CN113239761B - Face recognition method, device and storage medium - Google Patents
Face recognition method, device and storage medium Download PDFInfo
- Publication number
- CN113239761B CN113239761B CN202110472915.2A CN202110472915A CN113239761B CN 113239761 B CN113239761 B CN 113239761B CN 202110472915 A CN202110472915 A CN 202110472915A CN 113239761 B CN113239761 B CN 113239761B
- Authority
- CN
- China
- Prior art keywords
- particle
- sub
- optimized
- face
- block
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 49
- 239000013598 vector Substances 0.000 claims abstract description 89
- 238000012549 training Methods 0.000 claims abstract description 85
- 238000012706 support-vector machine Methods 0.000 claims abstract description 75
- 238000012360 testing method Methods 0.000 claims abstract description 54
- 238000007781 pre-processing Methods 0.000 claims abstract description 14
- 239000002245 particle Substances 0.000 claims description 206
- 238000005457 optimization Methods 0.000 claims description 17
- 238000012545 processing Methods 0.000 claims description 16
- 230000001815 facial effect Effects 0.000 claims description 15
- 210000000056 organ Anatomy 0.000 claims description 13
- 230000009977 dual effect Effects 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 7
- 238000012937 correction Methods 0.000 claims description 7
- 210000000697 sensory organ Anatomy 0.000 claims description 7
- 238000005498 polishing Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 53
- 238000001514 detection method Methods 0.000 description 9
- 230000011218 segmentation Effects 0.000 description 5
- 230000006978 adaptation Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000036544 posture Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
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
- G06V40/172—Classification, e.g. identification
-
- 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
-
- 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
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/245—Classification techniques relating to the decision surface
- G06F18/2453—Classification techniques relating to the decision surface non-linear, e.g. polynomial classifier
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/24—Aligning, centring, orientation detection or correction of the image
- G06V10/243—Aligning, centring, orientation detection or correction of the image by compensating for image skew or non-uniform image deformations
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Multimedia (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Oral & Maxillofacial Surgery (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Human Computer Interaction (AREA)
- Biomedical Technology (AREA)
- Nonlinear Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a face recognition method, which comprises the steps of obtaining a nonlinear support vector machine model, training the nonlinear support vector machine model according to a pre-obtained training set and a pre-obtained testing set to obtain an optimized nonlinear support vector machine model and an optimized sub-block channel number, preprocessing the face image area to be recognized according to the optimized sub-block channel number to obtain a face feature vector, wherein the dimension of the face feature vector is equal to the product of the optimized sub-block channel number and the preset sub-block number, and finally inputting the face feature vector into the optimized nonlinear support vector machine model to obtain a face recognition result, wherein the face recognition result can consider the influence of random and variable interference of a background environment where a face is positioned on face polishing, and the face recognition accuracy and the face recognition efficiency are improved. The invention also correspondingly provides a face recognition device and a storage medium.
Description
Technical Field
The present invention relates to the field of face recognition technologies, and in particular, to a face detection method, a face detection device, and a computer readable storage medium.
Background
Face recognition is a technology for carrying out identity recognition based on facial feature information of a person, and the face feature is extracted and compared with feature information stored in a database to obtain a comparison result so as to further carry out identity recognition. At present, the accuracy of face recognition is inaccurate due to age change of personnel, change of dressing postures and the like. In particular, face recognition and extraction are difficult in places with high crowd concentration and changeable environments, the existing face recognition method cannot clearly separate the faces of target persons, and the detection efficiency and accuracy are low.
Disclosure of Invention
The embodiment of the invention provides a face detection method, a face detection device and a computer readable storage medium, which can effectively improve the efficiency and accuracy of face recognition.
A first aspect of an embodiment of the present invention provides a face recognition method, including:
acquiring a nonlinear support vector machine model, and training the nonlinear support vector machine model according to a pre-acquired training set and a pre-acquired testing set to obtain an optimized nonlinear support vector machine model and an optimized sub-block channel number; the training set and the testing set comprise a plurality of face feature vector samples with different sub-block channel numbers;
acquiring a face image to be recognized;
preprocessing the face image area to be identified according to the optimized number of sub-block channels to obtain a face feature vector, wherein the dimension of the face feature vector is equal to the product of the optimized number of sub-block channels and the preset number of sub-blocks;
and inputting the face feature vector into the optimized nonlinear support vector machine model to obtain a face recognition result.
Preferably, the nonlinear support vector machine model specifically includes:
s.t.y i (w T x i +b)≥1-ξ i ,ξ i ≥0,i＝1,2,…,n，
the optimization problem of the nonlinear support vector machine model is converted into the following dual problem:
wherein x is i Is the ith sample in samples of given scale n, w is the weight vector, y i Is x i The value of 1 or-1, xi i For non-negative relaxation variables, C is an adjustable penalty parameter, n is the number of samples, L (alpha) represents a Lagrangian function, alpha i Represents Lagrange multiplier, alpha j Lagrangian multiplier, x representing dual j Is the j-th sample of the given n-scale samples that is dual to the i-th sample, y j Is x j Is a label of (a).
Preferably, the preprocessing is performed on the face image area to be identified according to the optimized number of sub-block channels to obtain a face feature vector, which specifically includes:
after inclination correction is carried out on a face image area to be identified, binarization processing is carried out on the face area by adopting a self-adaptive threshold method;
dividing the face image to be recognized after binarization processing into n according to the positions occupied by facial image five-sense organ areas according to the number of preset sub-blocks 0 Sub-blocks, where n 0 The number of the sub-blocks is preset;
dividing each sub-block into m channels according to the optimized number of sub-block channels to obtain N channels, wherein m is equal to the optimized number of sub-block channels, and N is equal to the total number of channels;
and calculating the direction gradient value of each channel, and obtaining the face feature vector according to the direction gradient value of each channel.
