CN106778529A - A kind of face identification method based on improvement LDP - Google Patents

A kind of face identification method based on improvement LDP Download PDF

Info

Publication number
CN106778529A
CN106778529A CN201611062329.6A CN201611062329A CN106778529A CN 106778529 A CN106778529 A CN 106778529A CN 201611062329 A CN201611062329 A CN 201611062329A CN 106778529 A CN106778529 A CN 106778529A
Authority
CN
China
Prior art keywords
ldp
improvement
face
pixel
subregion
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
CN201611062329.6A
Other languages
Chinese (zh)
Inventor
王绎博
沙涛
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.)
Nanjing University of Science and Technology
Original Assignee
Nanjing University of Science and Technology
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 Nanjing University of Science and Technology filed Critical Nanjing University of Science and Technology
Priority to CN201611062329.6A priority Critical patent/CN106778529A/en
Publication of CN106778529A publication Critical patent/CN106778529A/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
    • G06V40/161—Detection; Localisation; Normalisation
    • 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
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2132—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on discrimination criteria, e.g. discriminant 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
    • G06V40/168—Feature extraction; Face representation
    • G06V40/171—Local features and components; Facial parts ; Occluding parts, e.g. glasses; Geometrical relationships

Landscapes

  • Engineering & Computer Science (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Health & Medical Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Health & Medical Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Multimedia (AREA)
  • Human Computer Interaction (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Image Processing (AREA)

Abstract

The invention discloses a kind of based on the face identification method for improving LDP, comprise the following steps：Facial image to be identified is converted to the gray level image of formed objects specification, the subregion of formed objects is then cut into；Using LDP methods are improved, the improvement LDP characteristic values of all subregions that face gray level image cuts out are extracted, obtain improving LDP block images and incorporating Structure Comparison information carrying out divided group；Weighting LDP characteristic values according to each sub-regions, extract the improvement LDP histogram features of the subregion, and are integrated into an entirety and are used to represent facial image；Above-mentioned treatment is carried out to the face images in known face database, the overall improvement LDP histogram features of all faces are drawn；The improvement LDP histogram features vector of facial image to be identified is extracted, the characteristic vector with all candidate images is made comparisons, and calculates card side's distance, and the facial image where choosing minimum value is used as matching face.The present invention has the advantages that calculating speed is fast, robustness is good, improves recognition of face success rate.

Description

A kind of face identification method based on improvement LDP
Technical field
It is particularly a kind of based on the face for improving LDP the invention belongs to Digital Image Processing and mode identification technology Recognition methods.
Background technology
Face recognition technology is referred to using COMPUTER DETECTION target image, and locating human face extracts effective identification letter Breath, matches existing standard faces storehouse, obtains a special kind of skill of face identity, as the body commonly used in people's work and life One of part discriminating means.The increasingly extensive of face recognition technology application solves illumination, attitude, expression change to specific method with deep The ability of change proposes requirements at the higher level, therefore continues what research was a need for face identification method.
Face recognition technology mainly includes face characteristic extraction, three steps of Dimensionality Reduction and tagsort.Typically we To extract effective face characteristic as committed step, its validity will directly influence the result of recognition of face.Face characteristic Global characteristics and the class of local feature two can be divided into：What global characteristics method reflected is the integrity attribute of face, the overall situation of main flow Characterization method includes PCA (PCA), Fisher face (LDA) and Independent component analysis (ICA) etc..It is local Characterization method lay particular emphasis on extract face minutia, the recognition methods based on local feature including local feature represent (LFA), Local binary patterns (LBP), local direction pattern (LDP) etc..Local feature to the illumination of face, express one's feelings and the change such as block not Sensitivity, therefore the extracting method based on local feature has more preferable robustness relative to the extracting method based on global characteristics, In recent years come obtained it is more concern with research.
Local binary patterns (LBP) are a kind of operators of the textural characteristics that can effectively describe image local information, by T.Ojala et al. was proposed in 1996.LBP algorithms are realized simply, insensitive for uniformity illumination variation, but to making an uproar at random Sound is undesirable with the graphical representation effect of nonuniformity illumination variation.
The content of the invention
It is an object of the invention to provide the recognition of face side based on improvement LDP that a kind of calculating speed is fast, robustness is good Method, to improve recognition of face success rate.
The technical solution for realizing the object of the invention is：A kind of face identification method based on improvement LDP, including it is as follows Step：
Step 1, facial image to be identified is converted to the gray level image of formed objects specification；
Step 2, step 1 gained face gray level image is cut into the subregion of formed objects；
Step 3, using LDP methods are improved, the improvement LDP for extracting all subregions that face gray level image cuts out is special Value indicative, obtains improving LDP block images；
Step 4, to each improvement LDP block image, incorporating Structure Comparison information carries out divided group；
Step 5, the weighting LDP characteristic values according to each sub-regions, the improvement LDP histograms for extracting the subregion are special Levy；
Step 6, by the improvement LDP histogram features of each sub-regions, is integrated into an entirety and is used to represent facial image；
Face images in known face database are carried out the treatment of above-mentioned steps 1~6 by step 7, draw known face The improvement LDP histogram features of all faces entirety in storehouse；
Step 8, extracts the improvement LDP histogram features vector of facial image to be identified, the feature with all candidate images Vector is made comparisons, and calculates card side's distance, and the facial image where choosing minimum value is used as matching face.
Further, using LDP methods are improved described in step 3, all sub-districts that face gray level image cuts out are extracted The improvement LDP characteristic values in domain, obtain improving LDP block images, specific as follows：
3 × 3 matrix-blocks centered on target pixel points and Kirsch operators are carried out into computing in initial processing stage, will Eight values for obtaining use m as the gray value of surrounding abutmentsiRepresent, i=0,1 ..., 7, maximum k during then this 8 are worth Individual value is entered as 1, and remaining 8-k value is entered as 0, so as to obtain an eight bit, is represented after being converted to decimal number The gray value of central pixel point, used as LDP characteristic values are improved, the process can be represented with equation below：
Wherein, xLDP′K () is the improvement LDP characteristic values of target pixel points, mkIt is k value maximum in mi；
Each sub-regions to cutting out carry out aforesaid operations respectively, change as each pixel in each sub-regions Enter LDP characteristic values.
Further, to each improvement LDP block image described in step 4, incorporating Structure Comparison information carries out divided group, It is specific as follows：
Each sub-regions are proceeded as follows first in units of pixel：
Wherein, miRefer to that, for 3 × 3 matrixes centered on each pixel, 8 abutment points of the pixel are passed through The gray value obtained after Kirsch operator operations；Refer to miAverage value；
According to all of pixel in every sub-regions, the Structure Comparison information of the subregion after each segmentation is obtained, it is public Formula is as follows：
For a sub-regions, it is believed that each pixel is a least unit, xLDP′(r, c) represents the picture of r rows c row The improvement LDP characteristic values of vegetarian refreshments；srcRefer to a Structure Comparison information for the improvement LDP codings of pixel；siTo consider In the subregion after the Structure Comparison information of all pixels point, the overall weighted value of the subregion；M, N are the row of subregion pixel Number and columns.
Further, the weighting LDP characteristic values described in step 5 according to each sub-regions, extract the improvement of the subregion LDP histogram features, formula is as follows：
Wherein, pillar height when H (τ) represents gray value for τ on the histogram, so as to show a sub-district with the histogram table The improvement LDP characteristic values of all pixels point in domain；xLDP′(r, c) represents the improvement LDP characteristic values of the pixel of r rows c row； M, N are the line number and columns of subregion pixel.
Further, the improvement LDP histogram features vector of facial image to be identified is extracted described in step 8, with all times Select the characteristic vector of image to make comparisons, calculate card side's distance, the facial image where choosing minimum value is used as matching face, formula It is as follows：
Wherein, H1、H2It is respectively that training sample improves LDP histograms and test sample improves LDP histograms, i represents image Subarea number, j represents the feature histogram number included in subregion,It is i-th jth of subregion in training sample Individual improvement LDP histograms,It is i-th j-th improvement LDP histogram of subregion, s in test sampleiTo consider the son In region after the Structure Comparison information of all pixels point, the overall weighted value of the subregion.
Compared with prior art, its remarkable advantage is the present invention：(1) the weighting treatment based on Structure Comparison information, face Recognition success rate is high；(2) calculated and judged using histogram, recognition speed is fast；(3) robustness to noise is more preferable, can To be more effectively applied to field of face identification.
Brief description of the drawings
Fig. 1 is schematic flow sheet of the present invention based on the face identification method for improving LDP.
Fig. 2 is 8 Kirsch operator schematic diagrames.
Fig. 3 is LDP algorithm coding process demonstration graphs.
Fig. 4 is to improve LDP algorithm coding process demonstration graphs.
Specific embodiment
The present invention is described in further detail with specific embodiment below in conjunction with the accompanying drawings.
With reference to Fig. 1, the present invention is based on the face identification method for improving LDP, it is characterised in that comprise the following steps：
Step 1, facial image to be identified is converted to the gray level image of formed objects specification；
Step 2, step 1 gained face gray level image is cut into the subregion of formed objects；
Step 3, using LDP methods are improved, the improvement LDP for extracting all subregions that face gray level image cuts out is special Value indicative, obtains improving LDP block images, specific as follows：
3 × 3 matrix-blocks centered on target pixel points and Kirsch operators are carried out into computing in initial processing stage, will Eight values for obtaining use m as the gray value of surrounding abutmentsiRepresent, i=0,1 ..., 7, maximum k during then this 8 are worth Individual value is entered as 1, and remaining 8-k value is entered as 0, so as to obtain an eight bit, is represented after being converted to decimal number The gray value of central pixel point, used as LDP characteristic values are improved, the process can be represented with equation below：
Wherein, xLDP′K () is the improvement LDP characteristic values of target pixel points, mkIt is miMiddle k maximum value；
Each sub-regions to cutting out carry out aforesaid operations respectively, change as each pixel in each sub-regions Enter LDP characteristic values.
Step 4, to each improvement LDP block image, incorporating Structure Comparison information carries out divided group, specific as follows：
Each sub-regions are proceeded as follows first in units of pixel：
Wherein, miRefer to that, for 3 × 3 matrixes centered on each pixel, 8 abutment points of the pixel are passed through The gray value obtained after Kirsch operator operations；Refer to miAverage value；
According to all of pixel in every sub-regions, the Structure Comparison information of the subregion after each segmentation is obtained, it is public Formula is as follows：
For a sub-regions, it is believed that each pixel is a least unit, xLDP′(r, c) represents the picture of r rows c row The improvement LDP characteristic values of vegetarian refreshments；srcRefer to a Structure Comparison information for the improvement LDP codings of pixel；siTo consider In the subregion after the Structure Comparison information of all pixels point, the overall weighted value of the subregion；M, N are the row of subregion pixel Number and columns.
Step 5, the weighting LDP characteristic values according to each sub-regions, the improvement LDP histograms for extracting the subregion are special Levy, formula is as follows：
Wherein, pillar height when H (τ) represents gray value for τ on the histogram, so as to show a sub-district with the histogram table The improvement LDP characteristic values of all pixels point in domain；xLDP′(r, c) represents the improvement LDP characteristic values of the pixel of r rows c row； M, N are the line number and columns of subregion pixel.
Step 6, by the improvement LDP histogram features of each sub-regions, is integrated into an entirety and is used to represent facial image；
Face images in known face database are carried out the treatment of above-mentioned steps 1~6 by step 7, draw known face The improvement LDP histogram features of all faces entirety in storehouse；
Step 8, extracts the improvement LDP histogram features vector of facial image to be identified, the feature with all candidate images Vector is made comparisons, and calculates card side's distance, and the facial image where choosing minimum value is used as matching face, and formula is as follows：
Wherein, H1、H2It is respectively that training sample improves LDP histograms and test sample improves LDP histograms, i represents image Subarea number, j represents the feature histogram number included in subregion,It is i-th jth of subregion in training sample Individual improvement LDP histograms,It is i-th j-th improvement LDP histogram of subregion, s in test sampleiTo consider the son In region after the Structure Comparison information of all pixels point, the overall weighted value of the subregion.
Embodiment 1
Face identification method based on improvement LDP proposed by the present invention, based on following steps：
1st, facial image to be identified is converted to the gray level image of formed objects specification；
According to the standard in Generic face storehouse, all people's face image is adjusted to 100 × 100 size.
The 2nd, facial image is cut into the subregion of formed objects；
Facial image can be cut into the subregion of 4 × 4 formed objects.
3rd, using the improvement LDP characteristic values for improving all subregions that LDP methods extraction facial image is cut out；
With target pixel points as the center of circle, R is radius to local binary patterns (LBP) algorithm, extracts P adjoining of impact point Point, then with the pixel value as threshold value, P binary coding of the point is calculated by formula (1) and formula (2)：
Here gcRepresent the pixel value of central point, gpRepresent with gcFor the center of circle, R are the P adjoining values of radius.In the method P=8, R=1 are chosen, i.e., 3 × 3 module LBP are used as the improved precondition of subsequent algorithm.
Local direction pattern (LDP) algorithm reference LBP coded systems, by contrasting pixel edge in different directions Return value is calculated eight-digit binary number coding, generally chooses Krisch masks to calculate all directions of pixel to edge return Value, from all directions to Krisch mask set { G0~G7As shown in Figure 2.
Eight adjacent pixels values centered on object pixel are obtained, obtains new to Kirsch mask computings with all directions respectively Eight adjacent pixels values, referred to as from all directions to edge return value | mi| (i=0,1 ..., 7).Then k maximum value number is chosen to make It is principal character, and 1 is entered as by this k, remaining (8-k) position is entered as 0, finally by this 8 number according to certain order group Into an eight bit, then it is converted into decimal number.As shown in Figure 3.The process can represent (its with equation below In | mk| it is | mi| middle k maximum value)：
Improvement LDP algorithms used in the present invention, in initial processing stage by 3 × 3 centered on target pixel points Matrix-block carries out computing with Kirsch operators, and eight for obtaining value uses m as the gray value of surrounding abutmentsi(i=0, 1 ..., 7) represent then by 8 values maximum k value be entered as 1, remaining (8-k) individual value is entered as 0, so as to obtain one Eight bit, is converted into after decimal number and the gray value of central pixel point is represented with this, and LDP features are improved as it Value, as shown in Figure 4.The process can be represented with equation below：
The result drawn during k=3 is taken in the method as required improvement LDP characteristic values.xLDP′K () refers to target picture The improvement LDP characteristic values (being now binary system) of vegetarian refreshments, mkIt is miMiddle k maximum value.
Each sub-piecemeal to cutting out carries out aforesaid operations respectively, changes as each pixel in each sub-regions Enter LDP characteristic values.
4th, to each improvement LDP block image, incorporate Structure Comparison information to carry out divided group；
In coding, LDP codings contain partial structurtes distributed intelligence, by low comparison structure and the pixel of comparison structure high Point distribution is put on an equal footing.In order to protrude influence of the different human face regions for face recognition result, protoplast's face image is split This 4 × 4 block images for going out incorporate Structure Comparison information to carry out divided group.4 × 4 piecemeals obtained after segmentation, to every One sub-regions are proceeded as follows first in units of pixel：
Wherein miRefer to that, for 3 × 3 matrixes centered on each pixel, 8 abutment points of the pixel are passed through The gray value obtained after Kirsch operator operations.
The subregion after each segmentation is obtained by all of pixel in every sub-regions using formula is calculated as below Structure Comparison information：
For a sub-regions, it is believed that each pixel is a least unit, xLDP′(r, c) represents the picture of r rows c row The improvement LDP characteristic values of vegetarian refreshments.srcRefer to a Structure Comparison information for the improvement LDP codings of pixel, siTo being considered as In the piecemeal after the Structure Comparison information of all pixels point, the overall weighted value of the piecemeal, due to taking 4 × 4 segmentations, in we M=25 in method, N=25.
5th, using by step 3) with step 4) after the weighting LDP eigenvalue extractings of each piecemeal piecemeal that obtains Improve LDP histogram features；
Concrete operations mode can be expressed as following formula：
Pillar height on the histogram when H (τ) represents that gray value is τ, in showing a word piecemeal with the histogram table The improvement LDP characteristic values of all pixels point.
6th, the improvement LDP histogram features of (splicing) each piecemeal are integrated for an entirety is used to represent facial image；
7th, above-mentioned steps 1 are carried out to the face images in known face database) -6), draw the institute in known face database There are the overall improvement LDP histogram features of face；
8th, the vectorial characteristic vector with all candidate images of improvement LDP histogram features for extracting facial image to be identified is made Compare, calculate card side's distance, the facial image where choosing minimum value is used as matching face；
Card side's distance is chosen herein to calculate the similarity between image two-by-two.When carrying out face match cognization, extract and survey The vectorial characteristic vector with all candidate images of LDP histogram features for attempting picture is made comparisons, and calculates card side's distance, chooses minimum Used as matching face, the operating process can be expressed as following formula to facial image where value：
Wherein H1、H2It is respectively that training sample improves LDP histograms and test sample improves LDP histograms, i represents image Piecemeal number, j represents the feature histogram number (j=1 in this method) included in segmented areas.
The present invention is based on C Plus Plus, is utilized respectively YALE standard faces storehouse, ORL standard faces storehouse and is tested, correct to know Rate is not respectively 92.59% and 81.42%.

Claims (5)

1. it is a kind of based on the face identification method for improving LDP, it is characterised in that to comprise the following steps：
Step 1, facial image to be identified is converted to the gray level image of formed objects specification；
Step 2, step 1 gained face gray level image is cut into the subregion of formed objects；
Step 3, using LDP methods are improved, extracts the improvement LDP features of all subregions that face gray level image cuts out Value, obtains improving LDP block images；
Step 4, to each improvement LDP block image, incorporating Structure Comparison information carries out divided group；
Step 5, the weighting LDP characteristic values according to each sub-regions extract the improvement LDP histogram features of the subregion；
Step 6, by the improvement LDP histogram features of each sub-regions, is integrated into an entirety and is used to represent facial image；
Face images in known face database are carried out the treatment of above-mentioned steps 1~6 by step 7, in drawing known face database The overall improvement LDP histogram features of all faces；
Step 8, extracts the improvement LDP histogram features vector of facial image to be identified, the characteristic vector with all candidate images Make comparisons, calculate card side's distance, the facial image where choosing minimum value is used as matching face.
2. it is according to claim 1 based on the face identification method for improving LDP, it is characterised in that to be used described in step 3 and changed Enter LDP methods, extract the improvement LDP characteristic values of all subregions that face gray level image cuts out, obtain improving LDP piecemeals Image, it is specific as follows：
3 × 3 matrix-blocks centered on target pixel points and Kirsch operators are carried out into computing in initial processing stage, will be obtained Eight value as surrounding abutments gray value, use miRepresent, i=0,1 ..., 7, maximum k value during then this 8 are worth 1 is entered as, remaining 8-k value is entered as 0, so as to obtain an eight bit, center is represented after being converted to decimal number The gray value of pixel, used as LDP characteristic values are improved, the process can be represented with equation below：
x LDP ′ ( k ) = Σ i = 0 7 b i ( m i - m k ) × 2 i
b i ( a ) = 0 , a < 0 1 , a &GreaterEqual; 0
Wherein, xLDP′K () is the improvement LDP characteristic values of target pixel points, mkIt is miMiddle k maximum value；
Each sub-regions to cutting out carry out aforesaid operations respectively, used as the improvement of each pixel in each sub-regions LDP characteristic values.
3. it is according to claim 1 based on the face identification method for improving LDP, it is characterised in that to each described in step 4 LDP block images are improved, incorporating Structure Comparison information carries out divided group, specific as follows：
Each sub-regions are proceeded as follows first in units of pixel：
m &OverBar; = 1 8 &Sigma; i = 0 7 m i
Wherein, miRefer to for 3 × 3 matrixes centered on each pixel, 8 abutment points of the pixel are by Kirsch The gray value obtained after operator operation；Refer to miAverage value；
According to all of pixel in every sub-regions, the Structure Comparison information of the subregion after each segmentation is obtained, formula is such as Under：
s r c ( x LDP &prime; ( r , c ) ) = 1 8 &Sigma; i = 0 7 ( m i - m &OverBar; ) 2
s i = 1 H &times; N &Sigma; r = 1 M &Sigma; c = 1 N s r c ( x LDP &prime; ( r , c ) )
For a sub-regions, it is believed that each pixel is a least unit, xLDP′(r, c) represents the pixel of r rows c row Improvement LDP characteristic values；srcRefer to a Structure Comparison information for the improvement LDP codings of pixel；siTo consider the son In region after the Structure Comparison information of all pixels point, the overall weighted value of the subregion；M, N for subregion pixel line number with Columns.
4. it is according to claim 1 based on the face identification method for improving LDP, it is characterised in that according to every described in step 5 The weighting LDP characteristic values of one sub-regions, extract the improvement LDP histogram features of the subregion, and formula is as follows：
H ( &tau; ) = &Sigma; r = 1 M &Sigma; c = 1 N f ( x LDP &prime; ( r , c ) , &tau; )
f ( a , &tau; ) = 1 , a = &tau; 0 , a &NotEqual; &tau;
Wherein, pillar height on the histogram when H (τ) represents that gray value is τ, in showing a sub-regions with the histogram table The improvement LDP characteristic values of all pixels point；xLDP′(r, c) represents the improvement LDP characteristic values of the pixel of r rows c row；M、N It is the line number and columns of subregion pixel.
5. it is according to claim 1 based on the face identification method for improving LDP, it is characterised in that to be extracted described in step 8 and treated Recognize the improvement LDP histogram features vector of facial image, the characteristic vector with all candidate images makes comparisons, calculate card side away from From the facial image where choosing minimum value is used as matching face, and formula is as follows：
( H 1 , H 2 ) = &Sigma; i , j s i ( H i , j 1 - H i , j 2 ) 2 ( H i , j 1 + H i , j 2 )
Wherein, H1、H2It is respectively that training sample improves LDP histograms and test sample improves LDP histograms, i represents image subsection Field Number, j represents the feature histogram number included in subregion,For i-th j-th of subregion changes in training sample Enter LDP histograms,It is i-th j-th improvement LDP histogram of subregion, s in test sampleiTo consider the subregion After the Structure Comparison information of middle all pixels point, the overall weighted value of the subregion.
CN201611062329.6A 2016-11-25 2016-11-25 A kind of face identification method based on improvement LDP Pending CN106778529A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201611062329.6A CN106778529A (en) 2016-11-25 2016-11-25 A kind of face identification method based on improvement LDP

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201611062329.6A CN106778529A (en) 2016-11-25 2016-11-25 A kind of face identification method based on improvement LDP

Publications (1)

Publication Number Publication Date
CN106778529A true CN106778529A (en) 2017-05-31

Family

ID=58901776

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201611062329.6A Pending CN106778529A (en) 2016-11-25 2016-11-25 A kind of face identification method based on improvement LDP

Country Status (1)

Country Link
CN (1) CN106778529A (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107273855A (en) * 2017-06-17 2017-10-20 南京理工大学 Illumination invariant face identification method based on the linear Local Operator of higher derivative
CN107578005A (en) * 2017-09-01 2018-01-12 宜宾学院 A kind of Complex Wavelet Transform domain LBP face identification methods
CN107886116A (en) * 2017-10-30 2018-04-06 国家新闻出版广电总局广播科学研究院 The LOMO3D feature extracting methods identified again for video data pedestrian
CN108197529A (en) * 2017-11-27 2018-06-22 重庆邮电大学 A kind of human facial feature extraction method for merging DLDP and sobel
CN110348284A (en) * 2019-05-20 2019-10-18 天津科技大学 Face recognition features' extracting method is blocked based on local direction texture pattern
CN115798004A (en) * 2022-11-27 2023-03-14 深圳市巨龙创视科技有限公司 Face card punching method and device based on local area, electronic equipment and medium
CN116052135A (en) * 2023-04-03 2023-05-02 吉林大学 Foggy day traffic sign recognition method based on texture features

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120219224A1 (en) * 2011-02-28 2012-08-30 Yuanyuan Ding Local Difference Pattern Based Local Background Modeling For Object Detection
CN104915626A (en) * 2014-03-11 2015-09-16 重庆邮电大学 Face identification method and apparatus
CN105069447A (en) * 2015-09-23 2015-11-18 河北工业大学 Facial expression identification method
CN105956552A (en) * 2016-04-29 2016-09-21 中国人民解放军国防科学技术大学 Face black list monitoring method

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120219224A1 (en) * 2011-02-28 2012-08-30 Yuanyuan Ding Local Difference Pattern Based Local Background Modeling For Object Detection
CN104915626A (en) * 2014-03-11 2015-09-16 重庆邮电大学 Face identification method and apparatus
CN105069447A (en) * 2015-09-23 2015-11-18 河北工业大学 Facial expression identification method
CN105956552A (en) * 2016-04-29 2016-09-21 中国人民解放军国防科学技术大学 Face black list monitoring method

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
李杰: "基于改进 LDP 的人脸识别", 《计算机应用与软件》 *

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107273855A (en) * 2017-06-17 2017-10-20 南京理工大学 Illumination invariant face identification method based on the linear Local Operator of higher derivative
CN107578005A (en) * 2017-09-01 2018-01-12 宜宾学院 A kind of Complex Wavelet Transform domain LBP face identification methods
CN107886116A (en) * 2017-10-30 2018-04-06 国家新闻出版广电总局广播科学研究院 The LOMO3D feature extracting methods identified again for video data pedestrian
CN107886116B (en) * 2017-10-30 2021-03-26 国家新闻出版广电总局广播科学研究院 LOMO3D feature extraction method for pedestrian re-identification of video data
CN108197529A (en) * 2017-11-27 2018-06-22 重庆邮电大学 A kind of human facial feature extraction method for merging DLDP and sobel
CN108197529B (en) * 2017-11-27 2021-09-10 重庆邮电大学 Expression feature extraction method integrating DLDP and sobel
CN110348284A (en) * 2019-05-20 2019-10-18 天津科技大学 Face recognition features' extracting method is blocked based on local direction texture pattern
CN115798004A (en) * 2022-11-27 2023-03-14 深圳市巨龙创视科技有限公司 Face card punching method and device based on local area, electronic equipment and medium
CN116052135A (en) * 2023-04-03 2023-05-02 吉林大学 Foggy day traffic sign recognition method based on texture features
CN116052135B (en) * 2023-04-03 2023-07-11 吉林大学 Foggy day traffic sign recognition method based on texture features

Similar Documents

Publication Publication Date Title
CN106778529A (en) A kind of face identification method based on improvement LDP
CN107563385B (en) License plate character recognition method based on depth convolution production confrontation network
CN105095856B (en) Face identification method is blocked based on mask
CN105069447B (en) A kind of recognition methods of human face expression
CN102254188B (en) Palmprint recognizing method and device
CN102521561B (en) Face identification method on basis of multi-scale weber local features and hierarchical decision fusion
CN105139039A (en) Method for recognizing human face micro-expressions in video sequence
CN102663400B (en) LBP (length between perpendiculars) characteristic extraction method combined with preprocessing
CN104298981A (en) Face microexpression recognition method
CN105956560A (en) Vehicle model identification method based on pooling multi-scale depth convolution characteristics
CN105160317A (en) Pedestrian gender identification method based on regional blocks
CN108564040B (en) Fingerprint activity detection method based on deep convolution characteristics
CN101350063A (en) Method and apparatus for locating human face characteristic point
CN104616664A (en) Method for recognizing audio based on spectrogram significance test
CN106960181A (en) A kind of pedestrian&#39;s attribute recognition approach based on RGBD data
CN103279770B (en) Based on the person&#39;s handwriting recognition methods of stroke fragment and contour feature
CN104778472B (en) Human face expression feature extracting method
CN106709518A (en) Android platform-based blind way recognition system
CN105608441A (en) Vehicle type identification method and system
CN103729625A (en) Face identification method
CN104408711A (en) Multi-scale region fusion-based salient region detection method
CN104966054A (en) Weak and small object detection method in visible image of unmanned plane
Afakh et al. Aksara jawa text detection in scene images using convolutional neural network
CN104361319A (en) Fake fingerprint detection method based on SVM-RFE (support vector machine-recursive feature elimination)
CN104966305A (en) Foreground detection method based on motion vector division

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: 20170531