CN104318224B - A kind of face identification method and monitoring device - Google Patents

A kind of face identification method and monitoring device Download PDF

Info

Publication number
CN104318224B
CN104318224B CN201410658870.8A CN201410658870A CN104318224B CN 104318224 B CN104318224 B CN 104318224B CN 201410658870 A CN201410658870 A CN 201410658870A CN 104318224 B CN104318224 B CN 104318224B
Authority
CN
China
Prior art keywords
sample
nearest neighbor
default
monitoring device
class
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.)
Active
Application number
CN201410658870.8A
Other languages
Chinese (zh)
Other versions
CN104318224A (en
Inventor
仲崇亮
徐勇
林晓清
李静
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Entropy Technology Co Ltd
Original Assignee
SHENZHEN ZHONGKONG BIOMETRICS TECHNOLOGY Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by SHENZHEN ZHONGKONG BIOMETRICS TECHNOLOGY Co Ltd filed Critical SHENZHEN ZHONGKONG BIOMETRICS TECHNOLOGY Co Ltd
Priority to CN201410658870.8A priority Critical patent/CN104318224B/en
Publication of CN104318224A publication Critical patent/CN104318224A/en
Application granted granted Critical
Publication of CN104318224B publication Critical patent/CN104318224B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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
    • G06V40/168—Feature extraction; Face representation
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • General Health & Medical Sciences (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

The invention discloses a kind of face identification method and monitoring device, for improving the precision and efficiency of recognition of face.Methods described includes：Monitoring device obtains face image data；The monitoring device carries out local linear discriminant analysis LLDA to the face image data, extracts face characteristic data；The monitoring device carries out Classification and Identification according to default k nearest neighbor sample to the face characteristic data.