Preferably, the obtaining the nonlinear support vector machine model, and training the nonlinear support vector machine model according to a pre-obtained training set and a pre-obtained testing set to obtain an optimized nonlinear support vector machine model and an optimized sub-block channel number specifically includes:
according to a pre-acquired training set and a pre-acquired testing set, optimizing and training a punishment parameter and a kernel function parameter in the establishment process of the face feature vector by adopting a particle population algorithm to obtain an optimized punishment parameter, an optimized kernel function parameter and an optimized sub-block channel number;
and obtaining an optimized nonlinear support vector machine model according to the optimized penalty parameter and the optimized kernel function parameter.
Preferably, the particle population algorithm is adopted to optimize the number of sub-block channels in the process of building the face feature vector according to the pre-acquired training set and test set, and the penalty parameter and the kernel function parameter in the process of building the nonlinear support vector machine model are optimized and trained to obtain the optimized penalty parameter, the optimized kernel function parameter and the optimized number of sub-block channels, which specifically comprises:
particle determination: combining the punishment parameters and the kernel function parameters in the establishment process of the nonlinear support vector machine model into particles;
particle population acquisition: randomly generating different combinations of the sub-block channel number, the penalty parameter and the kernel function parameter to obtain a particle population, wherein each particle is expressed as: x is X i ＝(m i ，C i ，δ i )，X i Represents the ith particle, m i Represents the number of sub-block channels corresponding to the ith particle, C i Represents penalty parameter, delta, corresponding to the ith particle i Representing kernel function parameters corresponding to the ith particle;
and a particle fitness calculating step: acquiring multiple sets of training sets S i And multiple sets of test sets T i Wherein the training set S i The dimension of the face feature vector sample in the test set T is set according to the number of sub-block channels corresponding to the ith particle i The dimension in the face feature vector sample is set according to the number of sub-block channels corresponding to the ith particle;
using training sets S i Penalty parameter C for ith particle i Kernel function parameter delta corresponding to the ith particle i Training the nonlinear support vector machine model to obtain a five-sense organ recognition model M i ；
The five sense organs identification model M i Acting on test set T i Obtaining the identification accuracy of the ith particle, and taking the identification accuracy of the ith particle as the fitness of the ith particle;
a global optimal particle determining step: taking the particle with the highest adaptability in all particles as a global optimal particle;
particle updating: for each particle, updating the position and speed of each particle according to the fitness of each particle to obtain a new particle;
an optimized parameter obtaining step: judging whether a preset condition is met, if so, acquiring the global optimal particles to obtain optimized punishment parameters, optimized kernel function parameters and optimized sub-block channel numbers; otherwise, returning to the particle fitness calculating step.
A second aspect of an embodiment of the present invention provides a face recognition apparatus, including:
the optimization parameter and model acquisition module is used for acquiring a nonlinear support vector machine model, training the nonlinear support vector machine model according to a pre-acquired training set and a pre-acquired testing set, and obtaining an optimized nonlinear support vector machine model and an optimized sub-block channel number; the training set and the testing set comprise a plurality of face feature vector samples with different sub-block channel numbers;
the face image acquisition module is used for acquiring a face image to be identified;
the optimized face feature vector acquisition module is used for preprocessing the face image area to be recognized according to the optimized number of sub-block channels to obtain a face feature vector, wherein the dimension of the face feature vector is equal to the product of the optimized number of sub-block channels and the preset number of sub-blocks;
and the face recognition result acquisition module is used for inputting the face feature vector into the optimized nonlinear support vector machine model to obtain a face recognition result.
Preferably, the optimized face feature vector obtaining module is specifically configured to:
after inclination correction is carried out on a face image area to be identified, binarization processing is carried out on the face area by adopting a self-adaptive threshold method;
dividing the face image to be recognized after binarization processing into n according to the positions occupied by facial image five-sense organ areas according to the number of preset sub-blocks 0 Sub-blocks, where n 0 The number of the sub-blocks is preset;
dividing each sub-block into m channels according to the optimized number of sub-block channels to obtain N channels, wherein m is equal to the optimized number of sub-block channels, and N is equal to the total number of channels;
and calculating the direction gradient value of each channel, and obtaining the face feature vector according to the direction gradient value of each channel.
Preferably, the parameter optimization and model acquisition module includes:
the optimization parameter acquisition unit is used for optimizing and training the punishment parameters and the kernel function parameters in the establishment process of the face feature vector by adopting a particle population algorithm according to a training set and a testing set which are acquired in advance to obtain an optimized punishment parameter, an optimized kernel function parameter and an optimized sub-block channel number;
and the optimization model acquisition unit is used for obtaining an optimized nonlinear support vector machine model according to the optimized penalty parameter and the optimized kernel function parameter.
Preferably, the optimization parameter obtaining unit is specifically configured to perform:
particle determination: combining the punishment parameters and the kernel function parameters in the establishment process of the nonlinear support vector machine model into particles;
particle population acquisition: random generationDifferent combinations of the sub-block channel number, the penalty parameter and the kernel parameter result in a population of particles, wherein each particle is represented as: x is X i ＝(m i ，C i ，δ i )，X i Represents the ith particle, m i Represents the number of sub-block channels corresponding to the ith particle, C i Represents penalty parameter, delta, corresponding to the ith particle i Representing kernel function parameters corresponding to the ith particle;
and a particle fitness calculating step: acquiring multiple sets of training sets S i And multiple sets of test sets T i Wherein the training set S i The dimension of the face feature vector sample in the test set T is set according to the number of sub-block channels corresponding to the ith particle i The dimension in the face feature vector sample is set according to the number of sub-block channels corresponding to the ith particle;
using training sets S i Penalty parameter C for ith particle i Kernel function parameter delta corresponding to the ith particle i Training the nonlinear support vector machine model to obtain a five-sense organ recognition model M i ；
The five sense organs identification model M i Acting on test set T i Obtaining the identification accuracy of the ith particle, and taking the identification accuracy of the ith particle as the fitness of the ith particle;
a global optimal particle determining step: taking the particle with the highest adaptability in all particles as a global optimal particle;
particle updating: updating the position and the speed of each particle according to the fitness of each particle for each particle to obtain a new particle, and taking the new particle as the particle in the particle population;
an optimized parameter obtaining step: judging whether a preset condition is met, if so, acquiring the global optimal particles to obtain optimized punishment parameters, optimized kernel function parameters and optimized sub-block channel numbers; otherwise, returning to the particle fitness calculating step.
