CN109409212A - Face recognition method based on cascade BGP - Google Patents
Face recognition method based on cascade BGP Download PDFInfo
- Publication number
- CN109409212A CN109409212A CN201811062012.1A CN201811062012A CN109409212A CN 109409212 A CN109409212 A CN 109409212A CN 201811062012 A CN201811062012 A CN 201811062012A CN 109409212 A CN109409212 A CN 109409212A
- Authority
- CN
- China
- Prior art keywords
- bgp
- feature
- face
- cascade
- histogram
- 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
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F18/2148—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the process organisation or structure, e.g. boosting cascade
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
- G06V10/507—Summing image-intensity values; Histogram projection analysis
-
- 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
Abstract
The invention discloses a face recognition method based on cascade BGP, which comprises the following steps: s1, sequentially carrying out multistage BGP feature extraction on input face image data by adopting a BGP algorithm, wherein BGP feature vectors obtained by previous stage extraction are used as input of next stage BGP feature extraction, and a plurality of BGP feature vectors corresponding to each stage are finally obtained; s2, splicing and fusing the obtained multiple BGP feature vectors to obtain a multi-stage BGP feature vector; and S3, performing face recognition by using the multistage BGP feature vectors, and outputting a recognition result. The method has the advantages of simple implementation method, capability of extracting more abundant texture information, enhancement of feature discrimination and robustness, high identification precision, high efficiency and the like.
Description
Technical field
The present invention relates to technical field of face recognition more particularly to a kind of face identification methods based on cascade BGP.
Background technique
Recognition of face is and several known bodies of given storage under the premise of the still image of given scenario or dynamic video
The face database of part, the identity of verifying and someone or multiple people in identification scene.In recent years, recognition of face is figure always
As the hot issue of the area researches such as processing, computer vision, pattern-recognition, cognitive science, and has and differentiate obvious, acquisition
The features such as being easy is widely used in the occasions such as safety verification, quick payment, video surveys, personal identification.
Current face recognition algorithms can be divided into following a few classes: 1) recognition methods based on face local feature, such as office
Portion's binary pattern (Local Binary Pattern, abbreviation LBP), elastic graph matching, binary system gradient mode (Binary
The methods of Gradient Pattern, abbreviation BGP);2) recognition methods based on face global characteristics, such as linear discriminant analysis
(Linear Discriminant Analysis, abbreviation LDA), principal component analysis (Principle Component
Analysis, abbreviation PCA), the methods of independent component analysis (Independent Component Analysis, abbreviation ICA);
3) method combined based on global characteristics and local feature, the method such as combined based on eigenface and five features;4) based on deep
The method for spending study, such as the proposition and application of facenet.
The it is proposed of binary system gradient mode (BGP) is thought based on image gradient direction (IGO) and binary system describing mode
It is derived from the novel descriptor of Gradientfaces presumably, which replaces image pixel intensities come to people with image gradient direction (IGO)
Face is described, to realize that the robustness to illumination change, i.e., the aspect ratio that gradient field is extracted have more from the feature of intensity domain
There are identification and robustness.By measuring the relationship between the local pixel in image gradient domain in BGP algorithm, and by bottom office
Portion's structure efficient coding is one group of string of binary characters, not only increases judgement index, even more enormously simplifies computation complexity.
In order to find gradient field potential structure, BGP is and to be encoded to a series of two from multi-directional computing image gradient
System string can indicate the variation of small boundary and texture information, therefore have very strong identification, though in face of blocking, illumination,
Expression shape change etc., can also obtain preferable accuracy of identification, and when due to being encoded using BGP to image, each encoded radio
The information for having contained neighborhood territory pixel relationship, rather than just the strength information of pixel itself, therefore the image after BGP coding is to each
Kind environmental change more robust, especially has stronger illumination invariant.Basic description of BGP is as shown in Figure 1, wherein (a) is right
The eight adjacent pixels (value 115) that should be a center pixel, (b) are four direction, (c) are main binary string, coding
It is 0111, label 7.In recognition methods based on face local feature, binary system gradient mode (BGP) is a kind of succinct efficient
Face describe son, higher accuracy of identification is obtained in recognition of face based on BGP.