Description

A kind of face identification method and monitoring device
Technical field
The present invention relates to biological intelligence technical field, more particularly to a kind of face identification method and monitoring device.
Background technology
In the last few years, security protection turned into more and more important focus.Increasing residential area, kindergarten, school, public affairs Department and important exhibition can all be equipped with security personnel.But because personnel amount is huge, security personnel is difficult to remember each place The personnel in region, it is also difficult to accurate discrimination go out be not region personnel.Also, Migrant women is mixed into residential area, child The crime case in garden, school, company etc. is of common occurrence.So it is highly desirable using face recognition technology come to ensureing institute In the personal safety as well as the property safety of area people.
Linear discriminant analysis (Linear Discriminant Analysis, LDA) is kind of effective feature extraction and a drop The method of dimension.LDA has been successfully applied to various modes identification problem, such as recognition of face, text identification, also largely with The related machine learning application of image.In the late two decades, it is proposed that much to traditional LDA improved method come improve its precision and Efficiency, all it is just with the independent linear transformation in world coordinates system mostly.These methods are identical in having in data category What can be showed during the Gaussian Profile of covariance structure is fine.
But if data category distribution is more more complicated than Gauss, need calculating incidence matrix (each in incidence matrix Element is the distance between two data vectors), also feature decomposition is carried out to the matrix that whole training set generates.If input The dimension of data is very big and the size of whole training set is very huge, and the dimension of scatter diagram or correlation matrix will be very Greatly, and the feature decompositions of these matrixes will take and not practicable very much, and this method need to train all samples, popular Say to be exactly processing to great amount of samples.Traditional LDA is not appropriate for the processing of large sample.Therefore, face is largely influenceed The precision and efficiency of identification.
The content of the invention
The invention provides a kind of face identification method and monitoring device, for improving the precision and efficiency of recognition of face.
First aspect present invention provides a kind of face identification method, including：
Monitoring device obtains face image data；
The monitoring device carries out local linear discriminant analysis LLDA to the face image data, extracts face characteristic number According to；
The monitoring device carries out Classification and Identification according to default k nearest neighbor sample to the face characteristic data.
With reference to the first aspect of the present invention, in the first implementation of the first aspect of the present invention, the local line Property discriminant analysis LLDA include the local linear discriminant analysis VLLDA based on vector or the local linear discriminant analysis based on matrix MLLDA。
With reference to the first implementation of the first aspect of the present invention, in second of realization side of the first aspect of the present invention In formula, the face image data is expressed as vector by the local linear discriminant analysis VLLDA based on vector, described to be based on The face image data is expressed as matrix by the local linear discriminant analysis MLLDA of matrix.
Second with reference to the first of the first aspect of the present invention or first aspect implementation or first aspect is real Existing mode, in the third implementation of the first aspect of the present invention, the monitoring device is entered to the face image data Row local linear discriminant analysis LLDA, extraction face characteristic data specifically include：
The monitoring device is according to the default k nearest neighbor sample, default first formula and default second formula difference Scatter matrix S in the class of the face image data is calculatedwThe scatter matrix S between classb, the k nearest neighbor sample includes multiple The set of the training sample of class；
Default first formula is：
Default second formula is：
Wherein, c ' be k nearest neighbor sample classification number, m ' be k nearest neighbor sample average, l 'iFor the i-th class K in classification c ' The number of neighbour's sample, i.e.,m′iFor the average of the k nearest neighbor sample of the i-th class in classification c ', x 'ijFor classification c ' In the i-th class j-th of k nearest neighbor sample；
The monitoring device is according to scatter matrix S in the classw, scatter matrix S between the classbAnd Fei Sheer criterion meters Calculation obtains the characteristic vector of face image data；
The monitoring device carries out feature extraction to the characteristic vector of the face image data, obtains the face characteristic Data.
With reference to the third implementation of the first aspect of the present invention, in the 4th kind of realization side of the first aspect of the present invention In formula, the monitoring device calculates respectively according to the default k nearest neighbor sample, default first formula and default second formula Obtain scatter matrix S in the class of the face image datawThe scatter matrix S between classbAlso include before：
The monitoring device carries out dimension-reduction treatment according to principal component analysis PCA to training sample；
The monitoring device determines the default k nearest neighbor by k nearest neighbor algorithm from the training sample after dimension-reduction treatment Sample.
Second with reference to the first of the first aspect of the present invention or first aspect implementation or first aspect is real Existing mode, in the 5th kind of implementation of the first aspect of the present invention, the monitoring device is according to default k nearest neighbor sample pair The face characteristic data carry out Classification and Identification and specifically included：
The monitoring device is matched the face characteristic data with the default k nearest neighbor sample；
The monitoring device is obtained and the face characteristic data pair using the training sample matched by k nearest neighbor algorithm The test sample answered；
The monitoring device calculates the Euclidean distance between the test sample and the default k nearest neighbor sample；
When the Euclidean distance is less than preset threshold value, determine to include the face characteristic number in facial feature database According to；When the Euclidean distance is more than preset threshold value, determine not including the face characteristic data in facial feature database.
With reference to the 5th kind of implementation of the first aspect of the present invention, in the 6th kind of realization side of the first aspect of the present invention Do not include the face characteristic data in formula, in the determination facial feature database includes afterwards：
The monitoring device display warning picture, and the prompt tone that gives a warning.
Second aspect of the present invention provides a kind of monitoring device, including：
Acquiring unit, for obtaining face image data；
Analytic unit, for carrying out local linear discriminant analysis LLDA to the face image data, extract face characteristic Data；
Recognition unit, for carrying out Classification and Identification to the face characteristic data according to default k nearest neighbor sample.
With reference to the second aspect of the present invention, in the first implementation of the second aspect of the present invention, the analysis is single Member specifically includes：
First computing module, for according to the default k nearest neighbor sample, default first formula and default second formula Scatter matrix S in the class of the face image data is calculated respectivelywThe scatter matrix S between classb, the k nearest neighbor sample includes The set of the training sample of multiple classes；
Default first formula is：
Default second formula is：
Wherein, c ' be k nearest neighbor sample classification number, m ' be k nearest neighbor sample average, l 'iFor the i-th class K in classification c ' The number of neighbour's sample, i.e.,m′iFor the average of the k nearest neighbor sample of the i-th class in classification c ', x 'ijFor classification c ' In the i-th class j-th of k nearest neighbor sample；
Second computing module, for according to scatter matrix S in the classw, scatter matrix S between the classbAnd Fei Sheer is accurate The characteristic vector of face image data is then calculated；
Extraction module, for carrying out feature extraction to the characteristic vector of the face image data, it is special to obtain the face Levy data.
With reference to the first implementation of the second aspect of the present invention, in second of realization side of the second aspect of the present invention In formula, the monitoring device also includes：
Processing unit, for carrying out dimension-reduction treatment to training sample according to principal component analysis PCA；
Determining unit, for determining the default k nearest neighbor from the training sample after dimension-reduction treatment by k nearest neighbor algorithm Sample.
Second with reference to the first of the second aspect of the present invention or second aspect implementation or second aspect is real Existing mode, in the third implementation of the second aspect of the present invention, the recognition unit specifically includes：
Matching module, for the face characteristic data to be matched with the default k nearest neighbor sample；
3rd computing module, for being obtained and the face characteristic by k nearest neighbor algorithm using the training sample matched Test sample corresponding to data；