A third aspect of the embodiment of the present invention provides a computer readable storage medium, where the computer readable storage medium includes a stored computer program, where the computer program when executed controls a device where the computer readable storage medium is located to execute the face recognition method according to the foregoing embodiment.
Compared with the prior art, the face recognition method provided by the embodiment of the invention comprises the steps of obtaining a nonlinear support vector machine model, training the nonlinear support vector machine model according to a pre-obtained training set and a pre-obtained testing set to obtain an optimized nonlinear support vector machine model and an optimized sub-block channel number, preprocessing the face image area to be recognized according to the optimized sub-block channel number to obtain a face feature vector, wherein the dimension of the face feature vector is equal to the product of the optimized sub-block channel number and the preset sub-block number, and finally inputting the face feature vector into the optimized nonlinear support vector machine model to obtain a face recognition result, wherein the influence of uneven face polishing and random and variable interference of a background environment where a face is located can be considered, and the face recognition accuracy and the face recognition efficiency of the face are improved. In addition, the embodiment of the invention also correspondingly provides a face recognition device and a computer readable storage medium.
Drawings
Fig. 1 is a flowchart of a face detection method provided in an embodiment of the present invention;
fig. 2 is a block diagram of a face detection apparatus according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, a flowchart of a face detection method according to an embodiment of the present invention is shown.
The face detection method provided by the embodiment of the invention comprises the steps of S11 to S14:
step S11, a nonlinear support vector machine model is obtained, and training is carried out on the nonlinear support vector machine model according to a training set and a testing set which are obtained in advance, so that an optimized nonlinear support vector machine model and an optimized sub-block channel number are obtained; the training set and the testing set comprise a plurality of face feature vector samples with different sub-block channel numbers.
Specifically, the support vector machine (Support Vector Machine, SVM) is a generalized linear classifier (generalized linear classifier) for binary classification of data in a supervised learning (supervised learning) manner, and is solved by maximizing the classification interval of hyperplanes and converting the optimal classification surface problem into a dual problem by a lagrangian function. For the linear inseparable problem, the original sample in the low-dimensional space is converted into the linear inseparable sample in the high-dimensional characteristic space through kernel function mapping, so that the correct identification of the linear inseparable sample is realized. In an embodiment of the invention, by introducing a non-negative relaxation variable ζ i And penalty coefficient C, can obtain solving the support vector machine model of the nonlinear separable:
the optimization problem of the nonlinear support vector machine model is converted into the following dual problem:
wherein x is i Is the ith sample in samples of given scale n, w is the weight vector, y i Is x i The value of 1 or-1, xi i For non-negative relaxation variables, C is an adjustable penalty parameter, n is the number of samples, L (alpha) represents a Lagrangian function, alpha i Represents Lagrange multiplier, alpha j Lagrangian multiplier, x representing dual j Is of a given scale n and of sampleThe j-th sample of the pair of i samples, y j Is x j Is a label of (a).
In the embodiment of the invention, the training set and the test set are face feature vectors, and the face feature vectors can be used for acquiring a face image to be trained when the training set and the test set are implemented, performing inclination correction processing on a face region of the face image to be trained, performing binarization segmentation to obtain a binarized face image, and dividing the binarized face image into n according to the position occupied by a facial feature region 0 Sub-blocks; for each sub-block j, the center of the sub-block j is taken as the origin of coordinates, the sub-block j is equally divided into m channels (i.e. m symmetric regions), and then the directional gradient value of the channel k is calculated. Thus, the channel k in the sub-block j of the facial image facial region corresponds to a gradient value alpha jk ，1≤j≤n 0 K is more than or equal to 1 and less than or equal to m, so that the face feature vector of the face image to be trained is obtained asAnd taking the face feature vector as a sample. It will be appreciated that the dimension of the face feature vector is equal to n 0 * m. In order to facilitate the subsequent training of the number of sub-block channels (i.e. the number of channels contained in each sub-block), a large number of face feature vectors with different sub-block channel numbers need to be obtained, so as to obtain a large number of face feature vector samples, and further obtain the training set. It will be appreciated that the test set is also obtained by means of the training set, except that each sample in the training set carries a known tag, and the tag of each sample in the test set is unknown.
In one possible case, in the above-mentioned facial region segmentation processing of the binarized face image, the peer-to-peer segmentation may be directly performed directly based on the facial region of the binarized face image to obtain m' channels, i.e., n in this case 0 ＝1。
In another possible case, in the above-mentioned facial region segmentation processing of the binarized face image, the facial region of the binarized face image may be first divided into n 0 Sub-block (n) 0 >1) At this time, each sub-block is equally divided into m' channels. It will be appreciated that in this way, the amount of computation in the subsequent training process can be effectively reduced, and thus, embodiments of the present invention preferably employ this way.