As above-mentioned, binary system gradient mode has the characteristics that calculating is simple, identification is strong, robustness is good, is very suitable for answering
In recognition of face for being difficult to differentiate between, BGP feature specifically has the advantage that BGP is defined in image gradient directional diagram, tool
There are good Gradient Features, the variation such as intensity of illumination can be successfully managed;And tactic pattern and more spatial discriminations are used in BGP
Rate, structuring BGP function as edge detector, this be the key that accurately identify and succinctly indicate, meanwhile, more spatial discriminations
Rate strategy increases the ability of descriptor covering different radii neighborhood territory pixel.
But although higher accuracy of identification can be obtained in recognition of face currently based on BGP, currently based on
BGP realize recognition of face when, be usually all using BGP only to feature of face extraction, directly from the feature that extraction obtains to
Amount is identified, and obtained texture information is not practical abundant enough after a BGP is extracted, for original image,
Can still there be certain information loss, the precision identified based on the BGP eigen vector extracted is still to be improved.
Summary of the invention
The technical problem to be solved in the present invention is that, for technical problem of the existing technology, the present invention provides one
Plant simple implementation method, accuracy of identification and the high-efficient face identification method based on cascade BGP.
In order to solve the above technical problems, technical solution proposed by the present invention are as follows:
A kind of face identification method based on cascade BGP, step include:
S1. multistage BGP is extracted: successively being carried out multistage BGP feature based on BGP algorithm to the face image data of input and is mentioned
It takes, wherein input of the BGP characteristic image obtained by upper level BPG feature extraction as next stage BPG feature extraction, final
The multiple BGP feature vectors at different levels to correspondence;
S2. multi-stage characteristics splice: carrying out splicing fusion to obtained multiple BGP feature vectors, it is special to obtain multistage BGP
Levy vector;
S3. recognition of face: recognition of face is carried out using the multistage BGP feature vector, exports recognition result.
As a further improvement of the present invention: in the step S1, original facial image will be inputted and carry out a BGP feature
It extracts, obtains the 1st grade of BGP characteristic image and the 1st grade of BGP feature vector, then successively carry out (i-1)-th grade of BGP characteristic image
BGP feature extraction obtains i-stage characteristic image and i-stage BGP feature vector, and wherein i=2,3,4 ... n, n are
The BGP feature extraction series of required execution.
As a further improvement of the present invention: carrying out the step of single-stage BGP feature extraction in the step S1 using BGP algorithm
Suddenly are as follows:
S11. feature extraction is carried out to input image data based on BGP algorithm, obtains BGP characteristic image;
S12. the BGP characteristic image step S11 obtained is divided into the sub-block not overlapped;
S13. the BGP histogram for each sub-block that the step S12 is obtained is counted;
S14. all sub-block histograms step S13 obtained splice in order obtains final BGP feature vector.
As a further improvement of the present invention: the BGP characteristic image being divided into kk sub-block in the step S12
A1、 A2... ..., Ak·k;The BGP feature vector obtained in the step S14 is P=[XA1,XA2,...XAk,
...XAk·k], wherein XA1For A1The statistic histogram ... ... of sub-block, XAkFor AkThe statistic histogram of sub-block.
As a further improvement of the present invention: the BGP feature extraction of three-level or more is specifically executed in the step S1;Single-stage
The BGP characteristic image that extraction obtains is divided into the sub-block of 32*32 when BGP feature extraction.
As a further improvement of the present invention: multistage BGP feature vector obtained in the step S2 is Pfinal=[P1,
P2,...Pn], wherein P1For the 1st grade of BGP feature vector, P2For the 2nd grade of BGP feature vector, PnFor n-th grade of BGP feature vector.
As a further improvement of the present invention: in the step S3, the specific method using histogram intersection core, which calculates, to be made
Feature vector for training sample and the characteristic similarity between the feature vector as test sample, to carry out recognition of face.
As a further improvement of the present invention: specific calculating histogram intersection core according to the following formula is simultaneously similar as the feature
Degree:
Wherein, p and q is the histogram of the training sample and test sample respectively, and k is the pixel number for including in histogram
Amount, p (i) and q (i) are respectively the corresponding histogram component of ith pixel, i=1,2,3 ... k.
As a further improvement of the present invention: further including by the histogram intersection core divided by training sample, test sample
In one of histogram in all pixels number, obtain final standardized feature similarity result for recognition of face.