4th computing module, for calculating the Euclidean distance between the test sample and the default k nearest neighbor sample；
First determining module, for after the Euclidean distance is less than preset threshold value, determining to wrap in facial feature database Containing the face characteristic data；
Second determining module, for after the Euclidean distance is more than preset threshold value, determining in facial feature database not Include the face characteristic data.
With reference to the third implementation of the second aspect of the present invention, in the 4th kind of realization side of the second aspect of the present invention In formula, the recognition unit also includes：
Output module, for after it is determined that not including the face characteristic data in the facial feature database, showing Alert picture, and the prompt tone that gives a warning.
As can be seen from the above technical solutions, the embodiment of the present invention has advantages below：Monitoring device obtains facial image Data；The monitoring device carries out local linear discriminant analysis LLDA to the face image data, extracts face characteristic data； The monitoring device carries out Classification and Identification according to default k nearest neighbor sample to the face characteristic data.The present invention is based on to big The processing of data, using distributed structure/architecture, rely on distributed treatment, distributed data base and cloud storage, the void of cloud computing Planization technology.Therefore, when data category distribution is more complicated, by carrying out local linear differentiation point to the face image data LLDA is analysed, so as to need to only train the sub-fraction of whole training set before testing, and according to default k nearest neighbor sample to the people Face characteristic carries out Classification and Identification, improves the precision and efficiency of recognition of face.By the above method, monitoring device is to place The personnel in region carry out man face image acquiring, are stored in database and are used as training sample.After the completion of collection, it is possible to dealing Personnel carry out man face image acquiring and are used as test sample.Described training sample be for establishing model, using they Feature can be exactly to be determined by the feature of training sample with structural classification device, some parameters of grader.Described test Sample may be considered new unknown sample, and test sample is classified by the grader having had built up.So After carry out recognition of face, the personnel not in database are irised out in display picture with rectangle frame, and send caution sound prompting Security personnel couple can be checked with personnel, to take precautions against the personal safety as well as the property safety for specifying place.Because this method is not required to train institute Have a sample, but a portion sample, be exactly in this way to a large amount of more suitable for the processing to big data, popular saying The processing of sample.And traditional LDA is not appropriate for the processing of large sample.
Brief description of the drawings
Fig. 1 is a kind of one embodiment schematic flow sheet of face identification method in the embodiment of the present invention；
Fig. 2 is a kind of another embodiment schematic flow sheet of face identification method in the embodiment of the present invention；
Fig. 3 is a kind of another embodiment schematic flow sheet of face identification method in the embodiment of the present invention；
Fig. 4 is a kind of another embodiment schematic flow sheet of face identification method in the embodiment of the present invention；
Fig. 5 is a kind of one embodiment structural representation of monitoring device in the embodiment of the present invention；
Fig. 6 is a kind of another example structure schematic diagram of monitoring device in the embodiment of the present invention；
Fig. 7 is a kind of another example structure schematic diagram of monitoring device in the embodiment of the present invention；
Fig. 8 is a kind of another example structure schematic diagram of monitoring device in the embodiment of the present invention；
Fig. 9 is a kind of another example structure schematic diagram of monitoring device in the embodiment of the present invention.
Embodiment
To enable goal of the invention, feature, the advantage of the present invention more obvious and understandable, below in conjunction with the present invention Accompanying drawing in embodiment, the technical scheme in the embodiment of the present invention is clearly and completely described, it is clear that disclosed below Embodiment be only part of the embodiment of the present invention, and not all embodiment.Based on the embodiment in the present invention, this area The every other embodiment that those of ordinary skill is obtained under the premise of creative work is not made, belongs to protection of the present invention Scope.
A kind of face identification method in the embodiment of the present invention is specifically described below, referring to Fig. 1, the present invention is implemented A kind of one embodiment of face identification method includes in example：
101st, monitoring device obtains face image data；
Monitoring device obtains face image data；It should be noted that monitoring device passes through USB (Universal Serial Bus, logical to get lines crossed) the existing face image data of any of interface or network node acquisition；Or pass through connection Photographing module obtains face image data, and specific acquisition modes are not construed as limiting herein.
102nd, above-mentioned monitoring device carries out local linear discriminant analysis LLDA to above-mentioned face image data, and extraction face is special Levy data；
Above-mentioned monitoring device carries out local linear discriminant analysis LLDA (Local Linear to above-mentioned face image data Discriminant Analysis, local linear discriminant analysis) extraction face characteristic data；It should be noted that above-mentioned part Linear discriminant analysis LLDA includes local linear discriminant analysis VLLDA (the Vector Local Linear based on vector Discriminant Analysis, the local linear discriminant analysis based on vector) or local linear discriminant analysis based on matrix MLLDA (Matrix Local Linear Discriminant Analysis, the local linear discriminant analysis based on matrix). Wherein, above-mentioned face image data is expressed as vector by the above-mentioned local linear discriminant analysis VLLDA based on vector, above-mentioned to be based on Above-mentioned face image data is expressed as matrix by the local linear discriminant analysis MLLDA of matrix.
103rd, above-mentioned monitoring device carries out Classification and Identification according to default k nearest neighbor sample to above-mentioned face characteristic data.
Above-mentioned monitoring device carries out Classification and Identification according to default k nearest neighbor sample to above-mentioned face characteristic data；Need Bright, above-mentioned default k nearest neighbor sample is K arest neighbors of the test sample determined in training sample, and K number does not limit It is fixed；It is compared with the facial feature database of preservation after classifying to above-mentioned face characteristic data, knows so as to complete face Other process.
In the embodiment of the present invention, monitoring device obtains face image data；Above-mentioned monitoring device is to above-mentioned facial image number According to local linear discriminant analysis LLDA is carried out, face characteristic data are extracted；Above-mentioned monitoring device is according to default k nearest neighbor sample pair Above-mentioned face characteristic data carry out Classification and Identification.The present invention, using distributed structure/architecture, is relied on based on the processing to big data Distributed treatment, distributed data base and the cloud storage of cloud computing, virtualization technology.Therefore, when data category distribution is more complicated When, by carrying out local linear discriminant analysis LLDA to above-mentioned face image data, so as to which only whole instruction need to be trained before testing Practice the sub-fraction of collection, and Classification and Identification is carried out to above-mentioned face characteristic data according to default k nearest neighbor sample, improve face The precision and efficiency of identification.
By the above method, monitoring device carries out man face image acquiring to the personnel of region, is stored in database As training sample.After the completion of collection, it is possible to carry out man face image acquiring to dealing personnel and be used as test sample.Described Training sample is for establishing model, can be exactly to lead to structural classification device, some parameters of grader using their feature The feature of training sample is crossed to determine.Described test sample may be considered new unknown sample, by having built The grader that has stood is classified to test sample.Then recognition of face is carried out, the personnel not in database are being shown Irised out in picture with rectangle frame, and send caution sound and remind security personnel couple to be checked with personnel, to take precautions against specified place Personal safety as well as the property safety.Because this method is not required to train all samples, but a portion sample, it is more suitable in this way In the processing to big data, popular saying is exactly the processing to great amount of samples.And traditional LDA is not appropriate for the processing of large sample.
Based on the face identification method in above-described embodiment, below for the specific of above-mentioned local linear discriminant analysis LLDA Process is illustrated, referring to Fig. 2, a kind of another embodiment of face identification method includes in the embodiment of the present invention：
