CN104036261B - Face identification method and system - Google Patents
Face identification method and system Download PDFInfo
- Publication number
- CN104036261B CN104036261B CN201410306005.7A CN201410306005A CN104036261B CN 104036261 B CN104036261 B CN 104036261B CN 201410306005 A CN201410306005 A CN 201410306005A CN 104036261 B CN104036261 B CN 104036261B
- Authority
- CN
- China
- Prior art keywords
- face picture
- classification
- layer
- target face
- target
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000004364 calculation method Methods 0.000 abstract description 6
- 238000005516 engineering process Methods 0.000 abstract description 4
- 238000000605 extraction Methods 0.000 description 5
- 241001269238 Data Species 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000010276 construction Methods 0.000 description 3
- 238000005192 partition Methods 0.000 description 3
- 230000015572 biosynthetic process Effects 0.000 description 2
- 238000004590 computer program Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000001035 drying Methods 0.000 description 1
- 230000001815 facial effect Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
Abstract
The present invention is related to areas of information technology with regard to a kind of face identification method and system, and main purpose is, compared to prior art, can to complete recognition of face by less operand.Method includes:The face picture collected is carried out into cluster and obtains multiple 1st layer of classification, and the face picture in continuing to classify at least one i-th layers by iterative manner carries out cluster and obtains multiple i+1 layers classification;The 1st layer of classification belonging to target face picture is identified, and continues+1 layer of classification of jth belonging to target face picture is identified in the jth layer classification belonging to target face picture by iterative manner;By iterative manner until when there is no+1 layer of classification of jth in the jth layer classification belonging to target face picture, from the jth layer classification belonging to target face picture, identifying the similar face picture of target face picture.Relative to existing technical scheme, the amount of calculation in technical scheme is very little, substantially increases recognition of face efficiency.
Description
Technical field
The present invention relates to areas of information technology, in particular to a kind of face identification method and system.
Background technology
At present, enter line retrieval to extensive face, need the work for carrying out to be feature extraction to be carried out to human face data first
And quantization, after these work, every face can obtain corresponding multi-dimensional feature data;And by comparing two face figures
Whether the characteristic of picture is similar to judge two facial images.And in large-scale human face data, finding similar face needs
Every image calculated, find nearest human face data, need great amount of calculation.
Existing scheme directly to cluster to magnanimity face picture in local library, by target face picture successively and each
Cluster is compared and cluster belonging to which with being found, then compares to find similar face figure in affiliated cluster with every face picture
Piece.This scheme to a certain degree accelerates face search procedure, but in the search of extensive face, this scheme is still needed to be carried out
Great amount of calculation:If cluster arranges less, in each cluster, data scale can be scanned for consuming in cluster than larger
Take a large amount of operation times;If partition clustering is excessive, finds affiliated cluster and be accomplished by, compared with intensive, cannot all meeting search and drawing
The requirement of real-time held up.
The content of the invention
In view of the above problems, it is proposed that the present invention so as to provide one kind overcome the problems referred to above or at least in part solve on
State the face identification method and system of problem.
According to one aspect of the present invention, there is provided a kind of face identification method, which includes:By the face picture collected
Carry out cluster and obtain multiple 1st layer of classification, and continue to enter the face picture at least one i-th layers of classification by iterative manner
Row cluster obtains multiple i+1 layer classification, and i carries out integer value backward from 1;Identify the 1st layer belonging to target face picture
Classification, and continue to identify the target face in the jth layer classification belonging to the target face picture by iterative manner
+ 1 layer of classification of jth belonging to picture, j sequentially carry out integer value backward from 1;By the iterative manner until in the target
When there is no+1 layer of classification of jth in the jth layer classification belonging to face picture, from the jth layer belonging to the target face picture point
In class, the similar face picture of the target face picture is identified.
Alternatively, aforesaid face identification method, wherein, it is described the face picture collected is carried out into cluster to obtain multiple
The step of 1st layer of classification, includes:According to the feature of the face picture collected, generate described in the face picture collected
Characteristic vector;Multiple initial center points are set, and the characteristic vector according to the face picture collected is described just with each
The face picture collected is divided into multiple 1st layer of classification by the distance of beginning central point, and according to each 1st layer point
The characteristic vector of the face picture of class, calculates the vector center point of each the 1st layer classification.
Alternatively, aforesaid face identification method, wherein, it is described the face picture collected is carried out into cluster to obtain multiple
The step of 1st layer of classification, also includes:Side between the initial center point and vector center point of each the 1st layer classification described in calculating
Difference;Size such as the variance exceedes predetermined threshold value, then reset initial center point, and again by the face collected
Picture is divided into multiple 1st layer of classification, and recalculates the vector center point of each the 1st layer classification.
Alternatively, aforesaid face identification method, wherein, the 1st layer of classification identified belonging to target face picture
The step of include:According to the feature of the target face picture, the characteristic vector of the target face picture is generated;Select vector
The 1st layer of minimum classification of distance between the characteristic vector of central point and the target face picture, as the target face figure
The 1st layer of classification belonging to piece.
Alternatively, aforesaid face identification method, wherein, the similar face figure for identifying the target face picture
The step of piece, includes:From the face picture of jth layer classification, the feature of characteristic vector and the target face picture is selected
At least one minimum face picture of distance between vector, as the described similar face picture of the target face picture.
According to another aspect of the present invention, a kind of face identification system is additionally provided, which includes:Sort module, for inciting somebody to action
The face picture collected carries out cluster and obtains multiple 1st layer of classification, and is continued at least one i-th layers points by iterative manner
Face picture in class carries out cluster and obtains multiple i+1 layer classification, and i sequentially carries out integer value backward from 1;Classification iteration is known
Other module, for identifying the 1st layer of classification belonging to target face picture, and is continued in the target face by iterative manner
+ 1 layer of classification of jth belonging to the target face picture is identified in jth layer classification belonging to picture, j carries out integer backward from 1
Value;Similar face picture identification module, for there is no jth+1 in the jth layer classification belonging to the target face picture
During layer classification, from the jth layer classification belonging to the target face picture, the similar face of the target face picture is identified
Picture.
Alternatively, aforesaid face identification system, wherein, also include:First eigenvector generation module, for according to institute
State the feature of the face picture collected, generate described in the characteristic vector of face picture collected;The sort module is arranged
Multiple initial center points, and the distance of the characteristic vector according to the face picture collected and each initial center point
The face picture collected is divided into multiple 1st layer of classification by distance, and according to each the 1st layer face picture classified
Characteristic vector, calculates the vector center point of each the 1st layer classification.
Alternatively, aforesaid face identification system, wherein, also include:Variance computing module, calculates described each the 1st layer
Variance between the initial center point and vector center point of classification;As the variance size exceed predetermined threshold value, then described point
Generic module resets initial center point, and the face picture collected is divided into multiple 1st layers of classification again, and again
The vector center point of each the 1st layer classification described in calculating.
Alternatively, aforesaid face identification system, wherein, also include:Second feature vector generation module, for according to institute
The feature of target face picture is stated, the characteristic vector of the target face picture is generated;The sort module selects vector center
The 1st layer of minimum classification of distance between point and the characteristic vector of the target face picture, as the target face picture institute
1st layer of classification of category.
Alternatively, aforesaid face identification system, wherein, the similar face picture identification module is from the jth layer point
In the face picture of class, at least one of distance minimum between characteristic vector and the characteristic vector of the target face picture is selected
Face picture, as the described similar face picture of the target face picture.
Face identification method of the invention and system, by being divided poly- again to clustering last layer cluster result
Class, it is sandwich construction to cluster the face picture collected by iterative manner, and successively finds target by iterative manner
The affiliated classification of face picture, until eventually finding the similar face picture of target face picture;Relative to existing technical side
Case, the amount of calculation in technical scheme are very little, substantially increase recognition of face efficiency.
Described above is only the general introduction of technical solution of the present invention, in order to better understand the technological means of the present invention,
And can be practiced according to the content of specification, and in order to allow the above and other objects of the present invention, feature and advantage can
Become apparent, below especially exemplified by the specific embodiment of the present invention.