As a further improvement of the present invention: similar with the feature of training sample in acquisition test sample in the step S3
It is identified after degree using nearest neighbor classifier.
Compared with the prior art, the advantages of the present invention are as follows:
1, the present invention is based on the face identification methods of cascade BGP passes through on the basis of realizing recognition of face using BGP
To input picture carry out multistage BGP feature extraction, will after obtained BGP feature fused in tandem at different levels composition multistage BGP feature into
Row identification, realizes the recognition of face of multi-stage cascade BGP, based on the mode of multistage BGP, can sufficiently excavate the deeper time of picture
The useful informations such as marginal information, texture information, gradient information, formation more stablizes more robust character representation, so as to
Richer texture information is enough extracted, recognition of face precision is improved.
2, the present invention is based on the face identification methods of cascade BGP, and BGP feature is first extracted on the basis of original image, obtains one
Width BGP characteristic pattern carries out a BGP feature extraction to the BGP characteristic pattern again, can further extract more depth, more hidden
Contain, the useful informations such as richer texture information and marginal information, is all used as single-stage BGP to calculate every grade of BGP feature coding image
The input of method obtains multiple groups combination of eigenvectors can be realized efficiently by feature vector after merging after multiple groups feature vector
Recognition of face.
3. recognition of face is realized based on structuring BGP, by structuring the present invention is based on the face identification method of cascade BGP
BGP extracts structure gradient mode and carries out recognition of face as binary string, to illumination, blocks etc. and to all have very strong robust
Property.
Detailed description of the invention
Fig. 1 is that BGP describes sub- schematic illustration substantially.
Fig. 2 is the implementation process schematic diagram of face identification method of the present embodiment based on cascade BGP.
Fig. 3 is structuring BGP and unstructured BGP schematic illustration.
Fig. 4 is that single-stage BGP seeks feature vector flow diagram in the present embodiment.
Fig. 5 is the BGP histogram for the sub-block that the specific embodiment of the invention obtains.
Fig. 6 is that multistage BGP seeks feature vector schematic diagram in the present embodiment
Fig. 7 is three-level BGP feature coding image schematic diagram obtained in the specific embodiment of the invention.
Fig. 8 is personal sector's image schematic diagram in the Yale of the invention used.
Fig. 9 be in the YaleB that uses of the present invention third group to the 5th group of image schematic diagram.
Figure 10 is the simulation result schematic diagram that the library Yale subgraph size influences discrimination.
Figure 11 is the simulation result schematic diagram for extending YaleB third group subgraph size and influencing.
Figure 12 is the simulation result schematic diagram for extending the 4th group of subgraph size of YaleB and influencing.
Figure 13 is the simulation result schematic diagram for extending the 5th group of subgraph size of YaleB and influencing.
Specific embodiment
Below in conjunction with Figure of description and specific preferred embodiment, the invention will be further described, but not therefore and
It limits the scope of the invention.
As shown in Fig. 2, face identification method step of the present embodiment based on cascade BGP includes:
S1. multistage BGP is extracted: successively being carried out multistage BGP feature based on BGP algorithm to the face image data of input and is mentioned
It takes, wherein input of the BGP characteristic image obtained by upper level BPG feature extraction as next stage BPG feature extraction, final
The multiple BGP feature vectors at different levels to correspondence;
S2. multi-stage characteristics splice: splicing fusion is carried out to obtained multiple BGP feature vectors, obtain multistage BGP feature to
Amount;
S3. recognition of face: recognition of face is carried out using multistage BGP feature vector, exports recognition result.
BGP (binary system gradient mode) replaces image pixel intensities that face is described using image gradient direction, passes through
Image after BGP coding, the information for describing pixel have contained the information of neighborhood territory pixel relationship, have been not only gray value information, and
The aspect ratio that gradient field is extracted has more identification and robustness from the feature of intensity domain.The present embodiment is realized using BGP
On the basis of recognition of face, by carrying out multistage BGP feature extraction to input picture, the series connection of obtained BGP features at different levels is melted
Multistage BGP feature is constituted after conjunction to be identified, realizes the recognition of face of multistage BGP, due to the face characteristic image after BGP coding
It is still gray level image, which still contains the abundant informations such as grayscale information, texture information, marginal information, obtains at every grade
BGP characteristic image on the basis of when applying BGP again, obtained feature vector can obtain deeper time relative to original image
Information, thus the mode based on multistage BGP can sufficiently excavate deeper marginal information, the texture information, ladder of picture
The useful informations such as information are spent, effectively enhancing forms to the characterization ability of face and more stablizes more robust character representation, thus
Richer texture information can be extracted, the identification and robustness of feature are further enhanced, improves accuracy of identification.
