identification module, for using the human face photo collection as the trained convolutional neural networks
Input, obtain needing the corresponding ID number of the personnel of attendance;
Student attendance sum statistical module, for according to obtained ID number, to the attendance data of the personnel for needing attendance into
Row statistics obtains the Excel table that record has attendance data;Attendance EP (end of program).
Further, it in network training module, in the training convolutional neural networks, is mutually tied using scrambling with bootstrap
The sample of conjunction expands method and solves the human face photo library training convolutional neural networks bring sample that the present invention uses small sample
This problem;The human face photo in the human face photo library is expanded using bootstrap, is specifically included such as lower unit:
Sample expansion unit, for set extraction quantity as 1000, and generated at random using MATLAB 1000 from 1 to
135 number, this process, which is equivalent to, randomly selects human face photo from the human face photo library using resampling technique;
Picture processing unit makes an uproar for adding the random carry out Gauss of choose 1000 human face photos, at random plus makes an uproar, revolves
Turn operation processing;
Reconfiguration unit, is used for and treated that human face photo is added in the human face photo library by this 1000, and right
The human face photo library is reconstructed.
Further, in student attendance sum statistical module, the attendance data for the personnel for needing attendance is counted, is obtained
Include: to the specific method for having the Excel table of attendance data is recorded
ID number storage unit, after human face photo input convolutional neural networks, convolutional neural networks can identify the face
The corresponding people of photo simultaneously exports its corresponding ID number, and the ID number of output is generated to the matrix a of a 1 × N;Wherein, N is input
The picture of convolutional neural networks is total, and the element in a is corresponding in turn to the ID number of input picture;
Attendance record unit, for defining the full 0 matrix b that a size is 1 × M;And Ergodic Matrices a, successively count in a
The number that these numbers occur from 1 to M, and successively update the element in matrix b;M is the total number of persons for needing attendance;It has traversed
Bi Hou, obtains updated matrix B, and each element in matrix B is followed successively by the attendance number of student of the class number from 1 to M;
Total attendance output unit, for automatically generating the attendance data for recording attendance personnel in need according to matrix B
Excel table.
It is a kind of new that technical solution provided by the invention has the benefit that technical solution proposed by the invention proposes
Face take pictures Work attendance method, devise a set of face based on convolutional neural networks and take pictures attendance checking system, for intelligentized
The attendance of student is recorded, it is practical.
Detailed description of the invention
Present invention will be further explained below with reference to the attached drawings and examples, in attached drawing:
Fig. 1 is that a kind of face of the convolutional neural networks based on small sample is taken pictures the stream of Work attendance method in the embodiment of the present invention
Cheng Tu;
Fig. 2 is the structural schematic diagram of convolutional neural networks in the embodiment of the present invention;
Fig. 3 is the schematic diagram that the convergence rate before Gabor filter is introduced in the embodiment of the present invention;
Fig. 4 is the schematic diagram that the convergence rate after Gabor filter is introduced in the embodiment of the present invention;
Fig. 5 is in the embodiment of the present invention through adding the picture example made an uproar, generated after rotation process;
Fig. 6 is that a kind of face of the convolutional neural networks based on small sample is taken pictures the module composition of system in the embodiment of the present invention
Schematic diagram;
Fig. 7 is input test collection picture schematic diagram in the embodiment of the present invention;
Fig. 8 is class's checking-in result schematic diagram in the embodiment of the present invention.
Specific embodiment
For a clearer understanding of the technical characteristics, objects and effects of the present invention, now control attached drawing is described in detail
A specific embodiment of the invention.
The embodiment provides a kind of face of convolutional neural networks based on small sample take pictures Work attendance method and
System.