201st, monitoring device obtains face image data；
Monitoring device obtains face image data；It should be noted that monitoring device passes through USB (Universal Serial Bus, logical to get lines crossed) the existing face image data of any of interface or network node acquisition；Or pass through connection Photographing module obtains face image data, and specific acquisition modes are not construed as limiting herein.
202nd, above-mentioned monitoring device is according to default k nearest neighbor sample, default first formula and default second formula difference Scatter matrix S in the class of above-mentioned face image data is calculatedwThe scatter matrix S between classb；
It should be noted that above-mentioned k nearest neighbor sample includes the set of the training sample of multiple classes；Above-mentioned default first formula For：
Above-mentioned default second formula is：
Wherein, c ' be k nearest neighbor sample classification number, m ' be k nearest neighbor sample average, l 'iFor the i-th class K in classification c ' The number of neighbour's sample, i.e.,m′iFor the average of the k nearest neighbor sample of the i-th class in classification c ', x 'ijFor classification c ' In the i-th class j-th of k nearest neighbor sample.
203rd, above-mentioned monitoring device is according to scatter matrix S in above-mentioned classw, scatter matrix S between above-mentioned classbAnd Fei Sheer is accurate The characteristic vector of face image data is then calculated；
It should be noted that above-mentioned specific algorithm refer to it is as follows：For example, in VLLDA algorithms, facial image number is inputted According to vector representation.For a given sample, from whole training sample xi∈RnThe sample is determined in (i=1,2 ..., L) K arest neighbors.This K arest neighbors is expressed as x 'i∈Rn(i=1,2 ..., K).They come from the individual classifications of c ', are expressed as c 'i(i =1,2 ..., c '), these classifications are a parts for whole classifications.If classification c 'iThere is l 'iIndividual fixed neighbour, i.e.,m′iRepresent in classification c 'iIn fixed neighbour average, x 'ijIt is the sample in classification c 'iIn j-th Fixed neighbour.Scatter matrix is as follows between scatter matrix and class in definition class：
Wherein, m ' is the average of all fixed neighbours.Then, the local Fei Sheer criterions based on vector can represent For：
It is understood that if each classification in the individual classifications of c ' only includes an arest neighbors of given sample, that Scatter matrix is exactly a null matrix in the class of these arest neighbors.So in above-mentioned VLLDA algorithms, it is only necessary to utilize class Between scatter matrix, and criterion reforms into：
On the other hand, if whole K arest neighbors of given sample are all from same classification, then they Scatter matrix is exactly a null matrix between class.So, it is only necessary to utilize scatter matrix in classAnd criterion becomes：
So for a given sample, make JVLF(Wv) maximize, and a projection matrix is obtained, the matrix Row are characteristic vectors, correspond to characteristic equationTop characteristic value, or matrixOrTop feature Value.
Different from VLLDA, the input data in MLLDA is represented as matrix.Assuming that there are the L instructions from c classification Practice sampleIn order to extract the feature of sample, we determine its K from training set first Arest neighbors.These neighbours are expressed as A 'i(i=1,2 ..., K), and come from the individual classification C ' of C 'i(i=1,2 ..., C '), this c ' is individual Classification is a part for whole classifications.Then, we define in the class of the K fixed neighbours of a sample scatter matrix and Scatter matrix is as follows between class：
Wherein, L 'iIt is in classification C 'iThe number of fixed neighbour in classification, i.e.,M′iRepresent in class Other C 'iIn fixed neighbour average, A 'ijIt is in classification C 'iIn j-th of fixed neighbour, M ' is represented to random sample sheet K fixed neighbours average.So the local Fei Sheer criterions based on matrix can be expressed as：
Similarly, if each classification in the individual classifications of C ' only has a fixed neighbour, then this K fixed Scatter matrix is a null matrix in the class of neighbour.So MLLDA only needs to utilize scatter matrix between class, and differentiate standard It is then：
On the other hand, if whole K arest neighbors of given sample are all from same classification, then they Scatter matrix is exactly a null matrix between class.So we only need to utilize scatter matrix in class, and criterion becomes Into：
It is understood that for a given sample, make JMLF(Wm) maximize, and a projection matrix is obtained, The matrix column is characteristic vector, correspond to characteristic equationTop characteristic value, or matrixOrTop characteristic value.
204th, above-mentioned monitoring device carries out feature extraction to the characteristic vector of above-mentioned face image data, obtains above-mentioned face Characteristic；
Above-mentioned monitoring device carries out feature extraction to the characteristic vector of above-mentioned face image data, obtains above-mentioned face characteristic Data.
205th, above-mentioned monitoring device carries out Classification and Identification according to default k nearest neighbor sample to above-mentioned face characteristic data.
Above-mentioned monitoring device carries out Classification and Identification according to default k nearest neighbor sample to above-mentioned face characteristic data；Above-mentioned prison Control equipment and Classification and Identification is carried out to above-mentioned face characteristic data according to default k nearest neighbor sample；It is it should be noted that above-mentioned default K nearest neighbor sample be the K arest neighbors of test sample determined in training sample, K number is not construed as limiting；It is special to above-mentioned face Sign data are compared after being classified with the facial feature database of preservation, so as to complete face recognition process；Above-mentioned classification Nearest neighbor classifier is may be selected in method or other graders are classified to above-mentioned face characteristic data, is not restricted specifically.
In the embodiment of the present invention, above-mentioned monitoring device is according to default k nearest neighbor sample, default first formula and default the Scatter matrix between scatter matrix and class is calculated in the class of above-mentioned face image data in two formula respectively, and accurate using Fei Sheer The characteristic vector of face image data is then calculated, so as to extract face characteristic data by Partial data structure, improves The precision and efficiency of recognition of face.
Based on the face identification method in above-described embodiment, in the embodiment of the present invention, the recognition of face for large database concept, Data Dimensionality Reduction first can be carried out to it, to carry out characteristic extraction, referring to Fig. 3, a kind of face is known in the embodiment of the present invention Another embodiment of other method includes：
301st, monitoring device obtains face image data；
Monitoring device obtains face image data；It should be noted that monitoring device passes through in USB interface or network node Any one obtains existing face image data；Or face image data is obtained by the photographing module of connection, it is specific to obtain Mode is not construed as limiting herein.
302nd, above-mentioned monitoring device carries out dimension-reduction treatment according to principal component analysis PCA to training sample；
Above-mentioned monitoring device is according to principal component analysis PCA (Principal Component Analysis, principal component point Analysis) dimension-reduction treatment is carried out to training sample；PCA conversion is carried out to training sample first, obtains the projection square after PCA projections Battle array.
303rd, above-mentioned monitoring device determines above-mentioned default K by k nearest neighbor algorithm from the training sample after dimension-reduction treatment Neighbour's sample；
Above-mentioned monitoring device determines above-mentioned default k nearest neighbor by k nearest neighbor algorithm from the training sample after dimension-reduction treatment Sample；K nearest neighbor algorithm refers to prior art, wherein it is determined that above-mentioned default k nearest neighbor sample is true i.e. from above-mentioned training sample Determine K nearest samples of test sample.
304th, above-mentioned monitoring device is according to default k nearest neighbor sample, default first formula and default second formula difference Scatter matrix S in the class of above-mentioned face image data is calculatedwThe scatter matrix S between classb；
It should be noted that above-mentioned k nearest neighbor sample includes the set of the training sample of multiple classes；Above-mentioned default first formula For：
Above-mentioned default second formula is：
Wherein, c ' be k nearest neighbor sample classification number, m ' be k nearest neighbor sample average, l 'iFor the i-th class K in classification c ' The number of neighbour's sample, i.e.,m′iFor the average of the k nearest neighbor sample of the i-th class in classification c ', x 'ijFor classification c ' In the i-th class j-th of k nearest neighbor sample.
305th, above-mentioned monitoring device is according to scatter matrix S in above-mentioned classw, scatter matrix S between above-mentioned classbAnd Fei Sheer is accurate The characteristic vector of face image data is then calculated；
Specific algorithm refers to step 203, and here is omitted.
306th, above-mentioned monitoring device carries out feature extraction to the characteristic vector of above-mentioned face image data, obtains above-mentioned face Characteristic；
Above-mentioned monitoring device carries out feature extraction to the characteristic vector of above-mentioned face image data, obtains above-mentioned face characteristic Data.