In the embodiment of the invention, the influence of illumination intensity on the facial segmentation of the face image is considered, so that the processed face image is closer to the original effect. In the embodiment of the invention, in the preprocessing process of the face image, the binarization processing of the face image can adopt the following self-adaptive threshold scheme:
wherein t represents the total number of pixel points with gray values larger than 125 in the face image, and s represents the total number of pixels in the face image area.
Step S12, a face image to be recognized is obtained.
And step S13, preprocessing the face image area to be recognized according to the optimized number of sub-block channels to obtain a face feature vector, wherein the dimension of the face feature vector is equal to the product of the optimized number of sub-block channels and the preset number of sub-blocks.
And S14, inputting the face feature vector into the optimized nonlinear support vector machine model to obtain a face recognition result.
In specific implementation, a plurality of face images M can be input at one time k And k=1, 2 and …, preprocessing each face image according to the number of sub-block channels to be optimized to obtain a corresponding face feature vector, and inputting the corresponding face feature vector into the optimized nonlinear support vector machine model to obtain a corresponding face recognition result.
The face recognition method provided by the embodiment of the invention comprises the steps of obtaining a nonlinear support vector machine model, training the nonlinear support vector machine model according to a pre-obtained training set and a pre-obtained testing set to obtain an optimized nonlinear support vector machine model and an optimized sub-block channel number, preprocessing the face image area to be recognized according to the optimized sub-block channel number to obtain a face feature vector, wherein the dimension of the face feature vector is equal to the product of the optimized sub-block channel number and the preset sub-block number, and finally inputting the face feature vector into the optimized nonlinear support vector machine model to obtain a face recognition result, wherein the face recognition result can be considered to account the influence of uneven face polishing and random and changeable interference of a background environment where a face is located on the face, and the face recognition accuracy and the face recognition efficiency are improved.
In an optional implementation manner, the step S1 "obtaining a nonlinear support vector machine model, and training the nonlinear support vector machine model according to a pre-obtained training set and a pre-obtained testing set to obtain an optimized nonlinear support vector machine model and an optimized sub-block channel number", which specifically includes:
according to a pre-acquired training set and a pre-acquired testing set, optimizing and training a punishment parameter and a kernel function parameter in the establishment process of the face feature vector by adopting a particle population algorithm to obtain an optimized punishment parameter, an optimized kernel function parameter and an optimized sub-block channel number;
and obtaining an optimized nonlinear support vector machine model according to the optimized penalty parameter and the optimized kernel function parameter.
In the embodiment of the invention, the face is identified by adopting the face feature vector and the nonlinear support vector machine, so that the number of sub-block channels in the process of establishing the face feature vector, punishment parameters and kernel function parameters in the process of establishing the nonlinear support vector machine model can influence the accuracy of face identification. Furthermore, the number of sub-block channels, the penalty parameter and the kernel function parameter need to be optimized to obtain an optimized penalty parameter, an optimized kernel function parameter and an optimized number of sub-block channels.
Further, according to a pre-acquired training set and test set, a particle population algorithm is adopted to optimize the number of sub-block channels in the process of building the face feature vector, and the penalty parameter and the kernel function parameter in the process of building the nonlinear support vector machine model are optimized and trained to obtain an optimized penalty parameter, an optimized kernel function parameter and an optimized sub-block channel number, which specifically comprises:
particle determination: combining the punishment parameters and the kernel function parameters in the establishment process of the nonlinear support vector machine model into particles;
particle population acquisition: randomly generating different combinations of the sub-block channel number, the penalty parameter and the kernel function parameter to obtain a particle population, wherein each particle is expressed as: x is X i ＝(m i ，C i ，δ i )，X I Represents the ith particle, m i Represents the number of sub-block channels corresponding to the ith particle, C i Represents penalty parameter, delta, corresponding to the ith particle i Representing kernel function parameters corresponding to the ith particle;
and a particle fitness calculating step: acquiring multiple sets of training sets S I And multiple sets of test sets T i Wherein the training set S I The dimension of the face feature vector sample in the test set T is set according to the number of sub-block channels corresponding to the ith particle i The dimension in the face feature vector sample is set according to the number of sub-block channels corresponding to the ith particle;
using training sets S i Penalty parameter C for ith particle i Kernel function parameter delta corresponding to the ith particle i Training the nonlinear support vector machine model to obtain a five-sense organ recognition model M i ；
The five sense organs identification model M i Acting on test set T i Obtaining the identification accuracy of the ith particle, and taking the identification accuracy of the ith particle as the fitness of the ith particle;
a global optimal particle determining step: taking the particle with the highest adaptability in all particles as a global optimal particle;
particle updating: for each particleX i According to each particle X i Updates the position and velocity of each particle to obtain a new particle X i ' and particle X i Updated to new particle X i '；
An optimized parameter obtaining step: judging whether a preset condition is met, if so, taking the global optimal particles as optimal particles to obtain optimized punishment parameters, optimized kernel function parameters and optimized sub-block channel numbers; otherwise, returning to the particle fitness calculating step.
In specific implementation, the value ranges of three parameters to be optimized, namely, the value range of each parameter in the sub-block channel number m, the penalty parameter C and the kernel function parameter delta, can be preset, and the sub-block channel number m, the penalty parameter C and the kernel function parameter delta are randomly combined after being randomly valued, so that an initial particle population is obtained. Then training the non-support vector machine model by acquiring a pre-selected generated training set and a test set, wherein the training set and the test set comprise a plurality of face feature vector samples with different sub-block channel numbers, and in order to better test the effect of the recognition accuracy of the different sub-block channel numbers input into different models, the training set is divided into a plurality of groups of training sets, the dimension of the face feature vector sample of each group of training set is set according to the sub-block channel number corresponding to the ith particle, and the dimension of the face feature vector sample in the same group of training set is equal to m i *n 0 And utilize training set S i Penalty parameter C for ith particle i Kernel function parameter delta corresponding to the ith particle i Training the nonlinear support vector machine model to obtain a five-sense organ recognition model M i Further enable the official to recognize the model M i Acting on test set T i Obtaining particles X i The accuracy of face recognition is obtained, and the number of channels in the sub-block is m i The face feature vector of (2) is input into penalty parameter C i The kernel function parameter is delta i Is continuously optimized by an algorithm to find out the optimal particles (m * ，C * ，δ * ) Thereby obtaining the optimized sub-block channel number, the optimized punishment parameter and the optimized kernel function parameter, further obtaining an optimized nonlinear support vector machine model, and further according to the training set S i And test set T i The recognition accuracy is obtained from the following formula:
wherein r (m, C, delta) is the recognition accuracy obtained by the classifier which is obtained by the social media optimization and the training set and is used for the test set under the given m, C and delta of the optimized nonlinear support vector machine model.