Description of the drawings
By the detailed description for reading hereafter preferred embodiment, various other advantages and benefit are common for this area
Technical staff will be clear from understanding.Accompanying drawing is only used for the purpose for illustrating preferred embodiment, and is not considered as to the present invention
Restriction.And in whole accompanying drawing, it is denoted by the same reference numerals identical part.In the accompanying drawings:
The flow chart that Fig. 1 shows face identification method according to an embodiment of the invention;
Fig. 2 shows the operating diagram of face identification method according to an embodiment of the invention;
The flow chart that Fig. 3 shows face identification method according to an embodiment of the invention;
The flow chart that Fig. 4 shows face identification method according to an embodiment of the invention;
The flow chart that Fig. 5 shows face identification method according to an embodiment of the invention;
Fig. 6 shows the block diagram of face identification system according to an embodiment of the invention;
Fig. 7 shows the block diagram of face identification method according to an embodiment of the invention;
Fig. 8 shows the block diagram of face identification method according to an embodiment of the invention.
Specific embodiment
The exemplary embodiment of the disclosure is more fully described below with reference to accompanying drawings.Although the disclosure is shown in accompanying drawing
Exemplary embodiment, it being understood, however, that may be realized in various forms the disclosure and should not be by embodiments set forth here
Limited.On the contrary, there is provided these embodiments are able to be best understood from the disclosure, and can be by the scope of the present disclosure
Complete conveys to those skilled in the art.
As illustrated, An embodiment provides a kind of face identification method, which includes:
Step 110, by the face picture collected carry out cluster obtain it is multiple 1st layers classification, and by iterative manner after
The continuous face picture at least one i-th layers of classification carries out cluster and obtains multiple i+1 layer classification, and i carries out integer backward from 1
Value.In the present embodiment, as shown in Fig. 2 for example, wherein C1 classification includes C11 ..., and C1m etc. is more for the multistratum classification structure of formation
Individual classification, classifies including CN1, CN2 etc. in C11 classification again.
Step 120, is identified the 1st layer of classification belonging to target face picture, and is continued in target person by iterative manner
+ 1 layer of classification of jth belonging to target face picture is identified in jth layer classification belonging to face picture, j sequentially carries out whole backward from 1
Number value.
Step 130, by iterative manner until there is no+1 layer of jth in the jth layer classification belonging to target face picture
During classification, from the jth layer classification belonging to target face picture, the similar face picture of target face picture is identified.
In the technical scheme of the present embodiment, pass through iteration by partition clustering again being carried out to cluster last layer cluster result
Mode by the face picture collected cluster for sandwich construction, and the institute that target face picture is successively found by iterative manner
Category classification, until eventually finding the similar face picture of target face picture;Relative to existing technical scheme, the skill of the present invention
Amount of calculation in art scheme is very little, substantially increases recognition of face efficiency.
As shown in figure 3, an alternative embodiment of the invention provides a kind of face identification method, wherein, step 110 is wrapped
Include:
Step 111, according to the feature of the face picture collected, generates the characteristic vector of the face picture collected.This
Extraction of the embodiment based on the face picture feature to having collected, the feature for having collected face picture can be extracted in advance and be stored
In feature database.
Step 112, arranges multiple initial center points, and the characteristic vector according to the face picture collected is initial with each
The face picture collected is divided into multiple 1st layer of classification by the distance of central point, and according to the people of each the 1st layer classification
The characteristic vector of face picture, calculates the vector center point of each the 1st layer classification.
The then technical scheme according to the present embodiment, all according to the face picture characteristic vector collected and initial center point it
Between distance, assign it in closest classification, afterwards calculate vector center point;Multilayer can be quickly formed so repeatedly
Cluster structure.
An alternative embodiment of the invention provides a kind of face identification method, wherein, step 110 also includes:
113, calculate the variance between the initial center point and vector center point of each the 1st layer classification.The present embodiment is based on
Extraction to target face picture feature, the feature of target face picture can be with extract real-time.
114, the such as size of variance exceedes predetermined threshold value, then reset initial center point, and again by the people for having collected
Face picture is divided into multiple 1st layer of classification, and recalculates the vector center point of each the 1st layer classification.
In the technical scheme of the present embodiment, if variance<0.000001 (example can take other values), then it represents that point
All it is the close face picture of feature in class, there is otherwise in presentation class feature gap significant discomfort to close and be placed in same classification
Face picture, so needing to reclassify.Now, the workflow of the face identification method of the present embodiment can with as shown in figure 4,
In figure, calculate to the step of specifying the number of plies and refer to, be the cluster structure for specifying the number of plies by the face picture collected cluster.
As shown in figure 5, an alternative embodiment of the invention provides a kind of face identification method, wherein, step 120 is wrapped
Include:
Step 121, according to the feature of target face picture, generates the characteristic vector of target face picture.
Step 122, selects 1st layer minute of distance minimum between vector center point and the characteristic vector of target face picture
Class, classifies as the 1st layer belonging to target face picture.
The then technical scheme according to the present embodiment, in each Rotating fields, all by the characteristic vector of target face picture and institute
In the classification of category upper strata, the vector center point of multiple sub-classifications is compared, and can be quickly found out belonging to target face picture most
Subclassification.
An alternative embodiment of the invention provides a kind of face identification method, wherein, step 130 includes:
From the face picture of jth layer classification, distance between characteristic vector and the characteristic vector of target face picture is selected
At least one minimum face picture, as the similar face picture of target face picture.
With reference to above example, it is assumed that it is 10,000,000 to have collected face picture, needs searched targets face picture wherein
Similar face picture when:
If the 1, using direct manner of comparison, need the characteristic vector of target face picture and the spy for collecting face picture
Levy vector and compare 10,000,000 times.
If the 2,10,000,000 data being divided into 10000 clusters, then being needed target face figure using traditional cluster mode
Piece is compared 10000 times with the vector center point of cluster, has 1000 datas during each cluster is average, then each classification is internal needs
Relatively 1000 times, totally compare 10000+k × 1000 time, number of comparisons is if k takes 10:10000+10*1000=20000
It is secondary.When the similar face picture of target face picture is found in cluster by existing nearest neighbor algorithm, k represents that selection k is near
Adjacent central point, 10 is common value.
3rd, using the technical scheme of the present embodiment, if being divided into 2 layers, 100 clusters of ground floor, the second layer are poly- for 200
Class, then have 500 datas in average each cluster of the second layer, and number of comparisons about 100+m × 200+n × 500 time such as take m=3, n
=10, number of comparisons is:100+3 × 200+10 × 500=11100 time, and 1,2 compare, and can substantially reduce number of comparisons.Together
Reason, m represent that ground floor chooses m neighbour's central point, and n represents that the second layer chooses n neighbour's central point, and 3,10 is common value.
As shown in fig. 6, another embodiment of the present invention additionally provides a kind of face identification system, which includes:
Sort module 610, obtains multiple 1st layer of classification for the face picture collected is carried out cluster, and by repeatedly
Continue to carry out the face picture at least one i-th layers of classification cluster for mode and obtain the classification of multiple i+1 layers, i from 1 backward
Integer value is carried out sequentially.In the present embodiment, as shown in Fig. 2 for example, wherein C1 classification includes the multistratum classification structure of formation
Multiple classification such as C11 ... C1m, classify including CN1, CN2 etc. in C11 classification again.
Classification iteration identification module 620, for identifying the 1st layer of classification belonging to target face picture, and passes through iteration
Mode continues+1 layer of classification of jth belonging to target face picture, j are identified in the jth layer classification belonging to target face picture
Integer value is carried out backward from 1.
Similar face picture identification module 630, for belonging to target face picture jth layer classification in do not exist jth+
During 1 layer of classification, from the jth layer classification belonging to target face picture, the similar face picture of target face picture is identified.
In the technical scheme of the present embodiment, pass through iteration by partition clustering again being carried out to cluster last layer cluster result
Mode by the face picture collected cluster for sandwich construction, and the institute that target face picture is successively found by iterative manner
Category classification, until eventually finding the similar face picture of target face picture;Relative to existing technical scheme, the skill of the present invention
Amount of calculation in art scheme is very little, substantially increases recognition of face efficiency.
As shown in fig. 7, an alternative embodiment of the invention provides a kind of face identification system, wherein, also include:
First eigenvector generation module 640, for the feature according to the face picture collected, generates the people for having collected
The characteristic vector of face picture.Extraction of the present embodiment based on the face picture feature to having collected, the spy for having collected face picture
Levy and can be extracted and stored in feature database in advance.