BGP's the realization process includes: a series of adjacent pixel of a given center pixel and parts (than 8 as shown in figure 1 adjacent pictures
Element), according to formula (1) based on the symmetrical adjacent pixel of two in each direction, it is (main and auxiliary to calculate a pair of of binary coding
Help), as shown in (b) (c) of Fig. 1, available 4 pairs of binary numbers from the four directions such as G1, G2, G3, G4;
The label of center pixel is obtained by four main binary codings, i.e. BGP binary digit indicates such as formula (2).
It is converted into decimal number such as formula (3):
Eight binary numbers are obtained on four direction, main and auxiliary binary number in each direction is always complementary
, each direction only needs bit to embody.For succinct expression, the present embodiment only needs main binary system
Position calculates label (according to formula (3)), and table 1 enumerates several typical cases, and each nine grids center pixel passes through BGP in table 1
A value after coding, between 0 to 15 being encoded as.
Table 1:BGP typical case
Structuring BGP and unstructured BGP is as shown in figure 3, four direction BGP descriptors have 16 kinds of different labels, each
Label is made of 8 binary digits, including four " 1 " He Siwei " 0 ", wherein there is eight labels to have continuous position " 1 ", is such as schemed
Shown in middle bold box, the position " 1 " of other eight labels be it is discontinuous, as shown in frame thin in figure, above-mentioned continuous " 1 " indicates line
More stable localized variation is managed, describes the orientation of local edge substantially, and statistics obtains this mode in typical BGP spy
The overwhelming majority is accounted in sign coded image.The above-mentioned mode with continuous " 1 " is structuring BGP, other modes are referred to as non-structural
Mode, non-structural mode more represents noise jamming, and number is considerably less in BGP coding, so back statistic histogram is special
Ignored when sign, statistical framework BGP.
For a width gray scale picture, can be acquired by binary system gradient mode feature vector (dimension d1), if
Color image then needs to be converted to gray level image first.The present embodiment is specifically based on above structure BGP and realizes recognition of face, by tying
Structure BGP extracts structure gradient mode and carries out recognition of face as binary string, to illumination, block etc. all have it is very strong
Robustness.
In the present embodiment step S1, it will specifically input original facial image and carry out a BGP feature extraction, and obtain the 1st grade
BGP characteristic image and the 1st grade of BGP feature vector, then (i-1)-th grade of BGP characteristic image is successively subjected to a BGP feature and is mentioned
It takes, obtains i-stage characteristic image and i-stage BGP feature vector, wherein i=2,3,4 ... n, n are the BGP of required execution
Feature extraction series.
As shown in figure 4, the step of carrying out single-stage BGP feature extraction using BGP algorithm in the present embodiment step S1 are as follows:
S11. feature extraction is carried out to input image data based on BGP algorithm, obtains BGP characteristic image;
S12. the BGP characteristic image that step S11 is obtained is divided into the sub-block A that k*k is not overlapped1、A2... ..., Ak*k;
S13. the BGP histogram for each sub-block that statistic procedure S12 is obtained;
S14. all sub-block histograms step S13 obtained splice in order obtains final BGP feature vector.
When single-stage BGP feature extraction, i.e. n=1, present embodiment assumes that take radius of neighbourhood R=1, then BGP neighborhood territory pixel number
P=8, piecemeal number M*N (specifically enabling M=N=5), since 8 neighborhood territory pixels determine that (4 main, 4 auxiliary in 8 directions
), then the BGP encoded radio of any one center pixel should be tetrad, and being converted into decimal value should be 0 to 15,
16 kinds of modes have 8 kinds of tactic patterns, ignore non-structural mode, then the BGP histogram dimension d=8 of each sub-block, facial image warp
Histogram dimension after single-stage BGP coding should be d1=200 (5*5*8), and obtained BGP feature vector is P=[XA1,
XA2,...XAk,...XAk·k], wherein XA1For A1The statistic histogram ... ... of sub-block, XAkFor AkThe statistic histogram of sub-block.
After carrying out BGP feature extraction to a width gray level image, a width BGP characteristic image is obtained.In concrete application embodiment
In, will a width BGP characteristic image carry out 5*5 piecemeal after obtain 25 sub-blocks, 25 sub-blocks put in order for from left to right, from
Obtain that the results are shown in Table 2 under.
Table 2:BGP characteristic image piecemeal sequence.
In a sub-block obtained in concrete application embodiment 8 kinds of structure BGP statistic histogram as shown in figure 5, its
In non-structural BGP ignore, it is seen that the histogram is the histogram after normalization, can be 8 kinds in effecting reaction this sub-block
The frequency that tactic pattern occurs, uses XA1Indicate A1The frequency that the statistic histogram of sub-block, i.e. 8 kinds of tactic patterns occur, is expressed as 8
Row vector is tieed up, similarly XAkIndicate AkThe statistic histogram of sub-block splices the histogram vectors of all sub-blocks, composition it is final to
Amount P is the feature vector that this gray level image is calculated through binary system gradient mode, and feature vector P is indicated are as follows:
P=[XA1,XA2,...XAk,...,XA25] (4)
As shown in fig. 6, the present embodiment first extracts BGP feature on the basis of original image, a width BGP characteristic pattern, the figure are obtained
Still there is grayscale information, texture information, gradient information etc., a BGP feature extraction is carried out to the BGP characteristic pattern again, it can
Further to extract the useful informations such as more depth, more implicit, richer texture information and marginal information, in this mode according to
After the secondary progress feature extraction to coding characteristic image, available multistage BGP characteristic image;By every grade of BGP feature coding image
Input all as single-stage BGP algorithm, then available multiple groups feature vector, finally by multiple groups combination of eigenvectors to get arriving
Feature vector after multistage BGP effect.
Since original image and every grade of BGP feature coding image are equal in magnitude, and partitioned mode is identical, then every grade of BGP is pressed
Feature vector is calculated according to formula (4), when n=1 (represent coding primary), obtains feature vector P1;When n=2, feature vector is obtained
P2... and so on, the final fused feature vector P of n grades of BGPfinal:
Pfinal=[P1,P2,...,Pn] (5)
That is multistage BGP feature vector obtained in step S2 is Pfinal=[P1,P2,...,Pn], wherein P1For the 1st grade of BGP
Feature vector, P2For the 2nd grade of BGP feature vector, PnFor n-th grade of BGP feature vector.
In the present embodiment step S3, the specific method using histogram intersection core calculates the feature vector as training sample
Characteristic similarity between the feature vector as test sample, to carry out recognition of face, i.e., in the recognition of face stage, specifically
The characteristic similarity of training sample and test sample is calculated using the method for histogram intersection core.
If p and q are two histograms containing k bin, component is p (i) and q (i) respectively, wherein i=1,2,3 ...
k.Then their similarity measurement distance is as shown in formula (6):
The present embodiment specifically calculates histogram intersection core according to formula (6) and as characteristic similarity, and wherein p and q are right respectively
It should be the histogram of training sample and test sample, k is the pixel quantity for including in histogram, and p (i) and q (i) are respectively i-th
The corresponding histogram component of a pixel, i=1,2,3 ... k, with by two histograms of histogram intersection kernel representation in each bin
Shared number of pixels.
The present embodiment further passes through can be real divided by all pixels number in one of histogram with histogram intersection core
Now standardize so that end value is in [0,1], specifically by histogram intersection core divided by training sample, test sample wherein
All pixels number in one histogram, final standardized feature similarity result is calculated for face in (7) according to the following formula
Identification.
The present embodiment further uses nearest neighbor classifier after obtaining the characteristic similarity of test sample and training sample
It is identified to realize, realization principle is simple, it may be convenient to realize the assessment for describing sub performance itself.
