CN106529377A - Age estimating method, age estimating device and age estimating system based on image - Google Patents

Age estimating method, age estimating device and age estimating system based on image Download PDF

Info

Publication number
CN106529377A
CN106529377A CN201510586848.1A CN201510586848A CN106529377A CN 106529377 A CN106529377 A CN 106529377A CN 201510586848 A CN201510586848 A CN 201510586848A CN 106529377 A CN106529377 A CN 106529377A
Authority
CN
China
Prior art keywords
sample
age
detected
face
estimation
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.)
Pending
Application number
CN201510586848.1A
Other languages
Chinese (zh)
Inventor
陶海
林宇
柴兆虎
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.)
Wen'an Beijing Intelligent Technology Ltd By Share Ltd
Original Assignee
Wen'an Beijing Intelligent Technology Ltd By Share 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 Wen'an Beijing Intelligent Technology Ltd By Share Ltd filed Critical Wen'an Beijing Intelligent Technology Ltd By Share Ltd
Priority to CN201510586848.1A priority Critical patent/CN106529377A/en
Publication of CN106529377A publication Critical patent/CN106529377A/en
Pending legal-status Critical Current

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
    • G06V40/171—Local features and components; Facial parts ; Occluding parts, e.g. glasses; Geometrical relationships
    • 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
    • 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
    • G06V40/178—Human faces, e.g. facial parts, sketches or expressions estimating age from face image; using age information for improving recognition

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 an age estimating method, an age estimating device and an age estimating system based on an image, wherein the age estimating method based on the image comprises the steps of acquiring a to-be-detected face image and corresponding key points, wherein the key points are angular point positions of main parts; extracting an original characteristic of the to-be-detected face according to the to-be-detected face image and corresponding key points; performing dimension reduction processing on the original characteristic of the to-be-detected face, and acquiring a low-dimension characteristic vector of the to-be-detected face; and performing age estimation on the low-dimension characteristic vector of the to-be-detected face through an age estimating model. According to the age estimating method, the age estimating device and the age estimating system, through a face representation method for performing dimension reduction processing on the original characteristic, and the age estimating model, an average estimation error which is lower than 5.5 can be realized in dynamic scene application in age estimation, and an accuracy in a range in which an absolute error is lower than ten years reaches 88%. Furthermore, training of the age estimating model does not require identity information of samples, thereby greatly facilitating acquisition of training samples.