Illustratively, the following are the detailed steps of employing a particle swarm algorithm for an embodiment of the present invention:
step S1, setting the punishment parameters of the number of sub-block channels and the value range of kernel function parameters, wherein m is [1, 36], C is [10-3, 10];
step S2, acquiring a face image to be trained and a face image to be tested;
step S3, setting iteration times, enabling the current iteration times T to be equal to 1, and randomly generating N particles: x is X i ＝(m i ，C i ，δ i )，1≤i≤N；
Step S4, for each particle X i Calculate each particle X i Degree of adaptation F (X) i )：
Step S4.1 by rounding the number of sub-block channels m i According to the number m of sub-block channels i Preprocessing a face image to be trained and a face image to be tested to obtain a face feature vector of the face image to be trained and a face image feature vector to be trained, and further obtaining a training set S i And test set T i ；
Step S4.2, using C i 、δ i And training set S i Training the nonlinear support vector machine model to obtain a five-sense organ recognition model M i Acting on test set T i And the recognition accuracy thus obtained is taken as particle X i Degree of adaptation F (X) i )；
Step S5, taking the particle with the highest adaptability among the N particles as a global optimal particle gb;
step S6, for each particle, updating according to the following formula, and setting the position x 'of the i-th particle after updating' ij As new particles X i ：
x′ ij ←x ij +v′ ij ,1≤j≤3,
v′ ij ←v ij +c 1 r 1j (pb ij -x ij )+c 2 r 2j (gb ij -x ij )
Wherein c 1 And c 2 For learning rate, r 1j And r 2j A random number between 0 and 1; v' ij Representing the velocity, v, of the updated ith particle ij Indicating the speed of the ith particle before updating, pb ij Indicating the locally optimal solution, gb, of the ith particle ij Globally optimal solutions representing particle populations, i.e. globally optimal particles, x ij Indicating the position of the ith particle before updating, x' ij Representing the position of the i-th particle after updating;
step S7, let iteration times T=T+1, judge whether iteration times is greater than or equal to the maximum iteration times, if yes, obtain global optimal particle (m) * ，C * ，δ * ) Further obtaining the optimized sub-block channel number, the optimized punishment parameter and the optimized kernel function parameter; otherwise, return to step S4.
In still another optional implementation manner, the step S13 "preprocessing the face image area to be identified according to the optimized number of sub-block channels to obtain a face feature vector, where the dimension of the face feature vector is equal to the product of the optimized number of sub-block channels and the preset number of sub-blocks", specifically includes:
after the identified face image area is subjected to inclination correction, performing binarization processing on the face area by adopting a self-adaptive threshold method;
according to pre-predictionDividing the face image to be recognized after binarization processing into n according to the positions occupied by the facial five sense organs areas by the number of the sub-blocks 0 Sub-blocks, where n 0 The number of the sub-blocks is preset;
dividing each sub-block into m channels according to the optimized number of sub-block channels to obtain N channels, wherein m is equal to the optimized number of sub-block channels, and N is equal to the total number of channels;
and calculating the direction gradient value of each channel, and obtaining the face feature vector according to the direction gradient value of each channel.
It can be understood that the method for identifying the face image area is the same as the processing method of the training set, and will not be described in detail here.
Accordingly, referring to fig. 2, fig. 2 is a block diagram of a face recognition device according to an embodiment of the present invention. The face recognition device provided by the embodiment of the invention comprises:
the parameter optimization and model acquisition module 110 is configured to acquire a nonlinear support vector machine model, and train the nonlinear support vector machine model according to a training set and a testing set acquired in advance to obtain an optimized nonlinear support vector machine model and an optimized sub-block channel number; the training set and the testing set comprise a plurality of face feature vector samples with different sub-block channel numbers;
a face image acquisition module 120, configured to acquire a face image to be identified;
an optimized face feature vector obtaining module 130, configured to pre-process the face image area to be identified according to the optimized number of sub-block channels, to obtain a face feature vector, where the dimension of the face feature vector is equal to the product of the optimized number of sub-block channels and a preset number of sub-blocks;
and the face recognition result obtaining module 140 is configured to input the face feature vector to the optimized nonlinear support vector machine model to obtain a face recognition result.
In an optional implementation manner, the optimized face feature vector obtaining module 130 is specifically configured to:
after inclination correction is carried out on a face image area to be identified, binarization processing is carried out on the face area by adopting a self-adaptive threshold method;
dividing the face image to be recognized after binarization processing into n according to the positions occupied by the facial five-sense organ regions according to the number of preset sub-blocks 0 Sub-blocks, where n 0 The number of the sub-blocks is preset;
dividing each sub-block into m channels according to the optimized number of sub-block channels to obtain N channels, wherein m is equal to the optimized number of sub-block channels, and N is equal to the total number of channels;
and calculating the direction gradient value of each channel, and obtaining the face feature vector according to the direction gradient value of each channel.