In concrete application embodiment, using the three-level BGP coding characteristic figure of width facial image obtained by the above method
As shown in fig. 7, (b), (c), (d) respectively correspond image of the original image (a) after level-one, two-stage, three-level BGP coding, wherein
Fig. 7 (b) retains original image most information, can clearly identify the five features such as eyes, nose, mouth, shows that level-one BGP is special
Sign is that have stronger characterization ability than more completely characterization to original image one;Fig. 7 (c) clarity is compared to Fig. 7 (b)
Decline, but be the extraction of the information such as the gradient to Fig. 7 (b), edge, texture, and still be able to identify eye more visiblely
The five features such as eyeball, nose;Although finally, the decline of Fig. 7 (d) clarity is serious, but containing richer texture information,
It is the further extraction to its further feature for original image, can break through common description, can only to obtain image shallow
The obstacle of layer feature, extracts texture information more abundant, obtains the identification performance for having more identification, enhances to illumination, table
The robustness of the variations such as feelings improves recognition of face precision, obtains better recognition result.
In order to verify cascade BGP Feature fusion proposed in this paper, the present embodiment is marked in Yale and the YaleB of extension etc.
Quasi- face database is tested, including level-one, second level, three-level BGP Feature fusion, is respectively corresponded and is denoted as BGP1, BGP2,
BGP3.The library Yale that the present embodiment uses includes 15 research objects, everyone 11 face images, totally 165 width facial image.Its
In include illumination, expression, the variation such as block, personal sector's facial image is as shown in Figure 8 in Yale.In the YaleB of extension
Database includes 38 research objects, shares about 22000 face images, the corresponding 9 different postures and 64 of each object
A lighting condition, one of them widely used subset include from frontal pose all face-images (64 × 38=
2432) it, can be carried out in the experiment of illumination change, data set is divided into five different groups, and corresponding illumination effect is increasingly
Greatly, wherein shown in example face such as Fig. 9 (a)-(c) of third group to the 5th group.
Original image on condition that is divided into the subgraph not overlapped by the application of above-mentioned BGP feature extracting method, piecemeal subgraph
As the selection of size is just particularly significant, there can be direct influence on recognition result, if institute's block count is very little, extreme case is exactly
Original image just can not embody the local message of image in this way, if institute's block count is too many, extreme case is exactly that each pixel is one
A subgraph, and computation complexity can be considerably increased, and be readily incorporated noise.Yale, YaleB third group, YaleB the 4th
Influence experimental result of the piecemeal size to discrimination is as shown in Figure 10 to Figure 13 in the 5th group of group, YaleB, and wherein horizontal axis is point
Block subgraph size, takes 4*4,8*8,16*16,32*32 respectively, and the longitudinal axis is the face identification rate in database.
It can be obtained by Figure 10 to Figure 13, for piecemeal size in 32*32 discrimination highest, and on the whole, the performance of three-level BGP is strong
In second level BGP, second level BGP performance is better than level-one BGP, i.e. second level, three-level BGP enhances face table relative to original BGP feature
Sign ability improves the robustness of algorithm, to significantly improve face identification rate.Subgraph block count is more, local feature
Description is more detailed, and the identification and robustness for the BGP Feature fusion applied on this basis are stronger, but block number excessively causes
Cost prohibitive is calculated, and invalid noise can be introduced, three-level BGP, every grade of BGP is specifically used to divide in concrete application embodiment of the present invention
It is divided into 32*32 when block, cost of implementation and invalid noise etc. is reduced while to guarantee performance.
The present embodiment is further by three kinds of algorithms and algorithms most in use (LTV model, logarithm wavelet transformation (LWT), TCT-
MPCA, RLS) it is all made of 32*32 scale, is obtained in 345 groups of carry out Experimental comparisons of extension YaleB and analysis, piecemeal size
The experimental result arrived is as shown in table 3 below.
Table 3:YaleB database algorithms performance comparison table
It can be obtained by table 3, for the BGP series methods algorithm advantageous compared to traditional performance, realize the promotion of performance
And leap, the present invention is based on the face identification methods of cascade BGP, can obtain higher discrimination on the basis of BGP method
And better performance, especially under the influence of serious lighting condition, such as the 4th group and the 5th group obtains apparent performance and changes
It is kind.
The above-mentioned face identification method based on cascade BGP of the present invention, can sufficiently excavate the deeper edge of picture
The useful informations such as information, texture information, gradient information can extract richer feature in such a way that multistage BGP is merged,
And fusion feature more has identification and robustness.
Above-mentioned only presently preferred embodiments of the present invention, is not intended to limit the present invention in any form.Although of the invention
It has been disclosed in a preferred embodiment above, however, it is not intended to limit the invention.Therefore, all without departing from technical solution of the present invention
Content, technical spirit any simple modifications, equivalents, and modifications made to the above embodiment, should all fall according to the present invention
In the range of technical solution of the present invention protection.
Claims (10)
1. a kind of face identification method based on cascade BGP, which is characterized in that step includes:
S1. multistage BGP is extracted: BGP algorithm is based on to the face image data of input and successively carries out multistage BGP feature extraction,
In the input of the BGP characteristic image that is obtained by upper level BPG feature extraction as next stage BPG feature extraction, finally obtain pair
Answer multiple BGP feature vectors at different levels;
S2. multi-stage characteristics splice: splicing fusion is carried out to obtained multiple BGP feature vectors, obtain multistage BGP feature to
Amount;
S3. recognition of face: recognition of face is carried out using the multistage BGP feature vector, exports recognition result.
2. the face identification method according to claim 1 based on cascade BGP, which is characterized in that, will in the step S1
It inputs original facial image and carries out a BGP feature extraction, obtain the 1st grade of BGP characteristic image and the 1st grade of BGP feature vector,
(i-1)-th grade of BGP characteristic image is successively subjected to a BGP feature extraction again, i-stage characteristic image is obtained and i-stage BGP is special
Vector is levied, wherein i=2,3,4 ... n, n are the BGP feature extraction series of required execution.
3. the face identification method according to claim 1 based on cascade BGP, which is characterized in that adopted in the step S1
The step of carrying out single-stage BGP feature extraction with BGP algorithm are as follows:
S11. feature extraction is carried out to input image data based on BGP algorithm, obtains BGP characteristic image;
S12. the BGP characteristic image step S11 obtained is divided into the sub-block not overlapped;
S13. the BGP histogram for each sub-block that the step S12 is obtained is counted;
S14. all sub-block histograms step S13 obtained splice in order obtains final BGP feature vector.
4. the face identification method according to claim 3 based on cascade BGP, which is characterized in that will in the step S12
The BGP characteristic image is divided into kk sub-block A1、A2... ..., Ak·k;The BGP feature obtained in the step S14 to
Amount is P=[XA1,XA2,...XAk,...XAk·k], wherein XA1For A1The statistic histogram ... ... of sub-block, XAkFor AkThe system of sub-block
Count histogram.
5. the face identification method based on cascade BGP described according to claim 1~any one of 4, which is characterized in that institute
State the BGP feature extraction that three-level or more is specifically executed in step S1;The BGP feature for obtaining extraction when single-stage BGP feature extraction
Image is divided into the sub-block of 32*32.
6. the face identification method based on cascade BGP described according to claim 1~any one of 4, which is characterized in that institute
Stating multistage BGP feature vector obtained in step S2 is Pfinal=[P1,P2,...Pn], wherein P1For the 1st grade of BGP feature vector,
P2For the 2nd grade of BGP feature vector, PnFor n-th grade of BGP feature vector.
7. the face identification method based on cascade BGP described according to claim 1~any one of 4, which is characterized in that institute
It states in step S3, the specific method using histogram intersection core is calculated as the feature vector of training sample and as test sample
Feature vector between characteristic similarity, to carry out recognition of face.
8. the face identification method according to claim 7 based on cascade BGP, which is characterized in that specifically count according to the following formula
Calculate histogram intersection core and as the characteristic similarity:
Wherein, p and q is the histogram of the training sample and test sample respectively, and k is the pixel quantity for including, p in histogram
It (i) is respectively the corresponding histogram component of ith pixel, i=1,2,3 ... k with q (i).
9. the face identification method according to claim 8 based on cascade BGP, which is characterized in that further including will be described straight
Side's figure intersects core divided by all pixels number in one of histogram in training sample, test sample, obtains final standard
Change characteristic similarity result and is used for recognition of face.
10. the face identification method based on cascade BGP described according to claim 1~any one of 4, which is characterized in that
It is identified after obtaining the characteristic similarity of test sample and training sample using nearest neighbor classifier in the step S3.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811062012.1A CN109409212A (en) | 2018-09-12 | 2018-09-12 | Face recognition method based on cascade BGP |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811062012.1A CN109409212A (en) | 2018-09-12 | 2018-09-12 | Face recognition method based on cascade BGP |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109409212A true CN109409212A (en) | 2019-03-01 |
Family
ID=65464726
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811062012.1A Pending CN109409212A (en) | 2018-09-12 | 2018-09-12 | Face recognition method based on cascade BGP |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109409212A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110599776A (en) * | 2019-10-15 | 2019-12-20 | 福州市协成智慧科技有限公司 | Traffic data processing system |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104112122A (en) * | 2014-07-07 | 2014-10-22 | 叶茂 | Vehicle logo automatic identification method based on traffic video |
CN105117688A (en) * | 2015-07-29 | 2015-12-02 | 重庆电子工程职业学院 | Face identification method based on texture feature fusion and SVM |
CN106599870A (en) * | 2016-12-22 | 2017-04-26 | 山东大学 | Face recognition method based on adaptive weighting and local characteristic fusion |
-
2018
- 2018-09-12 CN CN201811062012.1A patent/CN109409212A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104112122A (en) * | 2014-07-07 | 2014-10-22 | 叶茂 | Vehicle logo automatic identification method based on traffic video |
CN105117688A (en) * | 2015-07-29 | 2015-12-02 | 重庆电子工程职业学院 | Face identification method based on texture feature fusion and SVM |
CN106599870A (en) * | 2016-12-22 | 2017-04-26 | 山东大学 | Face recognition method based on adaptive weighting and local characteristic fusion |
Non-Patent Citations (3)
Title |
---|
LIWEI LI等: "《Face Recognition Algorithm Based on Cascading BGP Feature Fusion》", 《2018 CHINESE CONTROL AND DECISION CONFERENCE(CCDC)》 * |
LIWEI LI等: "《Face Recognition Algorithm Based on Cascading BGP Feature Fusion》", 《第30届中国控制与决策会议》 * |
WEILIN HUANG 等: "《Robust face recognition with structural binary gradient patterns》", 《PATTERN RECOGNITION》 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110599776A (en) * | 2019-10-15 | 2019-12-20 | 福州市协成智慧科技有限公司 | Traffic data processing system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Lin | Face detection in complicated backgrounds and different illumination conditions by using YCbCr color space and neural network | |
CN112818862A (en) | Face tampering detection method and system based on multi-source clues and mixed attention | |
Zheng et al. | Attention-based spatial-temporal multi-scale network for face anti-spoofing | |
JP2009514110A (en) | Human detection by pause | |
CN110175248B (en) | Face image retrieval method and device based on deep learning and Hash coding | |
Yang et al. | Spatiotemporal trident networks: detection and localization of object removal tampering in video passive forensics | |
CN109740572A (en) | A kind of human face in-vivo detection method based on partial color textural characteristics | |
CN108108760A (en) | A kind of fast human face recognition | |
Paul et al. | Extraction of facial feature points using cumulative histogram | |
CN107220598A (en) | Iris Texture Classification based on deep learning feature and Fisher Vector encoding models | |
CN110188646B (en) | Human ear identification method based on fusion of gradient direction histogram and local binary pattern | |
Barodi et al. | An enhanced artificial intelligence-based approach applied to vehicular traffic signs detection and road safety enhancement | |
CN117197763A (en) | Road crack detection method and system based on cross attention guide feature alignment network | |
Madan et al. | Traffic Sign Classification using Hybrid HOG-SURF Features and Convolutional Neural Networks. | |
CN112070116B (en) | Automatic artistic drawing classification system and method based on support vector machine | |
Guo et al. | Pedestrian detection via multi-scale feature fusion convolutional neural network | |
CN109409212A (en) | Face recognition method based on cascade BGP | |
Lafuente-Arroyo et al. | Traffic sign classification invariant to rotations using support vector machines | |
Adeyanju et al. | Development of an american sign language recognition system using canny edge and histogram of oriented gradient | |
Fu et al. | Pedestrian detection by feature selected self-similarity features | |
CN113963164A (en) | Texture feature extraction method based on grouping neighborhood intensity difference coding | |
Fradi et al. | A new multiclass SVM algorithm and its application to crowd density analysis using LBP features | |
Wu et al. | Recursive algorithms for image segmentation based on a discriminant criterion | |
Lin et al. | A deep learning based bank card detection and recognition method in complex scenes | |
Li et al. | Finely Crafted Features for Traffic Sign Recognition |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190301 |