Description

It is a kind of based on the age estimation method of image, apparatus and system
Technical field
The present invention relates to computer biometrics technology, more particularly to a kind of based on the age estimation method of image, apparatus and system.
Background technology
With going deep into for Research on Face Recognition Technology, one of popular research topic of computer field of biological recognition is had become based on the estimation of Age of facial image.The estimation of Age or verification mode taken in the case of Most current depend on the subjective estimation of people, or depend on related certificate, such as passport, identity card etc..These mode speed are slow, it is high, unfriendly to spend, the shortcomings of easily forge.Based on disadvantage mentioned above, prior art employs some and realizes estimation of Age based on face recognition technology, so that many original estimation of Age can also be realized covering using the technology where being difficult to implement, the age distribution of shop personnel is analyzed in such as market monitoring, and the automation services terminal of special service is provided for all ages and classes user etc., therefore the technology has wide market prospect.
It is as follows about realizing that the method for estimation of Age implements process by face recognition technology in prior art：The first step, intercepts human face region to be detected；Second step, detects the key point of the human face region；3rd step, extracts the characteristic vector of facial image to be estimated, generates candidate's age pullulation module vector；4th step, judges whether pullulation module subspace trains；If not training the pullulation module subspace, pullulation module subspace is trained；If training the pullulation module subspace, into next step；5th step, according to the pullulation module subspace for training, by all candidate's age pullulation module vector projections of facial image to be estimated in pullulation module subspace, then reconstructs complete candidate's age pullulation module vector from projection vector；6th step, finds best age pullulation module by the reconstructed error compared between reconstructed image and original image, and its current age is estimated in position of the facial image to be estimated in best age pullulation module.
Therefore, during inventor's design face recognition technology realizes the method for estimation of Age, it is found that at least there are the following problems in prior art：
Above-mentioned face recognition technology realizes that the shortcoming of estimation of Age scheme is：First, the face representation be it is a kind of owe complete expression, do not give full expression to relevant with age information in facial image；Second, the training need sample of pullulation module subspace is while two kinds of information of labelling age and identity, it is generally the case that this kind of sample is difficult to a large amount of acquisitions；3rd, estimation of Age needs more or less a hundred candidate's pullulation module vector is projected and reconstructed, the pullulation module minimum to find reconstructed error, and this process is numerous and diverse and time-consuming.
The content of the invention
In view of the above problems, it is proposed that the present invention so as to provide it is a kind of overcome the problems referred to above or solve the above problems at least in part, the technical scheme is that what is be achieved in that：
On the one hand, the invention provides a kind of age estimation method based on image, including：
Obtain facial image to be detected and correspondence key point；Corner location of the key point for main portions；
According to the facial image to be detected and correspondence key point, face primitive character to be detected is extracted；
The face primitive character to be detected is carried out into dimension-reduction treatment, face low dimensional characteristic vector to be detected is obtained；
Estimation of Age is carried out to the face low dimensional characteristic vector to be detected by estimation of Age model.
Preferably, the method also includes：
It is described according to the facial image to be detected and correspondence key point, extract face primitive character step to be detected, specifically include：
According to the facial image to be detected and correspondence key point, face local feature to be detected is obtained；
The face local feature to be detected is connected in series, face primitive character to be detected is obtained；
It is described that the face primitive character to be detected is carried out into dimension-reduction treatment, face low dimensional characteristic vector step to be detected is obtained, is specifically included：
Obtain Feature Dimension Reduction matrix and the face primitive character to be detected；
By the Feature Dimension Reduction matrix, dimension-reduction treatment is carried out to the face primitive character to be detected, obtain face low dimensional characteristic vector to be detected.
Preferably, the method also includes：
Obtain the age of the set of facial image sample information and correspondence image sample；
Determine the key point of each sample in the facial image sample information set；Corner location of the key point for main portions；
According to each sample and its key point in the facial image sample information set, the primitive character of each sample in the facial image sample information set is obtained；
The primitive character of each sample in the facial image sample information set is carried out into dimension-reduction treatment, the low dimensional characteristic vector of each sample is obtained；
Model is estimated by the corresponding age value sport career age of the low dimensional characteristic vector of each sample, the estimation of Age model is obtained, in case used by follow-up estimation of Age.
Preferably, described according to each sample and its key point in the facial image sample information set, the primitive character step for obtaining each sample in the facial image sample information set includes：
According to the key point of each sample in the facial image sample information set, the local feature of each sample is obtained；
Same sample image local feature in the facial image sample information set is connected in series, each sample primitive character is obtained.
Preferably, the primitive character by each sample in the facial image sample information set carries out dimension-reduction treatment, obtains the low dimensional characteristic vector step of each sample, including：