In another alternative embodiment, the parameter optimization and model acquisition module 110 includes:
the optimization parameter acquisition unit is used for optimizing and training the punishment parameters and the kernel function parameters in the establishment process of the face feature vector by adopting a particle population algorithm according to a training set and a testing set which are acquired in advance to obtain an optimized punishment parameter, an optimized kernel function parameter and an optimized sub-block channel number;
and the optimization model acquisition unit is used for obtaining an optimized nonlinear support vector machine model according to the optimized penalty parameter and the optimized kernel function parameter.
In yet another alternative embodiment, the optimization parameter acquiring unit includes:
particle determination: combining the punishment parameters and the kernel function parameters in the establishment process of the nonlinear support vector machine model into particles;
particle population acquisition: randomly generating different combinations of the sub-block channel number, the penalty parameter and the kernel function parameter to obtain a particle population, wherein each particle is expressed as: x is X i ＝(m i ，C i ，δ i )，X i Represents the ith particle, m i Represents the number of sub-block channels corresponding to the ith particle, C i Represents penalty parameter, delta, corresponding to the ith particle i Representing kernel function parameters corresponding to the ith particle;
and a particle fitness calculating step: acquiring multiple sets of training sets S i And multiple sets of test sets T i Wherein the training set S i The dimension of the face feature vector sample in the test set T is set according to the number of sub-block channels corresponding to the ith particle i The dimension in the face feature vector sample is set according to the number of sub-block channels corresponding to the ith particle;
using training sets S i Penalty parameter C for ith particle i Kernel function parameter delta corresponding to the ith particle i Training the nonlinear support vector machine model to obtain a five-sense organ recognition model M i ；
The five sense organs identification model M i Acting on test set T i Obtaining the identification accuracy of the ith particle, and taking the identification accuracy of the ith particle as the fitness of the ith particle;
a global optimal particle determining step: taking the particle with the highest adaptability in all particles as a global optimal particle;
particle updating: for each particle, updating the position and speed of each particle according to the fitness of each particle to obtain a new particle;
an optimized parameter obtaining step: judging whether a preset condition is met, if so, acquiring the global optimal particles to obtain optimized punishment parameters, optimized kernel function parameters and optimized sub-block channel numbers; otherwise, returning to the particle fitness calculating step.
It should be noted that, the above-mentioned face recognition device is configured to execute all the processes and steps of the face recognition method according to the embodiment of the present invention, and the working principles and functions of the two are in one-to-one correspondence, which is not repeated here.
Furthermore, the apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in the drawings of the embodiment of the device provided by the invention, the connection relation between the modules represents that the modules have communication connection, and can be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art will understand and implement the present invention without undue burden. In addition, the network function chain deployment device provided in the above embodiment and the network function chain deployment method provided in the embodiment of the present invention belong to the same concept, and specific implementation processes and specific technical schemes of the network function chain deployment device are detailed in the above method embodiment, and are not repeated here.
Correspondingly, an embodiment of the present invention further provides a computer readable storage medium, which is characterized in that the computer readable storage medium includes a stored computer program, where the device where the computer readable storage medium is located is controlled to execute steps S11 to S14 of the face recognition method when the computer program runs.
The computer readable storage medium may include: any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), an electrical carrier signal, a telecommunications signal, a software distribution medium, and so forth.
While the foregoing is directed to the preferred embodiments of the present invention, it will be appreciated by those skilled in the art that changes and modifications may be made without departing from the principles of the invention, such changes and modifications are also intended to be within the scope of the invention.
Claims (8)
1. A face recognition method, comprising:
acquiring a nonlinear support vector machine model, and optimizing and training punishment parameters and kernel function parameters in the establishment process of the nonlinear support vector machine model by adopting a particle population algorithm according to a pre-acquired training set and a pre-acquired testing set to obtain optimized punishment parameters, optimized kernel function parameters and optimized sub-block channel numbers; obtaining an optimized nonlinear support vector machine model according to the optimized penalty parameter and the optimized kernel function parameter; the training set and the testing set comprise a plurality of face feature vector samples with different sub-block channel numbers; the number of the sub-block channels is the number of channels contained in each sub-block;
acquiring a face image to be recognized;
preprocessing the face image area to be identified according to the optimized number of sub-block channels to obtain a face feature vector, wherein the dimension of the face feature vector is equal to the product of the optimized number of sub-block channels and the preset number of sub-blocks;
and inputting the face feature vector into the optimized nonlinear support vector machine model to obtain a face recognition result.
2. The face recognition method of claim 1, wherein the nonlinear support vector machine model is specifically:
the optimization problem of the nonlinear support vector machine model is converted into the following dual problem:
wherein x is i Is the ith sample in samples of given scale n, w is the weight vector, y i Is x i The value of 1 or-1, xi i For non-negative relaxation variables, C is an adjustable penalty parameter, n is the number of samples, L (alpha) represents a Lagrangian function, alpha i Represents Lagrange multiplier, alpha j Lagrangian multiplier, x representing dual j Is the j-th sample of the given n-scale samples that is dual to the i-th sample, y j Is x j Is a label of (a).
3. The face recognition method of claim 1, wherein the preprocessing is performed on the face image area to be recognized according to the optimized number of sub-block channels to obtain a face feature vector, and specifically includes:
after inclination correction is carried out on a face image area to be identified, binarization processing is carried out on the face area by adopting a self-adaptive threshold method;
dividing the face image to be recognized after binarization processing into n according to the positions occupied by facial image five-sense organ areas according to the number of preset sub-blocks 0 Sub-blocks, where n 0 The number of the sub-blocks is preset;
dividing each sub-block into m channels according to the optimized sub-block channel number to obtain N channels, wherein m is equal to the optimized sub-block channel number, and N is equal to the total channel number;
and calculating the direction gradient value of each channel, and obtaining the face feature vector according to the direction gradient value of each channel.