Sort module 610 arranges multiple initial center points, and the characteristic vector according to the face picture collected and each
The face picture collected is divided into multiple 1st layer of classification, and is classified according to each the 1st layer by the distance of initial center point
Face picture characteristic vector, calculate each the 1st layer classification vector center point.
The then technical scheme according to the present embodiment, all according to the face picture characteristic vector collected and initial center point it
Between distance, assign it in closest classification, afterwards calculate vector center point;Multilayer can be quickly formed so repeatedly
Cluster structure.
An alternative embodiment of the invention provides a kind of face identification system, wherein, also include:
Variance computing module 650, calculates the variance between the initial center point and vector center point of each the 1st layer classification.
, based on the extraction to target face picture feature, the feature of target face picture can be with extract real-time for the present embodiment.
Size such as variance exceedes predetermined threshold value, then sort module 610 resets initial center point, and has been received again
The face picture of collection is divided into multiple 1st layer of classification, and recalculates the vector center point of each the 1st layer classification.
In the technical scheme of the present embodiment, if variance<0.000001 (example can take other values), then it represents that point
All it is the close face picture of feature in class, there is otherwise in presentation class feature gap significant discomfort to close and be placed in same classification
Face picture, so needing to reclassify.Now, the workflow of the face identification method of the present embodiment can with as shown in figure 4,
In figure, calculate to the step of specifying the number of plies and refer to, be the cluster structure for specifying the number of plies by the face picture collected cluster.
As shown in figure 8, an alternative embodiment of the invention provides a kind of face identification system, wherein, also include:
Second feature vector generation module 660, for the feature according to target face picture, generates target face picture
Characteristic vector.
Sort module 610 selects distance is minimum between vector center point and the characteristic vector of target face picture the 1st layer
Classification, classifies as the 1st layer belonging to target face picture.
The then technical scheme according to the present embodiment, in each Rotating fields, all by the characteristic vector of target face picture and institute
In the classification of category upper strata, the vector center point of multiple sub-classifications is compared, and can be quickly found out belonging to target face picture most
Subclassification.
An alternative embodiment of the invention provides a kind of face identification system, wherein, similar face picture identification module
630 from the face picture of jth layer classification, selects distance minimum between characteristic vector and the characteristic vector of target face picture
At least one face picture, as the similar face picture of target face picture.
With reference to above example, it is assumed that it is 10,000,000 to have collected face picture, needs searched targets face picture wherein
Similar face picture when:
If the 1, using direct manner of comparison, need the characteristic vector of target face picture and the spy for collecting face picture
Levy vector and compare 10,000,000 times.
If the 2,10,000,000 data being divided into 10000 clusters, then being needed target face figure using traditional cluster mode
Piece is compared 10000 times with the vector center point of cluster, has 1000 datas during each cluster is average, then each classification is internal needs
Relatively 1000 times, totally compare 10000+k × 1000 time, number of comparisons is if k takes 10:10000+10*1000=20000
It is secondary.When the similar face picture of target face picture is found in cluster by existing nearest neighbor algorithm, k represents that selection k is near
Adjacent central point, 10 is common value.
3rd, using the technical scheme of the present embodiment, if being divided into 2 layers, 100 clusters of ground floor, the second layer are poly- for 200
Class, then have 500 datas in average each cluster of the second layer, and number of comparisons about 100+m × 200+n × 500 time such as take m=3, n
=10, number of comparisons is:100+3 × 200+10 × 500=11100 time, and 1,2 compare, and can substantially reduce number of comparisons.Together
Reason, m represent that ground floor chooses m neighbour's central point, and n represents that the second layer chooses n neighbour's central point, and 3,10 is common value.
Algorithm and display be not inherently related to any certain computer, virtual system or miscellaneous equipment provided herein.
Various general-purpose systems can also be used together based on teaching in this.As described above, construct required by this kind of system
Structure be obvious.Additionally, the present invention is also not for any certain programmed language.It is understood that, it is possible to use it is various
Programming language realizes the content of invention described herein, and the description done to language-specific above is to disclose this
Bright preferred forms.
In specification mentioned herein, a large amount of details are illustrated.It is to be appreciated, however, that the enforcement of the present invention
Example can be put into practice in the case where not having these details.In some instances, known method, structure is not been shown in detail
And technology, so as not to obscure the understanding of this description.
Similarly, it will be appreciated that in order to simplify the disclosure and help understand one or more in each inventive aspect, exist
Above to, in the description of the exemplary embodiment of the present invention, each feature of the present invention is grouped together into single enforcement sometimes
In example, figure or descriptions thereof.However, should the method for the disclosure be construed to reflect following intention:I.e. required guarantor
The more features of feature is expressly recited in each claim by the application claims ratio of shield.More precisely, such as following
Claims it is reflected as, inventive aspect is less than all features of single embodiment disclosed above.Therefore,
Thus the claims for following specific embodiment are expressly incorporated in the specific embodiment, wherein each claim itself
All as the separate embodiments of the present invention.
Those skilled in the art are appreciated that can be carried out adaptively to the module in the equipment in embodiment
Change and they are arranged in one or more different from embodiment equipment.Can be the module or list in embodiment
Unit or component are combined into a module or unit or component, and can be divided in addition multiple submodule or subelement or
Sub-component.In addition at least some in such feature and/or process or unit is excluded each other, can adopt any
Combine to all features disclosed in this specification (including adjoint claim, summary and accompanying drawing) and so disclosed
Where all processes or unit of method or equipment are combined.Unless expressly stated otherwise, this specification (includes adjoint power
Profit is required, summary and accompanying drawing) disclosed in each feature can it is identical by offers, be equal to or the alternative features of similar purpose carry out generation
Replace.
Although additionally, it will be appreciated by those of skill in the art that some embodiments described herein include other embodiments
In some included features rather than further feature, but the combination of the feature of different embodiments means in of the invention
Within the scope of and form different embodiments.For example, in the following claims, embodiment required for protection appoint
One of meaning can in any combination mode using.
The present invention all parts embodiment can be realized with hardware, or with one or more processor operation
Software module realize, or with combinations thereof realize.It will be understood by those of skill in the art that can use in practice
Microprocessor or digital signal processor (DSP) come realize in face identification system according to embodiments of the present invention some or
The some or all functions of person's whole part.The present invention is also implemented as performing one of method as described herein
Point or whole equipment or program of device (for example, computer program and computer program).It is such to realize this
Bright program can be stored on a computer-readable medium, or can have the form of one or more signal.It is such
Signal can be downloaded from internet website and be obtained, or provide on carrier signal, or be provided with any other form.
It should be noted that above-described embodiment the present invention will be described rather than limits the invention, and ability
Field technique personnel can design alternative embodiment without departing from the scope of the appended claims.In the claims,
Any reference symbol between bracket should not be configured to limitations on claims.Word "comprising" is not excluded the presence of not
Element listed in the claims or step.Word "a" or "an" before element does not exclude the presence of multiple such
Element.The present invention can come real by means of the hardware for including some different elements and by means of properly programmed computer
It is existing.If in the unit claim for listing equipment for drying, several in these devices can be by same hardware branch
To embody.The use of word first, second, and third does not indicate that any order.These words can be explained and be run after fame
Claim.
Claims (6)
1. a kind of face identification method, which includes:
The face picture collected is carried out into cluster and obtains multiple 1st layer of classification, and continued at least one by iterative manner
Face picture in i-th layer of classification carries out cluster and obtains multiple i+1 layer classification, and i carries out integer value backward from 1, wherein, extremely
Comprising n corresponding sub-classification, n in a few classification>2;
The 1st layer of classification belonging to target face picture is identified, and is continued in the target face picture institute by iterative manner
+ 1 layer of classification of jth belonging to the target face picture is identified in the jth layer classification of category, j sequentially carries out integer backward from 1 and takes
Value;
By the iterative manner until there is no+1 layer of classification of jth in the jth layer classification belonging to the target face picture
When, from the jth layer classification belonging to the target face picture, identify the similar face picture of the target face picture;
It is described by the face picture collected carry out cluster obtain it is multiple 1st layers classification the step of include:
According to the feature of the face picture collected, generate described in the characteristic vector of face picture collected;
Multiple initial center points, and the characteristic vector according to the face picture collected and each described initial center are set
The face picture collected is divided into multiple 1st layer of classification by the distance of point, and according to the people of each the 1st layer classification
The characteristic vector of face picture, calculates the vector center point of each the 1st layer classification;
Wherein, it is described by the face picture collected carry out cluster obtain it is multiple 1st layers classification the step of also include:
Variance between the initial center point and vector center point of each the 1st layer classification described in calculating;
Size such as the variance exceedes predetermined threshold value, then reset initial center point, and again by the people for having collected
Face picture is divided into multiple 1st layer of classification, and recalculates the vector center point of each the 1st layer classification.
2. face identification method according to claim 1, wherein, the 1st layer identified belonging to target face picture
The step of classification, includes:
According to the feature of the target face picture, the characteristic vector of the target face picture is generated;
The 1st layer of classification that distance between vector center point and the characteristic vector of the target face picture is minimum is selected, as institute
State the 1st layer of classification belonging to target face picture.
3. the face identification method according to any one of claim 1-2, wherein, it is described to identify the target face picture
Similar face picture the step of include:
From the face picture of jth layer classification, select between characteristic vector and the characteristic vector of the target face picture
At least one minimum face picture of distance, as the described similar face picture of the target face picture.
4. a kind of face identification system, which includes:
Sort module, for by the face picture collected carry out cluster obtain it is multiple 1st layers classification, and by iterative manner after
The continuous face picture at least one i-th layers of classification carries out cluster and obtains multiple i+1 layer classification, and i is sequentially carried out backward from 1
Integer value, wherein, comprising n corresponding sub-classification, n at least one classification>2;
Classification iteration identification module, for identifying the 1st layer of classification belonging to target face picture, and is continued by iterative manner
+ 1 layer of classification of jth belonging to the target face picture, j are identified in the jth layer classification belonging to the target face picture
Integer value is carried out backward from 1;
Similar face picture identification module, for there is no+1 layer of jth in the jth layer classification belonging to the target face picture
During classification, from the jth layer classification belonging to the target face picture, the similar face figure of the target face picture is identified
Piece;
First eigenvector generation module, for the feature according to the face picture collected, has been collected described in generation
The characteristic vector of face picture;The sort module arranges multiple initial center points, and according to the face picture collected
Characteristic vector and each initial center point distance, the face picture collected is divided into into multiple 1st layers
Classification, and the characteristic vector of the face picture according to each the 1st layer classification, calculate the vector center of each the 1st layer classification
Point;
Also include:
Variance computing module, calculates the variance between the initial center point and vector center point of each the 1st layer classification;
Size such as the variance exceedes predetermined threshold value, then the sort module resets initial center point, and again by institute
State the face picture collected and be divided into multiple 1st layer of classification, and recalculate the vector center point of each the 1st layer classification.
5. face identification system according to claim 4, wherein, also include:
Second feature vector generation module, for the feature according to the target face picture, generates the target face picture
Characteristic vector;
The sort module selects distance is minimum between vector center point and the characteristic vector of the target face picture the 1st layer
Classification, classifies as the 1st layer belonging to the target face picture.
6. the face identification system according to any one of claim 4-5, wherein,
In the face picture that the similar face picture identification module is classified from the jth layer, characteristic vector and the mesh are selected
At least one minimum face picture of distance between the characteristic vector of mark face picture, as described in the target face picture
Similar face picture.
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410306005.7A CN104036261B (en) | 2014-06-30 | 2014-06-30 | Face identification method and system |
PCT/CN2015/082550 WO2015197029A1 (en) | 2014-06-27 | 2015-06-26 | Human face similarity recognition method and system |
US15/322,350 US20170132457A1 (en) | 2014-06-27 | 2015-06-26 | Human face similarity recognition method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410306005.7A CN104036261B (en) | 2014-06-30 | 2014-06-30 | Face identification method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104036261A CN104036261A (en) | 2014-09-10 |
CN104036261B true CN104036261B (en) | 2017-03-29 |
Family
ID=51467027
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410306005.7A Expired - Fee Related CN104036261B (en) | 2014-06-27 | 2014-06-30 | Face identification method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104036261B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2015197029A1 (en) * | 2014-06-27 | 2015-12-30 | 北京奇虎科技有限公司 | Human face similarity recognition method and system |
CN104850600B (en) * | 2015-04-29 | 2019-05-28 | 百度在线网络技术(北京)有限公司 | A kind of method and apparatus for searching for the picture comprising face |
CN110348274B (en) * | 2018-04-08 | 2022-03-04 | 杭州海康威视数字技术股份有限公司 | Face recognition method, device and equipment |
CN108664920B (en) * | 2018-05-10 | 2022-12-20 | 东方网力科技股份有限公司 | Real-time large-scale cascading face clustering method and device |
CN110069989B (en) * | 2019-03-15 | 2021-07-30 | 上海拍拍贷金融信息服务有限公司 | Face image processing method and device and computer readable storage medium |
CN110874419B (en) * | 2019-11-19 | 2022-03-29 | 山东浪潮科学研究院有限公司 | Quick retrieval technology for face database |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102402508A (en) * | 2010-09-07 | 2012-04-04 | 华东师范大学 | Similar image search device and search method thereof |
CN103617217A (en) * | 2013-11-20 | 2014-03-05 | 中国科学院信息工程研究所 | Hierarchical index based image retrieval method and system |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7734067B2 (en) * | 2004-12-07 | 2010-06-08 | Electronics And Telecommunications Research Institute | User recognition system and method thereof |
-
2014
- 2014-06-30 CN CN201410306005.7A patent/CN104036261B/en not_active Expired - Fee Related
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102402508A (en) * | 2010-09-07 | 2012-04-04 | 华东师范大学 | Similar image search device and search method thereof |
CN103617217A (en) * | 2013-11-20 | 2014-03-05 | 中国科学院信息工程研究所 | Hierarchical index based image retrieval method and system |
Non-Patent Citations (1)
Title |
---|
《基于聚类的大型人脸检索系统》;刘燕 等;《计算机工程》;20050831;第31卷(第15期);第163页第1栏第6段至第2栏第3段 * |
Also Published As
Publication number | Publication date |
---|---|
CN104036261A (en) | 2014-09-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104036261B (en) | Face identification method and system | |
CN109685776B (en) | Pulmonary nodule detection method and system based on CT image | |
CN110660062B (en) | Point cloud instance segmentation method and system based on PointNet | |
Tang et al. | A multi-stage framework with context information fusion structure for skin lesion segmentation | |
CN110969087B (en) | Gait recognition method and system | |
US20080192995A1 (en) | Example-Based Diagnosis Decision Support | |
CN106255968A (en) | Natural language picture search | |
WO2015197029A1 (en) | Human face similarity recognition method and system | |
JP2005535952A (en) | Image content search method | |
CN107315795B (en) | The instance of video search method and system of joint particular persons and scene | |
CN107679539B (en) | Single convolution neural network local information and global information integration method based on local perception field | |
CN104268275B (en) | It is a kind of that data are done with the method that business is abstracted and path searching is analyzed | |
CN108764018A (en) | A kind of multitask vehicle based on convolutional neural networks recognition methods and device again | |
WO2016057000A1 (en) | Customs tariff code classification | |
Garro et al. | Scale space graph representation and kernel matching for non rigid and textured 3D shape retrieval | |
CN114529581A (en) | Multi-target tracking method based on deep learning and multi-task joint training | |
CN110909040B (en) | Business delivery auxiliary method and device and electronic equipment | |
KR102039244B1 (en) | Data clustering method using firefly algorithm and the system thereof | |
Prasomphan | Toward Fine-grained Image Retrieval with Adaptive Deep Learning for Cultural Heritage Image. | |
Chang et al. | Multi-granularity visualization of trajectory clusters using sub-trajectory clustering | |
CN109299304A (en) | Target image search method and system | |
CN110377774B (en) | Method, device, server and storage medium for person clustering | |
CN110032933B (en) | Image data acquisition method and device, terminal and storage medium | |
KR101953479B1 (en) | Group search optimization data clustering method and system using the relative ratio of distance | |
JP2017004052A (en) | Learning device, method, and program |
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 | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170329 Termination date: 20210630 |
|
CF01 | Termination of patent right due to non-payment of annual fee |