Referring to FIG. 1, Fig. 1 is that a kind of face of the convolutional neural networks based on small sample is taken pictures and examined in the embodiment of the present invention
The flow chart of diligent method, specifically comprises the following steps:
S101: the human face photo library to attendance personnel is established;It is described have to attendance personnel it is multiple;In the human face photo library
Comprising all facial expression photos to attendance personnel, the mug shot of multiple different expressions is each corresponding with to attendance personnel,
And a unique ID number is each corresponding with to attendance personnel;
S102: it establishes based on improved convolutional neural networks (the network structure such as Fig. 2 for adapting to face attendance of LeNet-5
It is shown), and the convolutional neural networks are trained using the human face photo library, obtain trained convolutional Neural net
Network;
S103: when attendance, using photographing device to needing all personnel of attendance to take pictures together, Face datection is utilized
Method the face in the photo of shooting is detected and is proposed, form a face photograph collection;
S104: it using the human face photo collection as the input of the trained convolutional neural networks, obtains needing attendance
The corresponding ID number of personnel;
S105: according to obtained ID number, counting the attendance data for the personnel for needing attendance, obtains record and turns out for work
The Excel table of data;Attendance EP (end of program).
In step S102, the convolutional neural networks share 7 layers, successively are as follows: input layer, convolutional layer, pond layer, convolution
Layer, pond layer, full articulamentum and output layer;
In step S102, mug shot used in the convolutional neural networks is the gray level image that pixel is 100*100,
The size of C1 layers of convolution kernel is 8*8, and convolution kernel type is 6;F2 layers of sampling area are 3*3, and sampling type is 6 kinds;C3 layers of convolution
The size of core is 8*8, and the type of convolution kernel is 12 kinds;F4 layers of sampling area are 3*3, and sampling type is 12 kinds;And it is set at C1 layers
Set 6 Gabor filters, the convolution template size of this 6 Gabor filters is 8*8, direction be respectively as follows: 0, π/6, π/
3,π/2,2π/3,5π/6;C3 layers are provided with 12 Gabor filters, and the convolution mask size of this 12 Gabor filters is
8*8, direction are respectively as follows: 0, π/12, π/6, π/4, π/3,5 π/12, pi/2,7 π/12,2 π/3,3 π/4,5 π/6,11 π/12.Fig. 3 and
Fig. 4 is respectively the convolutional Neural after what introducing Gabor of the convergence rate of the convolutional neural networks before introducing Gabor filter
The convergence rate of network, it can be seen that introduce after Gabor filter, convergence rate is faster.
In step S102, the activation primitive of the convolutional neural networks is ReLu function;The expression formula of the ReLu function
Are as follows: f (x)=max (0, x), when input signal<0, output is all 0, and in the case where input signal>0, output is equal to input.
In step S102, in the training convolutional neural networks, frequency of training x, learning rate y;Wherein, x and y are equal
For empirical value;The value that the value of x is 250, y is 0.06.
In step S102, in the training convolutional neural networks, expanded using the sample that scrambling and bootstrap combine
Method solves the human face photo library training convolutional neural networks bring small sample problem that the present invention uses small sample;Using certainly
It helps method to expand the human face photo in the human face photo library: noise being added at random to the human face photo and several what connection
It disturbs, and -10 ° and+10 ° of the random rotation of part human face photo (is illustrated in figure 5 through adding the figure made an uproar and generated after rotation process
Piece exemplary diagram), training Shi Douhui constructs different test sets every time in this way, is conducive to the feature for preferably extracting image;Tool
Body is as follows:
S201: being set and extract quantity as 1000, and generated 1000 numbers from 1 to 135 at random using matlab, this mistake
Journey, which is equivalent to, randomly selects human face photo from the human face photo library using resampling technique;
S202: the random carry out Gauss of choose 1000 human face photos is added make an uproar, at random plus make an uproar, rotation process processing;
S203: and treated that human face photo is added in the human face photo library by this 1000, and to the face
Photo library is reconstructed.
S103: when attendance, using photographing device to needing all personnel of attendance to take pictures, the side of Face datection is utilized
Method is detected and is proposed to the face in the photo of shooting, and a face photograph collection is formed;
S104: it using the human face photo collection as the input of the trained convolutional neural networks, obtains needing attendance
The corresponding ID number of personnel;
S105: according to obtained ID number, counting the attendance data for the personnel for needing attendance, obtains record and turns out for work
The Excel table of data;Attendance EP (end of program).
In step S105, the attendance data for the personnel for needing attendance is counted, obtaining record has attendance data
The specific method of Excel table includes:
S301: after human face photo inputs convolutional neural networks, convolutional neural networks can identify the corresponding people of the human face photo
And its corresponding ID number is exported, and the ID number of output is generated to the matrix a of a 1*N;Wherein, N is input convolutional neural networks
Picture sum, the element in a is corresponding in turn to the ID number of input picture;
S302: the full 0 matrix b that a size is 1*M is defined;And Ergodic Matrices a, successively from 1 to M, these are counted in statistics a
The number that word occurs, and successively update the element in matrix b;M is the total number of persons for needing attendance;After traversal, obtain more
Matrix B after new, each element in matrix B are followed successively by the attendance number of student of the class number from 1 to M;
S303: according to matrix B, the Excel table for recording the attendance data of attendance personnel in need is automatically generated.
In step S105, the attendance data for the personnel for needing attendance is counted, obtaining record has attendance data
The specific method of Excel table includes:
Referring to Fig. 6, for a kind of convolutional neural networks based on small sample face take pictures attendance checking system module composition
Schematic diagram, including sequentially connected: module 11, network training module 12, Face datection and extraction module are established in human face photo library
13, convolutional neural networks identification module 14 and student attendance sum statistical module 15;
Module 11 is established in human face photo library, for establishing the human face photo library to attendance personnel;The people for needing attendance
Member has multiple;In the human face photo library comprising attendance personnel in need facial expression photo, each to attendance personnel couple
There should be the mug shot of multiple different expressions, and each be corresponding with a unique ID number to attendance personnel;
Network training module 12, for establishing based on the improved convolutional neural networks for adapting to face attendance of LeNet-5, and
The convolutional neural networks are trained using the human face photo library, obtain trained convolutional neural networks;
Face datection and extraction module 13, when being used for attendance, using photographing device to needing all personnel of attendance to carry out
It takes pictures, the face in the photo of shooting is detected and proposed using the method for Face datection, form a face photograph collection;
Convolutional neural networks identification module 14, for using the human face photo collection as the trained convolutional Neural net
The input of network obtains needing the corresponding ID number of the personnel of attendance;
Student attendance sum statistical module 15, for the attendance data according to obtained ID number, to the personnel for needing attendance
It is counted, obtains the Excel table that record has attendance data;Attendance EP (end of program).
In the present embodiment, in network training module, in the training convolutional neural networks, using scrambling and bootstrap
The sample combined expands method and solves the human face photo library training convolutional neural networks bring that the present invention uses small sample
Small sample problem;The human face photo in the human face photo library is expanded using bootstrap: random to the human face photo
Noise and geometry interference is added, and by -10 ° and+10 ° of the random rotation of part human face photo, trains Shi Douhui structure every time in this way
Different test sets is built out, the feature for preferably extracting image is conducive to;It specifically includes such as lower unit:
Sample expansion unit, for set extraction quantity as 1000, and generated at random using MATLAB 1000 from 1 to
135 number, this process, which is equivalent to, randomly selects human face photo from the human face photo library using resampling technique;
Picture processing unit makes an uproar for adding the random carry out Gauss of choose 1000 human face photos, at random plus makes an uproar, revolves
Turn operation processing;
Reconfiguration unit, is used for and treated that human face photo is added in the human face photo library by this 1000, and right
The human face photo library is reconstructed.
In the present embodiment, it in student attendance sum statistical module, unites to the attendance data for the personnel for needing attendance
Meter, the specific method that obtaining record has the Excel table of attendance data include:
ID number storage unit, after human face photo input convolutional neural networks, convolutional neural networks can identify the face
The corresponding people of photo simultaneously exports its corresponding ID number, and the ID number of output is generated to the matrix a of a 1*N;Wherein, N is input
The picture of convolutional neural networks is total, and the element in a is corresponding in turn to the ID number of input picture;
Attendance record unit, for defining the full 0 matrix b that a size is 1*M;And Ergodic Matrices a, successively count in a
The number that these numbers occur from 1 to M, and successively update the element in matrix b;M is the total number of persons for needing attendance;It has traversed
Bi Hou, obtains updated matrix B, and each element in matrix B is followed successively by the attendance number of student of the class number from 1 to M;
Total attendance output unit, for automatically generating the attendance data for recording attendance personnel in need according to matrix B
Excel table.
The groundwork to be completed of the present invention is total attendance of intelligent record student.Teacher is by the class of multiple bats of attending class
After group picture inputs Face datection and extraction module, by recognition of face therein and extracts, classify through convolutional neural networks
Afterwards again by attendance number statistical module Lai intelligent record, class's Excel Attendance Sheet is ultimately generated.In order to verify convolutional neural networks
Identification module and attendance number statistical module performance, multiple performance test collection tests the performance of network, and Fig. 7 is input
The test set comprising 30 pictures, Fig. 8 is primary output result.
Test set is 30 photos of 15 people under 2 kinds of expressions, class number corresponding to the picture of input be followed successively by (6,13,
3,9,3,10,10,7,8,2,12,11,2,1,8,6,4,5,9,14,15,11,4,14,7,1,13,12,5,15), convolutional Neural
Network is successively identified to the picture of input and is exported its corresponding class number, the result of identification be followed successively by (6,13,3,15,3,
10,10,7,8,2,12,11,2,1,8,6,4,5,9,14,15,11,4,14,7,1,13,12,5,15)。
Class number corresponding to face picture by input and the class number exported after convolutional neural networks identification carry out
Comparison, it can be seen that the 4th photo of the identification input of the convolutional neural networks mistake in current test, is 9 by class number
Classmate be identified as class number be 15 classmate, error rate 3.33%.The system can will be missed it can be seen from the experimental result
Difference control is within 5%, total attendance of intelligentized record student.
The beneficial effects of the present invention are: technical solution proposed by the invention proposes that a kind of new face is taken pictures attendance side
Method devises a set of face based on convolutional neural networks and takes pictures attendance checking system, real for the attendance of intelligentized record student
It is strong with property.
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all in spirit of the invention and
Within principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (10)
- The Work attendance method 1. a kind of face of the convolutional neural networks based on small sample is taken pictures, it is characterised in that: the following steps are included:S101: the human face photo library to attendance personnel is established;It is described have to attendance personnel it is multiple;Include in the human face photo library All facial expression photos to attendance personnel are each corresponding with the human face photo of multiple different expressions to attendance personnel, and every It is a to be corresponding with a unique ID number to attendance personnel;S102: it establishes based on the improved convolutional neural networks for adapting to face attendance of LeNet-5, and utilizes the human face photo library The convolutional neural networks are trained, a trained convolutional neural networks are obtained;S103: when attendance, using photographing device to needing all personnel of attendance to be taken pictures together, the side of Face datection is utilized Method is detected to the face in the photo of shooting and is proposed each face, forms a face photograph collection;S104: using the human face photo collection as the input of the trained convolutional neural networks, the people for needing attendance is obtained The corresponding ID number of member;S105: according to obtained ID number, counting the attendance data for the personnel for needing attendance, and obtaining record has attendance data Excel table, attendance EP (end of program).
- The Work attendance method 2. a kind of face of the convolutional neural networks based on small sample as described in claim 1 is taken pictures, feature Be: in step S102, the convolutional neural networks share 7 layers, successively are as follows: input layer, convolutional layer, pond layer, convolutional layer, pond Change layer, full articulamentum and output layer.
- The Work attendance method 3. a kind of face of the convolutional neural networks based on small sample as described in claim 1 is taken pictures, feature Be: in step S102, mug shot used in the convolutional neural networks is the gray level image that pixel is 100 × 100, C1 The size of layer convolution kernel is 8 × 8, and convolution kernel type is 6;F2 layers of sampling area are 3 × 3, and sampling type is 6 kinds;C3 layers of convolution The size of core is 8 × 8, and the type of convolution kernel is 12 kinds;F4 layers of sampling area are 3 × 3, and sampling type is 12 kinds;And at C1 layers Provided with 6 Gabor filters, the convolution template size of this 6 Gabor filters is 8 × 8, direction be respectively as follows: 0, π/6, π/3,π/2,2π/3,5π/6;C3 layers are provided with 12 Gabor filters, and the convolution mask size of this 12 Gabor filters is equal It is 8 × 8, direction is respectively as follows: 0, π/12, π/6, π/4, π/3,5 π/12, pi/2,7 π/12,2 π/3,3 π/4,5 π/6,11 π/12.
- The Work attendance method 4. a kind of face of the convolutional neural networks based on small sample as described in claim 1 is taken pictures, feature Be: in step S102, the activation primitive of the convolutional neural networks is ReLu function;The expression formula of the ReLu function are as follows: f (x)=max (0, x).
- The Work attendance method 5. a kind of face of the convolutional neural networks based on small sample as described in claim 1 is taken pictures, feature It is: in step S102, in the training convolutional neural networks, frequency of training x, learning rate y;Wherein, x and y are Empirical value;The value that the value of x is 250, y is 0.06.
- The Work attendance method 6. a kind of face of the convolutional neural networks based on small sample as described in claim 1 is taken pictures, feature It is: in step S102, in the training convolutional neural networks, method solution is expanded using the sample that scrambling and bootstrap combine Certainly use the human face photo library training convolutional neural networks bring small sample problem of small sample;It will be described using bootstrap Human face photo in human face photo library is expanded, the specific steps are as follows:S201: being set and extract quantity as 1000, and generated 1000 numbers from 1 to 135 at random using MATLAB, this process phase When in randomly selecting human face photo from the human face photo library using resampling technique;S202: the random carry out Gauss of choose 1000 human face photos is added make an uproar, at random plus make an uproar, rotation process processing;S203: and treated that human face photo is added in the human face photo library by this 1000, and to the human face photo Library is reconstructed.
- The Work attendance method 7. a kind of face of the convolutional neural networks based on small sample as described in claim 1 is taken pictures, feature It is: in step S105, the attendance data for the personnel for needing attendance is counted, obtains the Excel that record has attendance data The specific method of table includes:S301: after human face photo inputs convolutional neural networks, convolutional neural networks can identify the corresponding people of the human face photo and defeated Its corresponding ID number out, and by the ID number of output generate a 1 × N matrix a;Wherein, N is the figure for inputting convolutional neural networks Piece is total, and the element in a is corresponding in turn to the ID number of input picture;S302: the full 0 matrix b that a size is 1 × M is defined;And Ergodic Matrices a, successively from 1 to M, these numbers go out in statistics a Existing number, and successively update the element in matrix b;M is the total number of persons for needing attendance;After traversal, after obtaining update Matrix B, each element in matrix B is followed successively by the attendance number of student of the class number from 1 to M;S303: according to matrix B, the Excel table for recording the attendance data of attendance personnel in need is automatically generated.
- The attendance checking system 8. a kind of face of the convolutional neural networks based on small sample is taken pictures, it is characterised in that: comprise the following modules:Module is established in human face photo library, for establishing the human face photo library to attendance personnel;It is described have to attendance personnel it is multiple;Institute It states comprising all facial expression photos to attendance personnel in human face photo library, is each corresponding with multiple different tables to attendance personnel The mug shot of feelings, and each a unique ID number is corresponding with to attendance personnel;Network training module for establishing based on the improved convolutional neural networks for adapting to face attendance of LeNet-5, and utilizes institute It states human face photo library to be trained the convolutional neural networks, obtains a trained convolutional neural networks;Face datection and extraction module, when being used for attendance, using photographing device to needing all personnel of attendance to clap together According to the face in the photo of shooting is detected and is proposed using the method for Face datection, forms a face photograph collection;Convolutional neural networks identification module, for using the human face photo collection as the defeated of the trained convolutional neural networks Enter, obtains needing the corresponding ID number of the personnel of attendance;Student attendance sum statistical module, for being united to the attendance data for the personnel for needing attendance according to obtained ID number Meter obtains the Excel table that record has attendance data;Attendance EP (end of program).
- The attendance checking system 9. a kind of face of the convolutional neural networks based on small sample as claimed in claim 8 is taken pictures, feature It is: in network training module, in the training convolutional neural networks, is expanded using the sample that scrambling and bootstrap combine Method solves the human face photo library training convolutional neural networks bring small sample problem that the present invention uses small sample;Using certainly It helps method to expand the human face photo in the human face photo library, specifically includes such as lower unit:Sample expansion unit for set extraction quantity as 1000, and generates using MATLAB 1000 from 1 to 135 at random Number, this process, which is equivalent to, randomly selects human face photo from the human face photo library using resampling technique;Picture processing unit makes an uproar for adding the random carry out Gauss of choose 1000 human face photos, at random plus makes an uproar, rotates behaviour It deals with;Reconfiguration unit, is used for and treated that human face photo is added in the human face photo library by this 1000, and to described Human face photo library is reconstructed.
- The attendance checking system 10. a kind of face of the convolutional neural networks based on small sample as claimed in claim 8 is taken pictures, feature It is: in student attendance sum statistical module, the attendance data for the personnel for needing attendance is counted, obtains record and turn out for work The specific method of the Excel table of data includes:ID number storage unit, after human face photo input convolutional neural networks, convolutional neural networks can identify the human face photo Corresponding people simultaneously exports its corresponding ID number, and the ID number of output is generated to the matrix a of a 1 × N;Wherein, N is input convolution The picture of neural network is total, and the element in a is corresponding in turn to the ID number of input picture;Attendance record unit, for defining the full 0 matrix b that a size is 1 × M;And Ergodic Matrices a, it successively counts in a from 1 To the number of these number appearance of M, and successively update the element in matrix b;M is the total number of persons for needing attendance;Traversal finishes Afterwards, updated matrix B is obtained, each element in matrix B is followed successively by the attendance number of student of the class number from 1 to M;Total attendance output unit, for automatically generating the Excel for recording the attendance data of attendance personnel in need according to matrix B Table.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811563353.7A CN109635766A (en) | 2018-12-20 | 2018-12-20 | The face of convolutional neural networks based on small sample is taken pictures Work attendance method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811563353.7A CN109635766A (en) | 2018-12-20 | 2018-12-20 | The face of convolutional neural networks based on small sample is taken pictures Work attendance method and system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109635766A true CN109635766A (en) | 2019-04-16 |
Family
ID=66075846
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811563353.7A Withdrawn CN109635766A (en) | 2018-12-20 | 2018-12-20 | The face of convolutional neural networks based on small sample is taken pictures Work attendance method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109635766A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111401331A (en) * | 2020-04-27 | 2020-07-10 | 支付宝(杭州)信息技术有限公司 | Face recognition method and device |
CN111611911A (en) * | 2020-05-19 | 2020-09-01 | 成都理工大学 | Class attendance checking method based on convolutional neural network and multi-face recognition |
CN113159006A (en) * | 2021-06-23 | 2021-07-23 | 杭州魔点科技有限公司 | Attendance checking method and system based on face recognition, electronic equipment and storage medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102073586A (en) * | 2010-12-23 | 2011-05-25 | 北京航空航天大学 | Gray generalized regression neural network-based small sample software reliability prediction method |
CN105426875A (en) * | 2015-12-18 | 2016-03-23 | 武汉科技大学 | Face identification method and attendance system based on deep convolution neural network |
CN106940905A (en) * | 2017-03-15 | 2017-07-11 | 上海巽晔计算机科技有限公司 | Classroom automatic roll-calling system based on WIFI and smart mobile phone |
CN108154276A (en) * | 2017-12-29 | 2018-06-12 | 天津市淘客科技有限公司 | Attendance analysis method based on image identification and machine learning |
CN108921038A (en) * | 2018-06-07 | 2018-11-30 | 河海大学 | A kind of classroom based on deep learning face recognition technology is quickly called the roll method of registering |
-
2018
- 2018-12-20 CN CN201811563353.7A patent/CN109635766A/en not_active Withdrawn
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102073586A (en) * | 2010-12-23 | 2011-05-25 | 北京航空航天大学 | Gray generalized regression neural network-based small sample software reliability prediction method |
CN105426875A (en) * | 2015-12-18 | 2016-03-23 | 武汉科技大学 | Face identification method and attendance system based on deep convolution neural network |
CN106940905A (en) * | 2017-03-15 | 2017-07-11 | 上海巽晔计算机科技有限公司 | Classroom automatic roll-calling system based on WIFI and smart mobile phone |
CN108154276A (en) * | 2017-12-29 | 2018-06-12 | 天津市淘客科技有限公司 | Attendance analysis method based on image identification and machine learning |
CN108921038A (en) * | 2018-06-07 | 2018-11-30 | 河海大学 | A kind of classroom based on deep learning face recognition technology is quickly called the roll method of registering |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111401331A (en) * | 2020-04-27 | 2020-07-10 | 支付宝(杭州)信息技术有限公司 | Face recognition method and device |
CN111611911A (en) * | 2020-05-19 | 2020-09-01 | 成都理工大学 | Class attendance checking method based on convolutional neural network and multi-face recognition |
CN113159006A (en) * | 2021-06-23 | 2021-07-23 | 杭州魔点科技有限公司 | Attendance checking method and system based on face recognition, electronic equipment and storage medium |
CN113159006B (en) * | 2021-06-23 | 2021-09-14 | 杭州魔点科技有限公司 | Attendance checking method and system based on face recognition, electronic equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109635766A (en) | The face of convolutional neural networks based on small sample is taken pictures Work attendance method and system | |
Kim et al. | First and second levels of intermedia agenda setting: Political advertising, newspapers, and Twitter during the 2012 US presidential election | |
CN108419091A (en) | A kind of verifying video content method and device based on machine learning | |
CN104298973B (en) | Facial image spinning solution based on self-encoding encoder | |
CN108921037B (en) | Emotion recognition method based on BN-acceptance double-flow network | |
CN103473555B (en) | Horrible video scene recognition method based on various visual angles multi-instance learning | |
Madden et al. | The double-edged crisis: Invisible Children’s social media response to the Kony 2012 campaign | |
CN104268140B (en) | Image search method based on weight self study hypergraph and multivariate information fusion | |
Luminet et al. | Flashbulb memories: New challenges and future perspectives | |
Zhao et al. | M3ED: Multi-modal multi-scene multi-label emotional dialogue database | |
Tella | Humour generation and multimodal framing of political actor in the 2015 Nigerian presidential election campaign memes | |
Wang et al. | Neurocognition-inspired design with machine learning | |
CN103024569A (en) | Method and system for performing parent-child education data interaction through smart television | |
Sorlin | Historical films as tools for historians | |
Björklund et al. | A visual history of HIV/AIDS: exploring the face of AIDS film archive | |
CN110322086A (en) | A kind of education writing on the blackboard checking system based on Intelligent hardware and cloud platform | |
Oikonomou et al. | Virtual Reality in Humanistic Prisons Education. The STEPS project. | |
CN109934803A (en) | A method of scoring for mobile terminal shooting low resolution picture based on deep learning | |
Hao et al. | Evaluation System of Foreign Language Teaching Quality Based on Spatiotemporal Feature Fusion | |
Vijay et al. | Group face recognition smart attendance system using convolution neural network | |
Finnegan | The daguerreotype, republican style and theories of the public image. | |
Geetha et al. | Implementation of a Blockchain Based Attendance Tracking System | |
Surasak et al. | Application of Deep Learning on Student Attendance Checking in Virtual Classroom | |
DOĞUKANLI et al. | Giorgio Agamben’in Dispozitif Kavramı Üzerinden Midjourney Eleştirisi | |
Lieberman | Will the Courts Meet the Challenge of Technology |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WW01 | Invention patent application withdrawn after publication |
Application publication date: 20190416 |
|
WW01 | Invention patent application withdrawn after publication |