4. The face recognition method of claim 1, wherein the optimizing training is performed on the punishment parameters and the kernel function parameters in the process of building the face feature vector by using a particle population algorithm according to a pre-acquired training set and a testing set to obtain the optimized punishment parameters, the optimized kernel function parameters and the optimized sub-block channel number, and the method specifically comprises:
particle determination: combining the punishment parameters and the kernel function parameters in the establishment process of the nonlinear support vector machine model into particles;
particle population acquisition: randomly generating different combinations of the sub-block channel number, the penalty parameter and the kernel function parameter to obtain a particle population, wherein each particle is expressed as: x is X i ＝(m i ，C i ，δ i )，X i Represents the ith particle, m i Represents the number of sub-block channels corresponding to the ith particle, C i Represents penalty parameter, delta, corresponding to the ith particle i Representing kernel function parameters corresponding to the ith particle;
and a particle fitness calculating step: acquiring multiple sets of training sets S i And multiple sets of test sets T i Wherein the training set S i The dimension of the face feature vector sample in the test set T is set according to the number of sub-block channels corresponding to the ith particle i The dimension in the face feature vector sample is set according to the number of sub-block channels corresponding to the ith particle;
using training sets S i Penalty parameter C for ith particle i Kernel function parameter delta corresponding to the ith particle i Training the nonlinear support vector machine model to obtain a five-sense organ recognition model M i ；
The five sense organs identification model M i Acting on test set T i Obtaining the identification accuracy of the ith particle, and taking the identification accuracy of the ith particle as the fitness of the ith particle;
a global optimal particle determining step: taking the particle with the highest adaptability in all particles as a global optimal particle;
particle updating: for each particle, updating the position and speed of each particle according to the fitness of each particle to obtain a new particle;
an optimized parameter obtaining step: judging whether preset conditions are met, if yes, acquiring the global optimal particles, and obtaining optimized punishment parameters, optimized kernel function parameters and optimized sub-block channel numbers; otherwise, returning to the particle fitness calculating step.
5. A face recognition device, comprising:
the optimization parameter and model acquisition module is used for acquiring a nonlinear support vector machine model, adopting a particle population algorithm to carry out optimization training on punishment parameters and kernel function parameters in the process of establishing the face feature vector according to a training set and a testing set which are acquired in advance, and obtaining the optimized punishment parameters, the optimized kernel function parameters and the optimized sub-block channel number; obtaining an optimized nonlinear support vector machine model according to the optimized penalty parameter and the optimized kernel function parameter; the training set and the testing set comprise a plurality of face feature vector samples with different sub-block channel numbers; the number of the sub-block channels is the number of channels contained in each sub-block;
the face image acquisition module is used for acquiring a face image to be identified;
the optimized face feature vector acquisition module is used for preprocessing the face image area to be recognized according to the optimized number of sub-block channels to obtain a face feature vector, wherein the dimension of the face feature vector is equal to the product of the optimized number of sub-block channels and the preset number of sub-blocks;
and the face recognition result acquisition module is used for inputting the face feature vector into the optimized nonlinear support vector machine model to obtain a face recognition result.
6. The apparatus for face recognition according to claim 5, wherein the optimized face feature vector obtaining module is specifically configured to:
after inclination correction is carried out on a face image area to be identified, binarization processing is carried out on the face area by adopting a self-adaptive threshold method;
dividing the face image to be recognized after binarization processing into n according to the positions occupied by facial image five-sense organ areas according to the number of preset sub-blocks 0 Sub-blocks, where n 0 Is preset asThe number of sub-blocks;
dividing each sub-block into m channels according to the optimized sub-block channel number to obtain N channels, wherein m is equal to the optimized sub-block channel number, and N is equal to the total channel number;
and calculating the direction gradient value of each channel, and obtaining the face feature vector according to the direction gradient value of each channel.
7. The face recognition device according to claim 5, wherein the optimization parameter acquiring unit is specifically configured to perform:
particle determination: combining the punishment parameters and the kernel function parameters in the establishment process of the nonlinear support vector machine model into particles;
particle population acquisition: randomly generating different combinations of the sub-block channel number, the penalty parameter and the kernel function parameter to obtain a particle population, wherein each particle is expressed as: x is X i ＝(m i ，C i ，δ i )，X i Represents the ith particle, m i Represents the number of sub-block channels corresponding to the ith particle, C i Represents penalty parameter, delta, corresponding to the ith particle i Representing kernel function parameters corresponding to the ith particle;
and a particle fitness calculating step: acquiring multiple sets of training sets S i And multiple sets of test sets T i Wherein the training set S i The dimension of the face feature vector sample in the test set T is set according to the number of sub-block channels corresponding to the ith particle i The dimension in the face feature vector sample is set according to the number of sub-block channels corresponding to the ith particle;
using training sets S i Penalty parameter C for ith particle i Kernel function parameter delta corresponding to the ith particle i Training the nonlinear support vector machine model to obtain a five-sense organ recognition model M i ；
The five sense organs identification model M i Acting on test set T i Obtaining the identification of the ith particleAccuracy, and taking the identification accuracy of the ith particle as the fitness of the ith particle;
a global optimal particle determining step: taking the particle with the highest adaptability in all particles as a global optimal particle;
particle updating: for each particle, updating the position and speed of each particle according to the fitness of each particle to obtain a new particle;
an optimized parameter obtaining step: judging whether a preset condition is met, if so, acquiring the global optimal particles to obtain optimized punishment parameters, optimized kernel function parameters and optimized sub-block channel numbers; otherwise, returning to the particle fitness calculating step.