307th, above-mentioned monitoring device carries out Classification and Identification according to default k nearest neighbor sample to above-mentioned face characteristic data.
Above-mentioned monitoring device carries out Classification and Identification according to default k nearest neighbor sample to above-mentioned face characteristic data；Above-mentioned prison Control equipment and Classification and Identification is carried out to above-mentioned face characteristic data according to default k nearest neighbor sample；It is it should be noted that above-mentioned default K nearest neighbor sample be the K arest neighbors of test sample determined in training sample, K number is not construed as limiting；It is special to above-mentioned face Sign data are compared after being classified with the facial feature database of preservation, so as to complete face recognition process；Above-mentioned classification Nearest neighbor classifier is may be selected in method or other graders are classified to above-mentioned face characteristic data, is not restricted specifically.
In the embodiment of the present invention, above-mentioned monitoring device carries out dimension-reduction treatment according to principal component analysis PCA to training sample, and Above-mentioned default k nearest neighbor sample is determined from the training sample after dimension-reduction treatment by k nearest neighbor algorithm；So as to more accurately from Face is identified in huge database.
Based on the face identification method in above-described embodiment, in the embodiment of the present invention, above-mentioned face characteristic data will be entered Row Classification and Identification is described specifically, referring to Fig. 4, in the embodiment of the present invention a kind of face identification method another embodiment bag Include：
401st, monitoring device obtains face image data；
Monitoring device obtains face image data；It should be noted that monitoring device passes through in USB interface or network node Any one obtains existing face image data；Or face image data is obtained by the photographing module of connection, it is specific to obtain Mode is not construed as limiting herein.
402nd, above-mentioned monitoring device carries out dimension-reduction treatment according to principal component analysis PCA to training sample；
Above-mentioned monitoring device carries out dimension-reduction treatment according to principal component analysis PCA to training sample；I.e. first to training sample PCA conversion is carried out, obtains the projection matrix after PCA projections.
403rd, above-mentioned monitoring device determines above-mentioned default K by k nearest neighbor algorithm from the training sample after dimension-reduction treatment Neighbour's sample；
Above-mentioned monitoring device determines above-mentioned default k nearest neighbor by k nearest neighbor algorithm from the training sample after dimension-reduction treatment Sample；K nearest neighbor algorithm refers to prior art, wherein it is determined that above-mentioned default k nearest neighbor sample is true i.e. from above-mentioned training sample Determine K nearest samples of test sample.
404th, above-mentioned monitoring device is according to default k nearest neighbor sample, default first formula and default second formula difference Scatter matrix S in the class of above-mentioned face image data is calculatedwThe scatter matrix S between classb；
It should be noted that above-mentioned k nearest neighbor sample includes the set of the training sample of multiple classes；Above-mentioned default first formula For：
Above-mentioned default second formula is：
Wherein, c ' be k nearest neighbor sample classification number, m ' be k nearest neighbor sample average, l 'iFor the i-th class K in classification c ' The number of neighbour's sample, i.e.,m′iFor the average of the k nearest neighbor sample of the i-th class in classification c ', x 'ijFor classification c ' In the i-th class j-th of k nearest neighbor sample.
405th, above-mentioned monitoring device is according to scatter matrix S in above-mentioned classw, scatter matrix S between above-mentioned classbAnd Fei Sheer is accurate The characteristic vector of face image data is then calculated；
Specific algorithm refers to step 203, and here is omitted.
406th, above-mentioned monitoring device carries out feature extraction to the characteristic vector of above-mentioned face image data, obtains above-mentioned face Characteristic；
Above-mentioned monitoring device carries out feature extraction to the characteristic vector of above-mentioned face image data, obtains above-mentioned face characteristic Data.
407th, above-mentioned monitoring device is matched above-mentioned face characteristic data with above-mentioned default k nearest neighbor sample；
Above-mentioned monitoring device is matched above-mentioned face characteristic data with above-mentioned default k nearest neighbor sample.
408th, above-mentioned monitoring device is obtained and above-mentioned face characteristic number using the training sample matched by k nearest neighbor algorithm According to corresponding test sample；
Above-mentioned monitoring device is obtained and above-mentioned face characteristic data pair using the training sample matched by k nearest neighbor algorithm The test sample answered.
409th, above-mentioned monitoring device calculates the Euclidean distance between above-mentioned test sample and above-mentioned default k nearest neighbor sample；
Above-mentioned monitoring device calculates the Euclidean distance between above-mentioned test sample and above-mentioned default k nearest neighbor sample.
410th, when above-mentioned Euclidean distance is less than preset threshold value, determine to include above-mentioned face characteristic in facial feature database Data；When above-mentioned Euclidean distance is more than preset threshold value, determine not including above-mentioned face characteristic data in facial feature database.
When above-mentioned Euclidean distance is less than preset threshold value, determine to include above-mentioned face characteristic number in facial feature database According to；When above-mentioned Euclidean distance is more than preset threshold value, determine not including above-mentioned face characteristic data in facial feature database.Need It is noted that when it is determined that not including above-mentioned face characteristic data in above-mentioned facial feature database, the monitoring device is shown Alert picture, and the prompt tone that gives a warning；The facial image got can specifically be irised out with rectangle frame in display picture, And send caution sound and remind security personnel to check a suspect, to ensure the personal safety as well as the property safety for specifying place.
In the embodiment of the present invention, above-mentioned monitoring device is calculated between above-mentioned test sample and above-mentioned default k nearest neighbor sample Euclidean distance, when above-mentioned Euclidean distance is less than preset threshold value, determine to include above-mentioned face characteristic in facial feature database Data；When above-mentioned Euclidean distance is more than preset threshold value, determine not including above-mentioned face characteristic data in facial feature database, And warning picture is shown, give a warning prompt tone, so as to complete face recognition process.
A kind of face identification method in the embodiment of the present invention is described above, below from the angle of device to this hair A kind of monitoring device in bright embodiment is described in detail, referring to Fig. 5, in the embodiment of the present invention a kind of monitoring device one Individual embodiment includes：
Acquiring unit 501, for obtaining face image data；
Analytic unit 502, for carrying out local linear discriminant analysis LLDA to above-mentioned face image data, extraction face is special Levy data；
Recognition unit 503, for carrying out Classification and Identification to above-mentioned face characteristic data according to default k nearest neighbor sample.
It should be noted that the mode of above-mentioned acquisition face image data can be by any in USB interface or network node It is individual to obtain existing face image data；Or face image data, specific acquisition modes are obtained by the photographing module of connection It is not construed as limiting herein.
Above-mentioned local linear discriminant analysis LLDA is included based on vectorial local linear discriminant analysis VLLDA or based on matrix Local linear discriminant analysis MLLDA.Wherein, the above-mentioned local linear discriminant analysis VLLDA based on vector is by above-mentioned face figure As data are expressed as vector, above-mentioned face image data is expressed as by the above-mentioned local linear discriminant analysis MLLDA based on matrix Matrix.
Above-mentioned default k nearest neighbor sample is K arest neighbors of the test sample determined in training sample, and K number is not made Limit；It is compared after classifying to above-mentioned face characteristic data with the facial feature database of preservation, so as to complete face Identification process, the processing mode of above-mentioned face image data can pass through extra large dupp Hadoop processing.
In the embodiment of the present invention, above-mentioned acquiring unit 501 obtains face image data；Above-mentioned analytic unit 502 is to above-mentioned Face image data carries out local linear discriminant analysis LLDA, extracts face characteristic data；Above-mentioned recognition unit 503 is according to default K nearest neighbor sample to above-mentioned face characteristic data carry out Classification and Identification.Therefore, when data category distribution is more complicated, by right Above-mentioned face image data carries out local linear discriminant analysis LLDA, the one of whole training set small so as to need to only train before testing Part, and Classification and Identification is carried out to above-mentioned face characteristic data according to default k nearest neighbor sample, improve the precision of recognition of face And efficiency.
Based on the monitoring device in above-described embodiment, optionally, as shown in fig. 6, above-mentioned analysis list in the embodiment of the present invention Member 502 specifically includes：
First computing module 601, for according to above-mentioned default k nearest neighbor sample, default first formula and default second Scatter matrix Sb between scatter matrix Sw and class is calculated in the class of above-mentioned face image data in formula respectively；
Second computing module 602, for according to scatter matrix Sb and taking house between scatter matrix Sw, above-mentioned class in above-mentioned class The characteristic vector of face image data is calculated in your criterion；
Extraction module 603, for carrying out feature extraction to the characteristic vector of above-mentioned face image data, obtain above-mentioned face Characteristic.
Detailed process refers to as follows：For example, in VLLDA algorithms, face image data vector representation is inputted.For One given sample, from whole training sample xi∈RnK arest neighbors of the sample is determined in (i=1,2 ..., L).This K Arest neighbors is expressed as x 'i∈Rn(i=1,2 ..., K).They come from the individual classifications of c ', are expressed as c 'i(i=1,2 ..., c '), these Classification is a part for whole classifications.If classification c 'iThere is l 'iIndividual fixed neighbour, i.e.,m′iRepresent in classification c′iIn fixed neighbour average, x 'ijIt is the sample in classification c 'iIn j-th of fixed neighbour.Define in class Scatter matrix is as follows between scatter matrix and class：
Wherein, m ' is the average of all fixed neighbours.Then, the local Fei Sheer criterions based on vector can represent For：
It is understood that if each classification in the individual classifications of c ' only includes an arest neighbors of given sample, that Scatter matrix is exactly a null matrix in the class of these arest neighbors.So in above-mentioned VLLDA algorithms, it is only necessary to utilize class Between scatter matrixAnd criterion reforms into：
On the other hand, if whole K arest neighbors of given sample are all from same classification, then they Scatter matrix is exactly a null matrix between class.So, it is only necessary to utilize scatter matrix in classAnd criterion becomes：
So for a given sample, make JVLF(Wv) maximize, and a projection matrix is obtained, the matrix Row are characteristic vectors, correspond to characteristic equationTop characteristic value, or matrixOrTop feature Value.
Different from VLLDA, the input data in MLLDA is represented as matrix.Assuming that there are the L instructions from c classification Practice sample.In order to extract the feature of sample, we determine its K from training set first Arest neighbors.These neighbours are expressed as A 'i(i=1,2 ..., K), and come from the individual classification C ' of C 'i(i=1,2 ..., C '), this c ' is individual Classification is a part for whole classifications.Then, we define in the class of the K fixed neighbours of a sample scatter matrix and Scatter matrix is as follows between class：
Wherein, L 'iIt is in classification C 'iThe number of fixed neighbour in classification, i.e.,M′iRepresent in class Other C 'iIn fixed neighbour average, A 'ijIt is in classification C 'iIn j-th of fixed neighbour, M ' is represented to random sample sheet K fixed neighbours average.So the local Fei Sheer criterions based on matrix can be expressed as：
Similarly, if each classification in the individual classifications of C ' only has a fixed neighbour, then this K fixed Scatter matrix is a null matrix in the class of neighbour.So MLLDA only needs to utilize scatter matrix between classAnd differentiate standard It is then：
On the other hand, if whole K arest neighbors of given sample are all from same classification, then they Scatter matrix is exactly a null matrix between class.So we only need to utilize scatter matrix in classAnd criterion becomes Into：
It is understood that for a given sample, make JMLF(Wm) maximize, and a projection matrix is obtained, The matrix column is characteristic vector, correspond to characteristic equationTop characteristic value, or matrixOrTop characteristic value；Feature extraction is carried out after obtaining features described above vector, so as to obtain above-mentioned face characteristic data.
In the embodiment of the present invention, the first computing module 601 is according to default k nearest neighbor sample, default first formula and pre- If scatter matrix between scatter matrix and class is calculated in the class of above-mentioned face image data in the second formula respectively, and by above-mentioned The characteristic vector of face image data is calculated using Fei Sheer criterions for second computing module 602, so as to pass through local data Structure extraction face characteristic data, improve the precision and efficiency of recognition of face.
Based on the monitoring device in above-described embodiment, optionally, as shown in fig. 7, above-mentioned monitoring is set in the embodiment of the present invention It is standby also to include：
Processing unit 701, for carrying out dimension-reduction treatment to training sample according to principal component analysis PCA；
Determining unit 702, for determining above-mentioned default K from the training sample after dimension-reduction treatment by k nearest neighbor algorithm Neighbour's sample.
It should be noted that above-mentioned carry out dimension-reduction treatment i.e. to training sample according to principal component analysis PCA to training sample PCA conversion is carried out, obtains the projection matrix after PCA projections；Above-mentioned k nearest neighbor algorithm refers to prior art, wherein it is determined that above-mentioned Default k nearest neighbor sample is the K nearest samples that test sample is determined from above-mentioned training sample.
In the embodiment of the present invention, above-mentioned processing unit 701 is carried out at dimensionality reduction according to principal component analysis PCA to training sample Reason, above-mentioned determining unit 702 determine above-mentioned default k nearest neighbor sample by k nearest neighbor algorithm from the training sample after dimension-reduction treatment This；So as to which more accurately face be identified from huge database.
Based on the monitoring device in above-described embodiment, optionally, as shown in figure 8, above-mentioned identification list in the embodiment of the present invention Member 503 specifically includes：
Matching module 801, for above-mentioned face characteristic data to be matched with above-mentioned default k nearest neighbor sample；
3rd computing module 802, it is special with above-mentioned face for being obtained using the training sample matched by k nearest neighbor algorithm Levy test sample corresponding to data；
4th computing module 803, for calculate the Euclidean between above-mentioned test sample and above-mentioned default k nearest neighbor sample away from From；
First determining module 804, for after above-mentioned Euclidean distance is less than preset threshold value, determining in facial feature database Include above-mentioned face characteristic data；
Second determining module 805, for after above-mentioned Euclidean distance is more than preset threshold value, determining in facial feature database Not comprising above-mentioned face characteristic data.
In the embodiment of the present invention, above-mentioned 4th computing module 803 calculates above-mentioned test sample and above-mentioned default k nearest neighbor sample Euclidean distance between this, when above-mentioned Euclidean distance is less than preset threshold value, above-mentioned first determining module 804 determines face characteristic Above-mentioned face characteristic data are included in database；When above-mentioned Euclidean distance is more than preset threshold value, above-mentioned second determining module 805 Determine not including above-mentioned face characteristic data in facial feature database, so as to complete face recognition process.
Based on the monitoring device in above-described embodiment, optionally, as shown in figure 9, above-mentioned identification list in the embodiment of the present invention Member 503 also includes：
Output module 901, for after it is determined that not including above-mentioned face characteristic data in above-mentioned facial feature database, showing Warn and accuse picture, and the prompt tone that gives a warning.
It should be noted that the specific implementation of the above-mentioned display of output module 901 warning picture can be in display picture The facial image got is irised out with rectangle frame in face, and sends caution sound and reminds security personnel to examine a suspect Look into, to ensure the personal safety as well as the property safety for specifying place.
In the embodiment of the present invention, after it is determined that not including above-mentioned face characteristic data in above-mentioned facial feature database, lead to The above-mentioned display of output module 901 warning picture, and the prompt tone that gives a warning are crossed, to make the precautionary measures in time, ensures and specifies The personal safety as well as the property safety in place.
The above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although with reference to the foregoing embodiments The present invention is described in detail, it will be understood by those within the art that：It still can be to foregoing each implementation Technical scheme described in example is modified, or carries out equivalent substitution to which part technical characteristic；And these modification or Replace, the essence of appropriate technical solution is departed from the spirit and scope of various embodiments of the present invention technical scheme.

Claims (8)

  1. A kind of 1. face identification method, it is characterised in that including：
    Monitoring device obtains face image data；
    The monitoring device is calculated respectively according to default k nearest neighbor sample, default first formula and default second formula Scatter matrix S in the class of the face image datawThe scatter matrix S between classb, the k nearest neighbor sample includes the training of multiple classes The set of sample；
    Default first formula is：
    Default second formula is：
    Wherein, c' be k nearest neighbor sample classification number, m' be k nearest neighbor sample average, l 'iFor the i-th class k nearest neighbor sample in classification c' This number, i.e.,m′iFor the average of the k nearest neighbor sample of the i-th class in classification c', x'ijFor the i-th class in classification c' J-th of k nearest neighbor sample；
    The monitoring device is according to scatter matrix S in the classw, scatter matrix S between the classbAnd Fei Sheer criterions calculate To the characteristic vector of face image data；
    The monitoring device carries out feature extraction to the characteristic vector of the face image data, obtains the face characteristic number According to；
    The monitoring device carries out Classification and Identification according to default k nearest neighbor sample to the face characteristic data.
  2. 2. face identification method according to claim 1, it is characterised in that the monitoring device is according to the default K Neighbour's sample, default first formula and default second formula are calculated in the class of the face image data and spread square respectively Battle array SwThe scatter matrix S between classbAlso include before：
    The monitoring device carries out dimension-reduction treatment according to principal component analysis PCA to training sample；
    The monitoring device determines the default k nearest neighbor sample by k nearest neighbor algorithm from the training sample after dimension-reduction treatment.
  3. 3. according to the face identification method described in any one of claim 1 to 2, it is characterised in that the monitoring device is according to pre- If k nearest neighbor sample to the face characteristic data carry out Classification and Identification specifically include：
    The monitoring device is matched the face characteristic data with the default k nearest neighbor sample；
    The monitoring device is obtained corresponding with the face characteristic data using the training sample matched by k nearest neighbor algorithm Test sample；
    The monitoring device calculates the Euclidean distance between the test sample and the default k nearest neighbor sample；
    When the Euclidean distance is less than preset threshold value, determine to include the face characteristic data in facial feature database；When When the Euclidean distance is more than preset threshold value, determine not including the face characteristic data in facial feature database.
  4. 4. face identification method according to claim 3, it is characterised in that do not wrapped in the determination facial feature database Include afterwards containing the face characteristic data：
    The monitoring device display warning picture, and the prompt tone that gives a warning.
  5. A kind of 5. monitoring device, it is characterised in that including：
    Acquiring unit, for obtaining face image data；
    First computing unit, for being calculated respectively according to default k nearest neighbor sample, default first formula and default second formula Obtain scatter matrix S in the class of the face image datawThe scatter matrix S between classb, the k nearest neighbor sample includes multiple classes The set of training sample；
    Default first formula is：
    Default second formula is：
    Wherein, c' be k nearest neighbor sample classification number, m' be k nearest neighbor sample average, l 'iFor the i-th class k nearest neighbor sample in classification c' This number, i.e.,m′iFor the average of the k nearest neighbor sample of the i-th class in classification c', x'ijFor the i-th class in classification c' J-th of k nearest neighbor sample；
    Second computing unit, for according to scatter matrix S in the classw, scatter matrix S between the classbAnd Fei Sheer criterion meters Calculation obtains the characteristic vector of face image data；
    Extraction unit, for carrying out feature extraction to the characteristic vector of the face image data, obtain the face characteristic number According to；
    Recognition unit, for carrying out Classification and Identification to the face characteristic data according to default k nearest neighbor sample.
  6. 6. monitoring device according to claim 5, it is characterised in that the monitoring device also includes：
    Processing unit, for carrying out dimension-reduction treatment to training sample according to principal component analysis PCA；
    Determining unit, for determining the default k nearest neighbor sample from the training sample after dimension-reduction treatment by k nearest neighbor algorithm This.
  7. 7. according to the monitoring device described in any one of claim 5 to 6, it is characterised in that the recognition unit specifically includes：
    Matching module, for the face characteristic data to be matched with the default k nearest neighbor sample；
    3rd computing module, for being obtained and the face characteristic data by k nearest neighbor algorithm using the training sample matched Corresponding test sample；
    4th computing module, for calculating the Euclidean distance between the test sample and the default k nearest neighbor sample；
    First determining module, for after the Euclidean distance is less than preset threshold value, determining to include institute in facial feature database State face characteristic data；
    Second determining module, for after the Euclidean distance is more than preset threshold value, determining not including in facial feature database The face characteristic data.
  8. 8. monitoring device according to claim 7, it is characterised in that the recognition unit also includes：
    Output module, for after it is determined that not including the face characteristic data in the facial feature database, display to alert Picture, and the prompt tone that gives a warning.
CN201410658870.8A 2014-11-18 2014-11-18 A kind of face identification method and monitoring device Active CN104318224B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201410658870.8A CN104318224B (en) 2014-11-18 2014-11-18 A kind of face identification method and monitoring device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201410658870.8A CN104318224B (en) 2014-11-18 2014-11-18 A kind of face identification method and monitoring device

Publications (2)

Publication Number Publication Date
CN104318224A CN104318224A (en) 2015-01-28
CN104318224B true CN104318224B (en) 2018-03-23

Family

ID=52373453

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201410658870.8A Active CN104318224B (en) 2014-11-18 2014-11-18 A kind of face identification method and monitoring device

Country Status (1)

Country Link
CN (1) CN104318224B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108256405A (en) * 2016-12-29 2018-07-06 中国移动通信有限公司研究院 A kind of face identification method and device

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108573235A (en) * 2018-04-20 2018-09-25 天津科技大学 The compact graph structure face characteristic extraction algorithm of multi-direction weight optimization
CN109255377A (en) * 2018-08-30 2019-01-22 北京信立方科技发展股份有限公司 Instrument recognition methods, device, electronic equipment and storage medium
CN109558535B (en) * 2018-11-05 2020-08-07 重庆中科云从科技有限公司 Personalized article pushing method and system based on face recognition
CN115188053A (en) * 2022-07-21 2022-10-14 浪潮卓数大数据产业发展有限公司 Automatic alarm method and system for detecting intrusion of stranger in residential building

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101706871A (en) * 2009-11-05 2010-05-12 上海交通大学 Isometric mapping based facial image recognition method
CN103632144A (en) * 2013-12-05 2014-03-12 江苏大学 Fuzzy two-dimensional k-nearest neighbor based face recognition method

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR100543707B1 (en) * 2003-12-04 2006-01-20 삼성전자주식회사 Face recognition method and apparatus using PCA learning per subgroup

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101706871A (en) * 2009-11-05 2010-05-12 上海交通大学 Isometric mapping based facial image recognition method
CN103632144A (en) * 2013-12-05 2014-03-12 江苏大学 Fuzzy two-dimensional k-nearest neighbor based face recognition method

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108256405A (en) * 2016-12-29 2018-07-06 中国移动通信有限公司研究院 A kind of face identification method and device

Also Published As

Publication number Publication date
CN104318224A (en) 2015-01-28

Similar Documents

Publication Publication Date Title
CN107194341B (en) Face recognition method and system based on fusion of Maxout multi-convolution neural network
CN108304788B (en) Face recognition method based on deep neural network
CN107871100B (en) Training method and device of face model, and face authentication method and device
Hoang Ngan Le et al. Robust hand detection and classification in vehicles and in the wild
WO2015101080A1 (en) Face authentication method and device
CN104318224B (en) A kind of face identification method and monitoring device
CN102663370B (en) Face identification method and system
CN104700094B (en) A kind of face identification method for intelligent robot and system
CN105956572A (en) In vivo face detection method based on convolutional neural network
CN103198303B (en) A kind of gender identification method based on facial image
CN105512620B (en) The training method and device of convolutional neural networks for recognition of face
CN105528584B (en) A kind of detection method and device of face image
CN104134364B (en) Real-time traffic sign identification method and system with self-learning capacity
JP2022521038A (en) Face recognition methods, neural network training methods, devices and electronic devices
CN105138954A (en) Image automatic screening, query and identification system
CN105138968A (en) Face authentication method and device
Wang et al. Face live detection method based on physiological motion analysis
CN105138951B (en) Human face portrait-photo array the method represented based on graph model
CN106056067B (en) Low-resolution face image recognition methods based on corresponding relationship prediction
Mahmud et al. Face detection and recognition system
Del Coco et al. Assessment of deep learning for gender classification on traditional datasets
CN111191519B (en) Living body detection method for user access of mobile power supply device
Farinella et al. Natural versus artificial scene classification by ordering discrete fourier power spectra
Ahlawat et al. Online invigilation: A holistic approach
CN104091174B (en) portrait style classification method based on support vector machine

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
TR01 Transfer of patent right
TR01 Transfer of patent right

Effective date of registration: 20180906

Address after: 523710 26 Pingshan 188 Industrial Avenue, Tangxia Town, Dongguan, Guangdong

Patentee after: Central Intelligence Polytron Technologies Inc

Address before: 518129 Guangdong Shenzhen Longgang District Bantian five and Xinhua new industrial zone, central control building 1 to nine floors.

Patentee before: Shenzhen Zhongkong Biometrics Technology Co., Ltd.

CP01 Change in the name or title of a patent holder
CP01 Change in the name or title of a patent holder

Address after: 523710, 26, 188 Industrial Road, Pingshan Town, Guangdong, Dongguan, Tangxia

Patentee after: Entropy Technology Co., Ltd

Address before: 523710, 26, 188 Industrial Road, Pingshan Town, Guangdong, Dongguan, Tangxia

Patentee before: Zhongkong Smart Technology Co.,Ltd.

CP02 Change in the address of a patent holder
CP02 Change in the address of a patent holder

Address after: No.32, Pingshan Industrial Road, Tangxia Town, Dongguan City, Guangdong Province, 523710

Patentee after: Entropy Technology Co., Ltd

Address before: 523710 26 Pingshan 188 Industrial Avenue, Tangxia Town, Dongguan, Guangdong

Patentee before: Entropy Technology Co., Ltd