Feature Dimension Reduction matrix is obtained by dimension-reduction algorithm；
By the Feature Dimension Reduction matrix, dimension-reduction treatment is carried out to described each sample primitive character, the low dimensional characteristic vector of each sample is obtained.
Preferably, the sport career age estimates that model enters row constraint using penalty；
The penalty is：
Wherein,x(i)For i-th sample multilayer neural network last hidden layer of the output layer output, y(i)For the actual age of i-th sample, m is training sample number, and k is age categories number, wijI-th sample age is represented for the credibility of j year, when | y(i)- j | during≤K,In the case of other, wij=0.
On the other hand, the invention provides a kind of estimation of Age device based on image, including：
Information acquisition unit, for obtaining facial image to be detected and correspondence key point；Corner location of the key point for main portions；
Feature extraction unit, for according to the facial image to be detected and correspondence key point, extracting face primitive character to be detected；
Dimensionality reduction unit, for the face primitive character to be detected is carried out dimension-reduction treatment, obtains face low dimensional characteristic vector to be detected；
Estimation of Age unit, for carrying out estimation of Age by estimation of Age model to the face low dimensional characteristic vector to be detected.
Preferably, the feature extraction unit is specifically included：
Local feature obtains subelement, for according to the facial image to be detected and correspondence key point, obtaining face local feature to be detected；
Primitive character obtains subelement, for the face local feature to be detected is connected in series, obtains face primitive character to be detected；
The dimensionality reduction unit, is additionally operable to obtain Feature Dimension Reduction matrix and the face primitive character to be detected；By the Feature Dimension Reduction matrix, dimension-reduction treatment is carried out to the face primitive character to be detected, obtain face low dimensional characteristic vector to be detected.
Preferably, the device also includes：
Sample information acquiring unit, for obtaining the age of the set of facial image sample information and correspondence image sample；
Position determination unit, for determining the key point of each sample in the facial image sample information set；Corner location of the key point for main portions；
Sample primitive character acquiring unit, for according to each sample and its key point in the facial image sample information set, obtaining the primitive character of each sample in the facial image sample information set；
Sample dimensionality reduction unit, for the primitive character of each sample in the facial image sample information set is carried out dimension-reduction treatment, obtains the low dimensional characteristic vector of each sample；
Model acquiring unit, estimates model for the corresponding age value sport career age of the low dimensional characteristic vector by each sample, obtains the estimation of Age model, in case used by follow-up estimation of Age.
Preferably, the sample primitive character acquiring unit obtains the local feature of each sample for the key point according to each sample in the facial image sample information set；Same sample image local feature in the facial image sample information set is connected in series, each sample primitive character is obtained.
Preferably, the sample dimensionality reduction unit, for obtaining Feature Dimension Reduction matrix by dimension-reduction algorithm；By the Feature Dimension Reduction matrix, dimension-reduction treatment is carried out to described each sample primitive character, the low dimensional characteristic vector of each sample is obtained.
Another further aspect, the invention provides a kind of estimation of Age system based on image, including：As above the estimation of Age device based on image described in any one.
Face representation method of the present invention by dimension-reduction treatment primitive character, and estimation of Age model, estimation of Age is allowd to be less than 5.5 averaged power spectrum error in the application of dynamic scene, accuracy rate of the absolute error less than 10 years old, in addition, the training of the estimation of Age model requires no knowledge about the identity information of sample, greatly facilitates the acquisition of training sample.
Description of the drawings
Fig. 1 is a kind of age estimation method flow chart based on image provided in an embodiment of the present invention；
Fig. 2 is a kind of estimation of Age apparatus structure schematic diagram based on image provided in an embodiment of the present invention；
Fig. 3 is a kind of estimation of Age system structure diagram based on image provided in an embodiment of the present invention；
Fig. 4 is the training flow chart of estimation of Age model in a kind of age estimation method based on image provided in an embodiment of the present invention.
Specific embodiment
For making the object, technical solutions and advantages of the present invention clearer, below in conjunction with accompanying drawing, embodiment of the present invention is described further in detail.
If Fig. 1 is to show a kind of age estimation method based on image provided in an embodiment of the present invention；The method includes：
101：Obtain facial image to be detected and correspondence key point；Corner location of the key point for main portions；
102：According to the facial image to be detected and correspondence key point, face primitive character to be detected is extracted；
103：The face primitive character to be detected is carried out into dimension-reduction treatment, face low dimensional characteristic vector to be detected is obtained；
104：Estimation of Age is carried out to the face low dimensional characteristic vector to be detected by estimation of Age model.
It should be noted that it is described according to the facial image to be detected and correspondence key point, face primitive character step to be detected is extracted, is specifically included：
According to the facial image to be detected and correspondence key point, face local feature to be detected is obtained；
The face local feature to be detected is connected in series, face primitive character to be detected is obtained；
It is described that the face primitive character to be detected is carried out into dimension-reduction treatment, face low dimensional characteristic vector step to be detected is obtained, is specifically included：
Obtain Feature Dimension Reduction matrix and the face primitive character to be detected；
By the Feature Dimension Reduction matrix, dimension-reduction treatment is carried out to the face primitive character to be detected, obtain face low dimensional characteristic vector to be detected.
Based on above example, a kind of training flow chart of estimation of Age model in age estimation method based on image is illustrated in figure 4；The training flow process is as follows：
401：Obtain the age of the set of facial image sample information and correspondence image sample；
402：Determine the key point of each sample in the facial image sample information set；Corner location of the key point for main portions；It should be noted that the key point can also be directly obtained；
403：According to each sample and its key point in the facial image sample information set, the primitive character of each sample in the facial image sample information set is obtained；
404：The primitive character of each sample in the facial image sample information set is carried out into dimension-reduction treatment, the low dimensional characteristic vector of each sample is obtained；
405：Model is estimated by the corresponding age value sport career age of the low dimensional characteristic vector of each sample, the estimation of Age model is obtained, in case used by follow-up estimation of Age.
Preferably, described according to each sample and its key point in the facial image sample information set, the primitive character step for obtaining each sample in the facial image sample information set includes：
According to the key point of each sample in the facial image sample information set, the local feature of each sample is obtained；
Same sample image local feature in the facial image sample information set is connected in series, each sample primitive character is obtained.
Preferably, the primitive character by each sample in the facial image sample information set carries out dimension-reduction treatment, obtains the low dimensional characteristic vector step of each sample, including：
Feature Dimension Reduction matrix is obtained by dimension-reduction algorithm；
By the Feature Dimension Reduction matrix, dimension-reduction treatment is carried out to described each sample primitive character, the low dimensional characteristic vector of each sample is obtained.
Preferably, the sport career age estimates that model enters row constraint using penalty；
The penalty is：
Wherein,x(i)For i-th sample multilayer neural network last hidden layer of the output layer output, y(i)For the actual age of i-th sample, m is training sample number, and k is age categories number, wijI-th sample age is represented for the credibility of j year, when | y(i)- j | during≤K,In the case of other, wij=0.
Based on above example, as shown in Fig. 2 for a kind of estimation of Age apparatus structure schematic diagram based on image provided in an embodiment of the present invention；The device includes：
Information acquisition unit 201, for obtaining facial image to be detected and correspondence key point；Corner location of the key point for main portions；
Feature extraction unit 202, for according to the facial image to be detected and correspondence key point, extracting face primitive character to be detected；
Dimension-reduction treatment unit 203, for the face primitive character to be detected is carried out dimension-reduction treatment, obtains face low dimensional characteristic vector to be detected；
Estimation of Age unit 204, for carrying out estimation of Age by estimation of Age model to the face low dimensional characteristic vector to be detected.
Preferably, the feature extraction unit is specifically included：
Local feature obtains subelement, for according to the facial image to be detected and correspondence key point, obtaining face local feature to be detected；
Primitive character obtains subelement, for the face local feature to be detected is connected in series, obtains face primitive character to be detected；
The dimensionality reduction unit, is additionally operable to obtain Feature Dimension Reduction matrix and the face primitive character to be detected；By the Feature Dimension Reduction matrix, dimension-reduction treatment is carried out to the face primitive character to be detected, obtain face low dimensional characteristic vector to be detected.
Preferably, the device also includes：
Sample information acquiring unit, for obtaining the age of the set of facial image sample information and correspondence image sample；
Position determination unit, for determining the key point of each sample in the facial image sample information set；Corner location of the key point for main portions；
Sample primitive character acquiring unit, for according to each sample and its key point in the facial image sample information set, obtaining the primitive character of each sample in the facial image sample information set；
Sample dimensionality reduction unit, for the primitive character of each sample in the facial image sample information set is carried out dimension-reduction treatment, obtains the low dimensional characteristic vector of each sample；
Model acquiring unit, estimates model for the corresponding age value sport career age of the low dimensional characteristic vector by each sample, obtains the estimation of Age model, in case used by follow-up estimation of Age.
Preferably, the sample primitive character acquiring unit obtains the local feature of each sample for the key point according to each sample in the facial image sample information set；Same sample image local feature in the facial image sample information set is connected in series, each sample primitive character is obtained.
Preferably, the sample dimensionality reduction unit, for obtaining Feature Dimension Reduction matrix by dimension-reduction algorithm；By the Feature Dimension Reduction matrix, dimension-reduction treatment is carried out to described each sample primitive character, the low dimensional characteristic vector of each sample is obtained.
Based on above example, below the training principle and estimation of Age principle of estimation of Age model are described in detail.
It is as follows that the training principle of the estimation of Age model implements process：
The first step obtains the age of the set of facial image sample information and correspondence image sample；
Second step determines the key point of each sample in the facial image sample information set；The key point can also be directly obtained by the first step simultaneously；
3rd step calculates local feature according to each sample and its key point in the facial image sample information set；The same sample image local feature is connected in series, primitive character is obtained.The primitive character is complete face representation.
Because no longer doing normalization to face on the whole, caused by the factors such as attitude, expression and shape of face difference, problem effectively can be avoided, meanwhile, local feature is extracted in the local field of key point and can also obtain finer and complete face representation, be beneficial to final estimation of Age.The primitive character description can be using HOG, LBP, Gabor etc., and the size of each key point local field, gridding parameter etc. can be freely set according to practical situation.
The primitive character is carried out dimension-reduction treatment by the 4th step；The Feature Dimension Reduction is processed, and the dimension reduction method of employing can be using PCA, LDA etc..The usual dimension of face primitive character that 3rd step is obtained is very high, and dimension-reduction treatment can reduce characteristic dimension, and play Noise Reduction, be beneficial to subsequent treatment.Below by way of illustrating to dimension-reduction treatment process as a example by PCA dimension reduction methods：Specifically in two steps：1, by PCA algorithms obtain Feature Dimension Reduction matrix, Feature Dimension Reduction matrix described herein be dimensionality reduction model, the dimensionality reduction model only in the training flow process of estimation of Age model obtain；2. by the Feature Dimension Reduction matrix, dimension-reduction treatment is carried out to the primitive character, obtain low dimensional characteristic vector.
5th step estimates model by the corresponding age value sport career age of the low dimensional characteristic vector of each sample, obtains the estimation of Age model, in case used by follow-up estimation of Age.
It should be noted that employing penalty during the estimation of Age model enters row constraint；The penalty is as follows：
Wherein,x(i)For i-th sample multilayer neural network last hidden layer of the output layer output, y(i)For the actual age of i-th sample, m is training sample number, and k is age categories number, wijI-th sample age is represented for the credibility of j year, when | y(i)- j | during≤K,In the case of other, wij=0.
The estimation of Age process of the age estimation method based on image is specific as follows：
The first step obtains facial image to be detected and correspondence key point；Corner location of the key point for main portions；For example：The corner location of the organs such as eye, nose, mouth.
Second step extracts face primitive character to be detected according to the detection facial image and correspondence key point；The step, specifically includes：
According to the facial image to be detected and correspondence key point, face local feature to be detected is obtained；
The face local feature to be detected is connected in series, face primitive character to be detected is obtained；
The face primitive character to be detected is carried out dimension-reduction treatment by the 3rd step, obtains face low dimensional characteristic vector to be detected；The step, specifically includes：
Obtain Feature Dimension Reduction matrix and the face primitive character to be detected；
By the Feature Dimension Reduction matrix, dimension-reduction treatment is carried out to the face primitive character to be detected, obtain face low dimensional characteristic vector to be detected.
4th step carries out estimation of Age by estimation of Age model to the face low dimensional characteristic vector to be detected.
It should be noted that the estimation of Age model is multilayer neural network；The face character information to be detected is belonged to the probability output of each age value by the multilayer neural network according to current sample；Using the probability output maximum age value as the face to be detected final estimation of Age value.
It should be noted that, the purpose of the present invention is to estimate the age based on facial image, and age estimation is complex concept the problems such as be related to physiology, sociology and psychology, not equivalent to physiological age, its classification is fuzzy, and it is also unbalanced to produce wrong estimated risk, and the child of 10 years old is misestimated the impression for bringing user for 15 years old and 50 years old huge difference, so estimation of Age can not be regarded as classification problem more than simply.For these situations, the present invention devises a kind of multilayer neural network with Softmax Regression as output layer, directly differentiate that sample belongs to the probability of all ages and classes value, and the lack of uniformity for the wrong calculated risk of the training process of estimation of Age model devises special penalty.The penalty is
x(i)For i-th sample last hidden layer of network output, y(i)For the actual age of i-th sample, m is training sample number, and k is age categories number, wijI-th sample age is represented for the credibility of j year, when | y(i)- j | during≤K,In the case of other, wij=0.
Network training adopts back-propagation gradient descent algorithm.
As shown in figure 3, for a kind of estimation of Age system structure diagram based on image provided in an embodiment of the present invention；The system includes：As above arbitrary estimation of Age device based on image.
The present invention is by the primitive character of dimension-reduction treatment, fine and complete face representation method, and for the grader of the special design of estimation of Age problem, establish an accurate estimation of Age system, the system can be less than 5.5 averaged power spectrum error, rate of accuracy reached of the absolute error less than 10 years old to 88% in the application of dynamic scene.Additionally, the training of the grader requires no knowledge about the identity information of sample, the acquisition of training sample is greatly facilitated.
Presently preferred embodiments of the present invention is the foregoing is only, protection scope of the present invention is not intended to limit.All any modification, equivalent substitution and improvements made within the spirit and principles in the present invention etc., are all contained in protection scope of the present invention.

Claims (12)

1. a kind of age estimation method based on image, it is characterised in that include：
Obtain facial image to be detected and correspondence key point；Corner location of the key point for main portions；
According to the facial image to be detected and correspondence key point, face primitive character to be detected is extracted；
The face primitive character to be detected is carried out into dimension-reduction treatment, face low dimensional feature to be detected is obtained Vector；
Estimation of Age is carried out to the face low dimensional characteristic vector to be detected by estimation of Age model.
2. age estimation method according to claim 1 based on image, it is characterised in that
It is described according to the facial image to be detected and correspondence key point, extract face primitive character to be detected Step, specifically includes：
According to the facial image to be detected and correspondence key point, face local feature to be detected is obtained；
The face local feature to be detected is connected in series, face primitive character to be detected is obtained；
It is described that the face primitive character to be detected is carried out into dimension-reduction treatment, obtain face low dimensional to be detected Characteristic vector step, specifically includes：
Obtain Feature Dimension Reduction matrix and the face primitive character to be detected；
By the Feature Dimension Reduction matrix, dimension-reduction treatment is carried out to the face primitive character to be detected, is obtained Take face low dimensional characteristic vector to be detected.
3. the age estimation method based on image according to claim 1 or claim 2, it is characterised in that should Method also includes：
Obtain the age of the set of facial image sample information and correspondence image sample；
Determine the key point of each sample in the facial image sample information set；Based on the key point Want the corner location at position；
According to each sample and its key point in the facial image sample information set, the face is obtained The primitive character of each sample in image pattern information aggregate；
The primitive character of each sample in the facial image sample information set is carried out into dimension-reduction treatment, is obtained Take the low dimensional characteristic vector of each sample；
Mould is estimated by the corresponding age value sport career age of the low dimensional characteristic vector of each sample Type, obtains the estimation of Age model, in case used by follow-up estimation of Age.
4. the age estimation method based on image according to claim 3, it is characterised in that described According to each sample and its key point in the facial image sample information set, the facial image is obtained In sample information set, the primitive character step of each sample includes：
According to the key point of each sample in the facial image sample information set, each sample is obtained Local feature；
Same sample image local feature in the facial image sample information set is connected in series, is obtained Each sample primitive character.
5. the age estimation method based on image according to claim 4, it is characterised in that described The primitive character of each sample in the facial image sample information set is carried out into dimension-reduction treatment, obtains every The low dimensional characteristic vector step of individual sample, including：
Feature Dimension Reduction matrix is obtained by dimension-reduction algorithm；
By the Feature Dimension Reduction matrix, dimension-reduction treatment is carried out to described each sample primitive character, obtained The low dimensional characteristic vector of each sample.
6. the age estimation method based on image according to claim 5, it is characterised in that described Sport career age estimates that model enters row constraint using penalty；
The penalty is：
J ( θ ) = - 1 m [ Σ i = 1 m Σ j = 1 k w i j n o r m log e θ j T x ( i ) Σ l = 1 k e θ l T x ( i ) ]
Wherein,x(i)It is last in the multilayer neural network of the output layer for i-th sample The output of one hidden layer, y(i)For the actual age of i-th sample, m is training sample number, and k is year Age class number, wijI-th sample age is represented for the credibility of j year, when | y(i)- j | during≤K,In the case of other, wij=0.
7. a kind of estimation of Age device based on image, it is characterised in that include：
Information acquisition unit, for obtaining facial image to be detected and correspondence key point；The key point is The corner location of main portions；
Feature extraction unit, for according to the facial image to be detected and correspondence key point, extracting to be checked Survey face primitive character；
Dimensionality reduction unit, for the face primitive character to be detected is carried out dimension-reduction treatment, obtains to be detected Face low dimensional characteristic vector；
Estimation of Age unit, for by estimation of Age model to the face low dimensional feature to be detected to Amount carries out estimation of Age.
8. estimation of Age device according to claim 7 based on image, it is characterised in that
The feature extraction unit is specifically included：
Local feature obtains subelement, for according to the facial image to be detected and correspondence key point, obtaining Take face local feature to be detected；
Primitive character obtains subelement, for the face local feature to be detected is connected in series, obtains Face primitive character to be detected；
The dimensionality reduction unit, is additionally operable to obtain Feature Dimension Reduction matrix and the face primitive character to be detected； By the Feature Dimension Reduction matrix, dimension-reduction treatment is carried out to the face primitive character to be detected, acquisition is treated Detection face low dimensional characteristic vector.
9. the estimation of Age device according to claim 7 or 8 based on image, it is characterised in that should Device also includes：
Sample information acquiring unit, for obtaining the set of facial image sample information and correspondence image sample Age；
Position determination unit, for determining the key of each sample in the facial image sample information set Point；Corner location of the key point for main portions；
Sample primitive character acquiring unit, for according to each sample in the facial image sample information set Sheet and its key point, obtain the primitive character of each sample in the facial image sample information set；
Sample dimensionality reduction unit, for by the original spy of each sample in the facial image sample information set Levying carries out dimension-reduction treatment, obtains the low dimensional characteristic vector of each sample；
Model acquiring unit, for the corresponding year of the low dimensional characteristic vector by each sample Age value sport career age estimates model, obtains the estimation of Age model, in case used by follow-up estimation of Age.
10. the estimation of Age device based on image according to claim 9, it is characterised in that institute Sample primitive character acquiring unit is stated for according to each sample in the facial image sample information set Key point, obtains the local feature of each sample；By same in the facial image sample information set This image local feature is connected in series, and obtains each sample primitive character.
The 11. estimation of Age devices based on image according to claim 10, it is characterised in that institute Sample dimensionality reduction unit is stated, for Feature Dimension Reduction matrix being obtained by dimension-reduction algorithm；By the Feature Dimension Reduction Matrix, carries out dimension-reduction treatment to described each sample primitive character, obtains the low dimensional feature of each sample Vector.
12. a kind of estimation of Age systems based on image, it is characterised in that include：Such as claim 7 Estimation of Age device into 11 described in any one based on image.
CN201510586848.1A 2015-09-15 2015-09-15 Age estimating method, age estimating device and age estimating system based on image Pending CN106529377A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510586848.1A CN106529377A (en) 2015-09-15 2015-09-15 Age estimating method, age estimating device and age estimating system based on image

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510586848.1A CN106529377A (en) 2015-09-15 2015-09-15 Age estimating method, age estimating device and age estimating system based on image

Publications (1)

Publication Number Publication Date
CN106529377A true CN106529377A (en) 2017-03-22

Family

ID=58349228

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510586848.1A Pending CN106529377A (en) 2015-09-15 2015-09-15 Age estimating method, age estimating device and age estimating system based on image

Country Status (1)

Country Link
CN (1) CN106529377A (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107180234A (en) * 2017-06-01 2017-09-19 四川新网银行股份有限公司 The credit risk forecast method extracted based on expression recognition and face characteristic
CN107403145A (en) * 2017-07-14 2017-11-28 北京小米移动软件有限公司 Image characteristic points positioning method and device
CN107545249A (en) * 2017-08-30 2018-01-05 国信优易数据有限公司 A kind of population ages' recognition methods and device
CN110378306A (en) * 2019-07-25 2019-10-25 厦门美图之家科技有限公司 Age prediction technique, device and image processing equipment
CN110674744A (en) * 2019-09-24 2020-01-10 京东方科技集团股份有限公司 Age identification method and device and electronic equipment
CN111062995A (en) * 2019-11-28 2020-04-24 重庆中星微人工智能芯片技术有限公司 Method and device for generating face image, electronic equipment and computer readable medium
CN111967382A (en) * 2020-08-14 2020-11-20 北京金山云网络技术有限公司 Age estimation method, and training method and device of age estimation model
TWI832660B (en) * 2023-01-06 2024-02-11 宏碁股份有限公司 Method and device for calculating normative distribution of visual acuity index

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101615248A (en) * 2009-04-21 2009-12-30 华为技术有限公司 Age estimation method, equipment and face identification system
CN102637251A (en) * 2012-03-20 2012-08-15 华中科技大学 Face recognition method based on reference features
CN102663413A (en) * 2012-03-09 2012-09-12 中盾信安科技(江苏)有限公司 Multi-gesture and cross-age oriented face image authentication method
CN103544486A (en) * 2013-10-31 2014-01-29 东南大学 Human age estimation method based on self-adaptation sign distribution
CN103927518A (en) * 2014-04-14 2014-07-16 中国华戎控股有限公司 Facial feature extraction method for facial analysis system
CN104021384A (en) * 2014-06-30 2014-09-03 深圳市创冠智能网络技术有限公司 Face recognition method and device
CN104463190A (en) * 2014-10-30 2015-03-25 华为技术有限公司 Age estimation method and equipment
CN104504376A (en) * 2014-12-22 2015-04-08 厦门美图之家科技有限公司 Age classification method and system for face images

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101615248A (en) * 2009-04-21 2009-12-30 华为技术有限公司 Age estimation method, equipment and face identification system
CN102663413A (en) * 2012-03-09 2012-09-12 中盾信安科技(江苏)有限公司 Multi-gesture and cross-age oriented face image authentication method
CN102637251A (en) * 2012-03-20 2012-08-15 华中科技大学 Face recognition method based on reference features
CN103544486A (en) * 2013-10-31 2014-01-29 东南大学 Human age estimation method based on self-adaptation sign distribution
CN103927518A (en) * 2014-04-14 2014-07-16 中国华戎控股有限公司 Facial feature extraction method for facial analysis system
CN104021384A (en) * 2014-06-30 2014-09-03 深圳市创冠智能网络技术有限公司 Face recognition method and device
CN104463190A (en) * 2014-10-30 2015-03-25 华为技术有限公司 Age estimation method and equipment
CN104504376A (en) * 2014-12-22 2015-04-08 厦门美图之家科技有限公司 Age classification method and system for face images

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
李红梅: "基于人脸图像的性别识别的研究与实现", 《中国优秀硕士学位论文全文数据库 信息科技辑》 *

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107180234A (en) * 2017-06-01 2017-09-19 四川新网银行股份有限公司 The credit risk forecast method extracted based on expression recognition and face characteristic
CN107403145A (en) * 2017-07-14 2017-11-28 北京小米移动软件有限公司 Image characteristic points positioning method and device
CN107403145B (en) * 2017-07-14 2021-03-09 北京小米移动软件有限公司 Image feature point positioning method and device
CN107545249A (en) * 2017-08-30 2018-01-05 国信优易数据有限公司 A kind of population ages' recognition methods and device
CN110378306A (en) * 2019-07-25 2019-10-25 厦门美图之家科技有限公司 Age prediction technique, device and image processing equipment
CN110674744A (en) * 2019-09-24 2020-01-10 京东方科技集团股份有限公司 Age identification method and device and electronic equipment
CN111062995A (en) * 2019-11-28 2020-04-24 重庆中星微人工智能芯片技术有限公司 Method and device for generating face image, electronic equipment and computer readable medium
CN111062995B (en) * 2019-11-28 2024-02-23 重庆中星微人工智能芯片技术有限公司 Method, apparatus, electronic device and computer readable medium for generating face image
CN111967382A (en) * 2020-08-14 2020-11-20 北京金山云网络技术有限公司 Age estimation method, and training method and device of age estimation model
TWI832660B (en) * 2023-01-06 2024-02-11 宏碁股份有限公司 Method and device for calculating normative distribution of visual acuity index

Similar Documents

Publication Publication Date Title
Yuan et al. Fingerprint liveness detection using an improved CNN with image scale equalization
CN106529377A (en) Age estimating method, age estimating device and age estimating system based on image
CN110443798B (en) Autism detection method, device and system based on magnetic resonance image
CN103942577B (en) Based on the personal identification method for establishing sample database and composite character certainly in video monitoring
CN106529468B (en) A kind of finger vein identification method and system based on convolutional neural networks
CN106529442B (en) A kind of pedestrian recognition method and device
CN104143079B (en) The method and system of face character identification
CN110084149B (en) Face verification method based on hard sample quadruple dynamic boundary loss function
CN106446754A (en) Image identification method, metric learning method, image source identification method and devices
CN111931758B (en) Face recognition method and device combining facial veins
CN108229330A (en) Face fusion recognition methods and device, electronic equipment and storage medium
CN105335732B (en) Based on piecemeal and differentiate that Non-negative Matrix Factorization blocks face identification method
CN106529504B (en) A kind of bimodal video feeling recognition methods of compound space-time characteristic
CN105956570B (en) Smiling face's recognition methods based on lip feature and deep learning
CN104298981A (en) Face microexpression recognition method
CN111222380B (en) Living body detection method and device and recognition model training method thereof
CN106897669A (en) A kind of pedestrian based on consistent iteration various visual angles transfer learning discrimination method again
CN103839033A (en) Face identification method based on fuzzy rule
CN110458792A (en) Method and device for evaluating quality of face image
CN105320948A (en) Image based gender identification method, apparatus and system
CN104679967B (en) A kind of method for judging psychological test reliability
CN110427923A (en) Infantile vomiting milk Activity recognition method, apparatus, computer equipment and storage medium
CN103714340A (en) Self-adaptation feature extracting method based on image partitioning
CN106778491A (en) The acquisition methods and equipment of face 3D characteristic informations
CN113506274A (en) Detection system for human cognitive condition based on visual saliency difference map

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
CB02 Change of applicant information
CB02 Change of applicant information

Address after: Unit 801, Unit 5, No. 2 Building, 9 Yuan, Fenghao East Road, Haidian District, Beijing 100094

Applicant after: Wen'an Beijing intelligent technology Limited by Share Ltd

Address before: 100085 4th Floor, No. 1 Courtyard, Shangdi East Road, Haidian District, Beijing

Applicant before: Wen'an Beijing intelligent technology Limited by Share Ltd

RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20170322