8. A computer readable storage medium, characterized in that the computer readable storage medium comprises a stored computer program, wherein the computer program, when run, controls a device in which the computer readable storage medium is located to perform the face recognition method according to any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110472915.2A CN113239761B (en) | 2021-04-29 | 2021-04-29 | Face recognition method, device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110472915.2A CN113239761B (en) | 2021-04-29 | 2021-04-29 | Face recognition method, device and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113239761A CN113239761A (en) | 2021-08-10 |
| CN113239761B true CN113239761B (en) | 2023-11-14 |
Family
ID=77131354
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110472915.2A Active CN113239761B (en) | 2021-04-29 | 2021-04-29 | Face recognition method, device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113239761B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104766063A (en) * | 2015-04-08 | 2015-07-08 | 宁波大学 | Living body human face identifying method |
| CN107679528A (en) * | 2017-11-24 | 2018-02-09 | 广西师范大学 | A kind of pedestrian detection method based on AdaBoost SVM Ensemble Learning Algorithms |
| WO2018072351A1 (en) * | 2016-10-20 | 2018-04-26 | 北京工业大学 | Method for optimizing support vector machine on basis of particle swarm optimization algorithm |
| CN109033954A (en) * | 2018-06-15 | 2018-12-18 | 西安科技大学 | A kind of aerial hand-written discrimination system and method based on machine vision |
| CN109359536A (en) * | 2018-09-14 | 2019-02-19 | 华南理工大学 | Passenger behavior monitoring method based on machine vision |
| CN111723700A (en) * | 2020-06-08 | 2020-09-29 | 国网河北省电力有限公司信息通信分公司 | Face recognition method and device and electronic equipment |
| CN111783704A (en) * | 2020-07-07 | 2020-10-16 | 中电万维信息技术有限责任公司 | Face recognition system based on particle swarm optimization radial basis |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101877981B1 (en) * | 2011-12-21 | 2018-07-12 | 한국전자통신연구원 | System for recognizing disguised face using gabor feature and svm classifier and method thereof |
-
2021
- 2021-04-29 CN CN202110472915.2A patent/CN113239761B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104766063A (en) * | 2015-04-08 | 2015-07-08 | 宁波大学 | Living body human face identifying method |
| WO2018072351A1 (en) * | 2016-10-20 | 2018-04-26 | 北京工业大学 | Method for optimizing support vector machine on basis of particle swarm optimization algorithm |
| CN107679528A (en) * | 2017-11-24 | 2018-02-09 | 广西师范大学 | A kind of pedestrian detection method based on AdaBoost SVM Ensemble Learning Algorithms |
| CN109033954A (en) * | 2018-06-15 | 2018-12-18 | 西安科技大学 | A kind of aerial hand-written discrimination system and method based on machine vision |
| CN109359536A (en) * | 2018-09-14 | 2019-02-19 | 华南理工大学 | Passenger behavior monitoring method based on machine vision |
| CN111723700A (en) * | 2020-06-08 | 2020-09-29 | 国网河北省电力有限公司信息通信分公司 | Face recognition method and device and electronic equipment |
| CN111783704A (en) * | 2020-07-07 | 2020-10-16 | 中电万维信息技术有限责任公司 | Face recognition system based on particle swarm optimization radial basis |
Non-Patent Citations (2)
| Title |
|---|
| 基于粒子群优化的支持向量机人脸识别;廖周宇;王钰婷;谢晓兰;刘建明;;计算机工程(12);第248-254页 * |
| 改进的支持向量机算法在人脸识别上的应用;谌璐;贺兴时;;纺织高校基础科学学报(01);第108-115页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113239761A (en) | 2021-08-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106169081B (en) | A kind of image classification and processing method based on different illumination | |
| CN105069400B (en) | Facial image gender identifying system based on the sparse own coding of stack | |
| Kotte et al. | An efficient approach for optimal multilevel thresholding selection for gray scale images based on improved differential search algorithm | |
| CN107341463B (en) | Face feature recognition method combining image quality analysis and metric learning | |
| Javed et al. | Multiplex cellular communities in multi-gigapixel colorectal cancer histology images for tissue phenotyping | |
| CN109559309B (en) | Multi-objective optimization infrared thermal image defect feature extraction method based on uniform evolution | |
| CN116137036B (en) | Gene detection data intelligent processing system based on machine learning | |
| CN110889332A (en) | Lie detection method based on micro expression in interview | |
| CN111931700A (en) | Corn variety authenticity identification method and identification system based on multiple classifiers | |
| CN109961425A (en) | A kind of water quality recognition methods of Dynamic Water | |
| CN111726472B (en) | Image anti-interference method based on encryption algorithm | |
| CN111126169B (en) | Face recognition method and system based on orthogonalization graph regular nonnegative matrix factorization | |
| CN115984930A (en) | Micro expression recognition method and device and micro expression recognition model training method | |
| CN111563556A (en) | Transformer substation cabinet equipment abnormity identification method and system based on color gradient weight | |
| CN113239761B (en) | Face recognition method, device and storage medium | |
| CN112861743A (en) | Palm vein image anti-counterfeiting method, device and equipment | |
| CN117649621A (en) | Fake video detection method, device and equipment | |
| CN110222793B (en) | Online semi-supervised classification method and system based on multi-view active learning | |
| CN114820550A (en) | Disease prediction system based on block chain and medical image | |
| CN114078194A (en) | Vehicle identity discrimination method | |
| CN113537173A (en) | Face image authenticity identification method based on face patch mapping | |
| CN112613415A (en) | Face nose type recognition method and device, electronic equipment and medium | |
| CN107341485B (en) | Face recognition method and device | |
| CN116386038B (en) | DC cell detection method and system | |
| CN111340111B (en) | Method for recognizing face image set based on wavelet kernel extreme learning machine |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |