CN110502989A - A kind of small sample EO-1 hyperion face identification method and system - Google Patents

A kind of small sample EO-1 hyperion face identification method and system Download PDF

Info

Publication number
CN110502989A
CN110502989A CN201910641561.2A CN201910641561A CN110502989A CN 110502989 A CN110502989 A CN 110502989A CN 201910641561 A CN201910641561 A CN 201910641561A CN 110502989 A CN110502989 A CN 110502989A
Authority
CN
China
Prior art keywords
hyperion
recognition model
sample
face recognition
training
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
CN201910641561.2A
Other languages
Chinese (zh)
Inventor
赵曰峰
吴蒙蒙
刘永雷
李秀秀
魏小琴
蔡阳健
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.)
Shandong Normal University
Original Assignee
Shandong Normal University
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 Shandong Normal University filed Critical Shandong Normal University
Priority to CN201910641561.2A priority Critical patent/CN110502989A/en
Publication of CN110502989A publication Critical patent/CN110502989A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • 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

Landscapes

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

Abstract

Present disclose provides small sample EO-1 hyperion face identification method and systems.Wherein, small sample EO-1 hyperion face identification method, comprising: building EO-1 hyperion human face recognition model step: EO-1 hyperion human face recognition model is made of the local classifiers and K nearest neighbor classifier being connected in parallel；Local classifiers are made of the Naive Bayes Classifier that several are connected in parallel；The training of EO-1 hyperion human face recognition model and Optimization Steps, process are as follows: obtain EO-1 hyperion facial image and as sample and mark face classification；EO-1 hyperion facial image sample is divided into the candidate regional area equal with Naive Bayes Classifier quantity, extracts the local feature of each candidate regional area, and is input to corresponding Naive Bayes Classifier and is trained；The low frequency component of spectrum Fourier transformation in high spectrum image sample is extracted as global characteristics, training K nearest neighbor classifier；The EO-1 hyperion human face recognition model that optimization training is completed；Test output EO-1 hyperion face classification step.

Description

A kind of small sample EO-1 hyperion face identification method and system
Technical field
The disclosure belongs to field of face identification more particularly to a kind of small sample EO-1 hyperion face identification method and system.
Background technique
Only there is provided background technical informations relevant to the disclosure for the statement of this part, it is not necessary to so constitute first skill Art.
Currently, face has been widely applied to personal verification, video monitoring, people as a kind of biological information Machine interaction etc. is compared with other biological characteristic information such as iris, fingerprint etc., face have it is untouchable, at a distance can It realizes the feature of purpose, but is influenced (such as posture, expression, light, block) by non-limiting condition identification process is still There are some challenges.
Inventors have found that there are following problems during automatically processing and identifying face: (a) illumination level and property Variation；(b) with the reduction of illumination level, signal-to-noise ratio is risen rapidly；(c) image obtained under unconfined condition, packet are handled Include night and remote etc.；(d) great amount of samples is difficult to obtain；The above problem may cause recognition of face calculating speed difference and identify Accuracy rate is low.
Summary of the invention
To solve the above-mentioned problems, the first aspect of the disclosure provides a kind of small sample EO-1 hyperion face identification method, With the advantages that convenient real-time, calculating speed is fast, and recognition accuracy is high.
To achieve the goals above, the disclosure adopts the following technical scheme that
A kind of small sample EO-1 hyperion face identification method, comprising:
Building EO-1 hyperion human face recognition model step: the EO-1 hyperion human face recognition model is classified by the part being connected in parallel Device and K nearest neighbor classifier are constituted；Local classifiers are made of the Naive Bayes Classifier that several are connected in parallel；
The training of EO-1 hyperion human face recognition model and Optimization Steps, process are as follows:
It obtains EO-1 hyperion facial image and as sample and marks face classification；
EO-1 hyperion facial image sample is divided into the candidate regional area equal with Naive Bayes Classifier quantity, is mentioned The local feature of each candidate regional area is taken, and is input to corresponding Naive Bayes Classifier and is trained；
The low frequency component of spectrum Fourier transformation in high spectrum image sample is extracted as global characteristics, training K arest neighbors Classifier；Local classifiers and EO-1 hyperion human face recognition model are separately optimized using particle swarm optimization algorithm, obtain optimization training The EO-1 hyperion human face recognition model of completion；
Test output EO-1 hyperion face classification step: local feature and the overall situation for extracting EO-1 hyperion facial image to be tested are special Sign is input in the EO-1 hyperion human face recognition model that optimization training is completed, and exports EO-1 hyperion face classification.
The second aspect of the disclosure provides a kind of small sample EO-1 hyperion face identification system.
A kind of small sample EO-1 hyperion face identification system, comprising:
Construct EO-1 hyperion human face recognition model module, be used for: the EO-1 hyperion human face recognition model is by being connected in parallel Local classifiers and K nearest neighbor classifier are constituted；The Naive Bayes Classifier structure that local classifiers are connected in parallel by several At；
The training of EO-1 hyperion human face recognition model and optimization module, are used for:
It obtains EO-1 hyperion facial image and as sample and marks face classification；
EO-1 hyperion facial image sample is divided into the candidate regional area equal with Naive Bayes Classifier quantity, is mentioned The local feature of each candidate regional area is taken, and is input to corresponding Naive Bayes Classifier and is trained；
The low frequency component of spectrum Fourier transformation in high spectrum image sample is extracted as global characteristics, training K arest neighbors Classifier；Local classifiers and EO-1 hyperion human face recognition model are separately optimized using particle swarm optimization algorithm, obtain optimization training The EO-1 hyperion human face recognition model of completion；
Test output EO-1 hyperion face class Modules, are used for: extracting the local feature of EO-1 hyperion facial image to be tested It is input to global characteristics in the EO-1 hyperion human face recognition model that optimization training is completed, exports EO-1 hyperion face classification.
A kind of computer readable storage medium is provided in terms of the third of the disclosure.
A kind of computer readable storage medium, is stored thereon with computer program, realization when which is executed by processor Step in small sample EO-1 hyperion face identification method described above.
4th aspect of the disclosure provides a kind of computer equipment.
A kind of computer equipment can be run on a memory and on a processor including memory, processor and storage Computer program, the processor are realized in small sample EO-1 hyperion face identification method described above when executing described program Step.
The beneficial effect of the disclosure is:
The EO-1 hyperion human face recognition model of disclosure building is by the local classifiers and K nearest neighbor classifier structure that are connected in parallel At；Local classifiers are made of the Naive Bayes Classifier that several are connected in parallel；EO-1 hyperion facial image sample is divided At the candidate regional area equal with Naive Bayes Classifier quantity, the local feature of each candidate regional area is extracted, and Corresponding Naive Bayes Classifier is input to be trained；Extract the low frequency point of spectrum Fourier transformation in high spectrum image sample Amount is used as global characteristics, training K nearest neighbor classifier；Local classifiers and EO-1 hyperion are separately optimized using particle swarm optimization algorithm Human face recognition model obtains the EO-1 hyperion human face recognition model for optimizing training completion；Without collecting great amount of samples and not by user The interference of behavior has the advantages that accuracy is high, calculating speed is fast and the flexible robust of identification.
Detailed description of the invention
The Figure of description for constituting a part of this disclosure is used to provide further understanding of the disclosure, and the disclosure is shown Meaning property embodiment and its explanation do not constitute the improper restriction to the disclosure for explaining the disclosure.
Fig. 1 is a kind of small sample EO-1 hyperion face identification method flow chart that the embodiment of the present disclosure provides.
Fig. 2 is the procedure chart of the training of EO-1 hyperion human face recognition model and optimization that the embodiment of the present disclosure provides.
Fig. 3 is a kind of small sample EO-1 hyperion face identification system structural schematic diagram that the embodiment of the present disclosure provides.
Specific embodiment
The disclosure is described further with embodiment with reference to the accompanying drawing.
It is noted that following detailed description is all illustrative, it is intended to provide further instruction to the disclosure.Unless another It indicates, all technical and scientific terms used herein has usual with disclosure person of an ordinary skill in the technical field The identical meanings of understanding.
It should be noted that term used herein above is merely to describe specific embodiment, and be not intended to restricted root According to the illustrative embodiments of the disclosure.As used herein, unless the context clearly indicates otherwise, otherwise singular Also it is intended to include plural form, additionally, it should be understood that, when in the present specification using term "comprising" and/or " packet Include " when, indicate existing characteristics, step, operation, device, component and/or their combination.
Embodiment 1
Fig. 1 is a kind of small sample EO-1 hyperion face identification method flow chart that the embodiment of the present disclosure provides.
As shown in Figure 1, a kind of small sample EO-1 hyperion face identification method of the present embodiment, comprising:
S101: building EO-1 hyperion human face recognition model step: the EO-1 hyperion human face recognition model is by the office that is connected in parallel Classifier and K nearest neighbor classifier are constituted；Local classifiers are made of the Naive Bayes Classifier that several are connected in parallel.
Naive Bayes Classification: Bayes' theorem formula is simplified, i.e., introduces when solving the problems, such as several it is assumed that making It is simply readily understood to obtain model.For the sample to be classified, sample to be sorted calculates each classification and occurs herein under the premise of occurring Probability, and that classification with maximum probability, the i.e. classification of sample to be sorted thus.It indicates are as follows:
Wherein, x indicates sample, yiIndicate that sample class, a indicate sample value；P(x|yi) indicate that sample x belongs to sample class Other yiProbability；P(yi) indicate sample class yiProbability；P(aj|yi) expression sample value be ajBelong to sample class yiIt is general Rate.
Different regional areas carries complementary classification information, and local classifiers are integrated to integrate these local classifiers The performance of classification can be promoted together.
The classification of K arest neighbors can be new simply by it to be compared to find with the most like record in training set The classification of unfiled sample.Its thinking is: when k sample in test sample and training sample feature space is most like (i.e. It is closest in feature space) when, if most of k sample belongs to the same classification, which also belongs to this classification, The neighbouring sample selected in KNN algorithm is all the object correctly classified.
Particle group optimizing (Particle Swarm Optimization, PSO) is a kind of random optimization skill based on population Art, it simulates the social action of the shoal of fish or flock of birds.Each of PSO individual not will use evolutionary operator to manipulate individual, But as other evolutionary computation algorithms, change evolutionary rate in search space, speed can be unique according to oneself Performance carries out dynamic adjustment.In particle swarm optimization algorithm, system initialization is to include potential solution (referred to as particle) It is random overall, and optimal solution is searched in problem space, while more newly-generated optimal particle.The initial optimum position of particle (pbest) the best particle position (gbest) and in generating process can all influence the new position of particle, in the algorithm each grain The speed of son is all updated according to gbest, pbest and other control parameters in each step.In addition, being produced to pbest and gbest Random number is given birth to weight speed, has fully considered fitness function, it is also defined as the performance for evaluating each particle.
Validation data set is inputted into each classifier, and defines corresponding label vector；It then, will be any between 0 to 1 Weight coefficient of one number as each label vector, optimally determines these weights by PSO algorithm.Specific method is collection Regard the fitness function for needing to minimize as at classification total false rate of the system to validation data set, it is each in group (Swarm) Particle is all a candidate weight matrix, it is so structured that W={ w1,w2..., wL, wherein wiIt is allocated to i-th of single classification The weight of device.
S102: the training of EO-1 hyperion human face recognition model and Optimization Steps.
As shown in Fig. 2, the process of the training of EO-1 hyperion human face recognition model and optimization are as follows:
S1021: it obtains EO-1 hyperion facial image and as sample and marks face classification；
S1022: EO-1 hyperion facial image sample is divided into the candidate part equal with Naive Bayes Classifier quantity Region, extracts the local feature of each candidate regional area, and is input to corresponding Naive Bayes Classifier and is trained；
S1023: the low frequency component of spectrum Fourier transformation in high spectrum image sample is extracted as global characteristics, training K Nearest neighbor classifier；Local classifiers and EO-1 hyperion human face recognition model are separately optimized using particle swarm optimization algorithm, obtain excellent Change the EO-1 hyperion human face recognition model that training is completed.
It is marked off from image with different size of candidate regional area, the classification results of these regional areas usually have Institute is different.Such as M different regional areaFrom original three-dimensional image Middle interception comes out, wherein LW×LHThe size of representation space dimension, LSIndicate the size (l of spectral DimensionsW<LW,lH<LH)。
It should be noted that the feature of regional area further include: local binary patterns (LBP) and Gabor face characteristic. LBP is a kind of powerful tool, for encoding appearance and grain details；Gabor can be right on a series of rougher scales Shape and appearance information are encoded, it is evident that their complementary information abundant make the feature of fusion have more identification.
The classification separating property of different regional areas is evaluated with Fisher ratio on this basis.Fisher ratio is usually For the classification performance of classification of assessment device, the present invention is utilized to evaluate the classification separating property of different regional areas.
Fisher ratio:Wherein FijBe i-th and jth class it Between two classification situations, i.e.,SbIndicate inter-class variance, SwIndicate variance within clusters.Inter-class variance is bigger, side in class Difference is smaller, then it represents that classification performance is better, so biggish β value indicates better classification separating capacity.Wherein, p (yiyj) indicate The probability that i-th class and jth class occur, ni、njRespectively indicate the sum of i-th, j class sample.
The process of portion's feature extraction, comprising:
EO-1 hyperion face regional area tensor representation: tensor is the powerful for solving high dimensional data, it provides one Effective mathematical framework, for analyzing the multifactor structure of image.Same local area is integrated into the tetradic, to extract Identification local feature, four dimensions respectively indicate: classification dimension, two spaces dimension and a spectral Dimensions；It fully considers Holotopy across between the neighborhood relevance and frequency spectrum dimension of Spatial Dimension.
Tensor resolution under constraint condition: tucker core tensor resolution (TKD) is unable to diagonalization, so it is mainly realized Reduce dimension purpose.It is intentional to obtain to apply orthogonality, sparsity constraints condition on the basic matrix or core tensor of TKD The adopted and unique factor or matrix, or high order tensor data compression at smaller size or is resolved into understandable point Amount, avoids high dimensional data from bringing computation complexity.
When obtaining the orthogonal gene polyadenylation signal (dictionary) and core coefficients tensor property from training data, tensor network is reused Contractor (NCON algorithm) is that training data generates the distinguishing feature (referred to as LFT feature) with label information.Using from The orthogonal gene polyadenylation signal that training data obtains extracts the distinguishing feature of test data.
When extracting global characteristics, EO-1 hyperion face figure is solved using the wavestrip fusion method of space-optical spectrum covariance Problem of misalignment between the wavestrip of picture, the Band fusion method add spectrum integral information better than simple space average information, sufficiently The variance of spatially and spectrally dimension is utilized, and is contained in and differentiates in feature.
The Polar-FFT value that will be obtained under pseudo- polar grid based on fast algorithm, goes to Polar by interpolation stage Coordinate.Since this grid is closer to Polar coordinates of targets, have reason to believe that precision can be improved in this method, to drop Low over-sampling requirement.Another very important benefit is that necessary interpolation can be executed by pure 1D operation without losing standard True property；
The zero frequency item and average value f (x, y) of Fourier transformation are directly proportional, whereinIt is being averaged for input picture Value, since coefficient is usually very big, so | F (0,0) | value it is generally also very big, therefore the dynamic range of image spectrum is compressed, That is, the detailed features of image are lost completely.In this case, the low frequency component of Fourier transform reflects whole category The profile of property and image.
These characteristics of low-frequency are robusts for localized variation (including expression, noise etc.)；Low frequency size be [40, When 40], low frequency coefficient has best classification performance up to 72% or so, and not only classification performance is best but also also reduces calculating again Miscellaneous degree.Respectively link together the low frequency coefficient of real and imaginary parts in experiment the global Fourier's feature vector for indicating one-dimensional, Its size is 40 × 40 × 2=3200.
S103: test output EO-1 hyperion face classification step: extract EO-1 hyperion facial image to be tested local feature and Global characteristics are input in the EO-1 hyperion human face recognition model that optimization training is completed, and export EO-1 hyperion face classification.
Embodiment 2
Fig. 3 is a kind of small sample EO-1 hyperion face identification system structural schematic diagram that the embodiment of the present disclosure provides.
As shown in figure 3, present embodiments providing a kind of small sample EO-1 hyperion face identification system, comprising:
(1) EO-1 hyperion human face recognition model module is constructed, be used for: the EO-1 hyperion human face recognition model is by being connected in parallel Local classifiers and K nearest neighbor classifier constitute；The Naive Bayes Classifier that local classifiers are connected in parallel by several It constitutes；
(2) training of EO-1 hyperion human face recognition model and optimization module, are used for:
It obtains EO-1 hyperion facial image and as sample and marks face classification；
EO-1 hyperion facial image sample is divided into the candidate regional area equal with Naive Bayes Classifier quantity, is mentioned The local feature of each candidate regional area is taken, and is input to corresponding Naive Bayes Classifier and is trained；
The low frequency component of spectrum Fourier transformation in high spectrum image sample is extracted as global characteristics, training K arest neighbors Classifier；Local classifiers and EO-1 hyperion human face recognition model are separately optimized using particle swarm optimization algorithm, obtain optimization training The EO-1 hyperion human face recognition model of completion；
In the EO-1 hyperion human face recognition model training and optimization module, the local feature of candidate regional area is extracted Process are as follows:
Candidate regional area is indicated that the four dimensions of tensor are respectively classification dimension, two spaces using the tetradic Dimension and a spectral Dimensions；
Tensor is decomposed under constraint condition, obtains orthogonal factor and core coefficients tensor property, and then generate with label The distinguishing feature of information is as local feature.
Wherein, the constraint condition includes orthogonality and sparsity constraints condition.
(3) test output EO-1 hyperion face class Modules, are used for: the part for extracting EO-1 hyperion facial image to be tested is special Global characteristics of seeking peace are input in the EO-1 hyperion human face recognition model that optimization training is completed, and export EO-1 hyperion face classification.
Embodiment 3
A kind of computer readable storage medium is present embodiments provided, computer program is stored thereon with, which is located Reason device realizes the step in small sample EO-1 hyperion face identification method as shown in Figure 1 when executing.
Embodiment 4
A kind of computer equipment of the present embodiment, including memory, processor and storage are on a memory and can be in processor The computer program of upper operation, the processor realize that small sample EO-1 hyperion face as shown in Figure 1 is known when executing described program Step in other method.
It should be understood by those skilled in the art that, embodiment of the disclosure can provide as method, system or computer program Product.Therefore, the shape of hardware embodiment, software implementation or embodiment combining software and hardware aspects can be used in the disclosure Formula.Moreover, the disclosure, which can be used, can use storage in the computer that one or more wherein includes computer usable program code The form for the computer program product implemented on medium (including but not limited to magnetic disk storage and optical memory etc.).
The disclosure is referring to method, the process of equipment (system) and computer program product according to the embodiment of the present disclosure Figure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructions The combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programs Instruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produce A raw machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute for real The device for the function of being specified in present one or more flows of the flowchart and/or one or more blocks of the block diagram.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates, Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one The step of function of being specified in a box or multiple boxes.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with Relevant hardware is instructed to complete by computer program, the program can be stored in a computer-readable storage medium In, the program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein, the storage medium can be magnetic Dish, CD, read-only memory (Read-Only Memory, ROM) or random access memory (Random AccessMemory, RAM) etc..
The foregoing is merely preferred embodiment of the present disclosure, are not limited to the disclosure, for the skill of this field For art personnel, the disclosure can have various modifications and variations.It is all within the spirit and principle of the disclosure, it is made any to repair Change, equivalent replacement, improvement etc., should be included within the protection scope of the disclosure.

Claims (10)

1. a kind of small sample EO-1 hyperion face identification method characterized by comprising
Building EO-1 hyperion human face recognition model step: the EO-1 hyperion human face recognition model by the local classifiers that are connected in parallel and K nearest neighbor classifier is constituted；Local classifiers are made of the Naive Bayes Classifier that several are connected in parallel；
The training of EO-1 hyperion human face recognition model and Optimization Steps, process are as follows:
It obtains EO-1 hyperion facial image and as sample and marks face classification；
EO-1 hyperion facial image sample is divided into the candidate regional area equal with Naive Bayes Classifier quantity, is extracted each The local feature of a candidate's regional area, and be input to corresponding Naive Bayes Classifier and be trained；
The low frequency component of spectrum Fourier transformation in high spectrum image sample is extracted as global characteristics, training K arest neighbors classification Device；Local classifiers and EO-1 hyperion human face recognition model are separately optimized using particle swarm optimization algorithm, obtain optimizing trained completion EO-1 hyperion human face recognition model；
Test output EO-1 hyperion face classification step: local feature and the global characteristics for extracting EO-1 hyperion facial image to be tested are defeated Enter in the EO-1 hyperion human face recognition model completed to optimization training, exports EO-1 hyperion face classification.
2. a kind of small sample EO-1 hyperion face identification method as described in claim 1, which is characterized in that extract candidate partial zones The process of the local feature in domain are as follows:
Candidate regional area is indicated that the four dimensions of tensor are respectively classification dimension, two spaces dimension using the tetradic With a spectral Dimensions；
Tensor is decomposed under constraint condition, obtains orthogonal factor and core coefficients tensor property, and then generate with label information Distinguishing feature as local feature.
3. a kind of small sample EO-1 hyperion face identification method as claimed in claim 2, which is characterized in that the constraint condition packet Include orthogonality and sparsity constraints condition.
4. a kind of small sample EO-1 hyperion face identification method as described in claim 1, which is characterized in that know in EO-1 hyperion face In other model training and Optimization Steps, further includes:
The classification separating property of different regional areas is evaluated using Fisher ratio.
5. a kind of small sample EO-1 hyperion face identification system characterized by comprising
Construct EO-1 hyperion human face recognition model module, be used for: the EO-1 hyperion human face recognition model is by the part that is connected in parallel Classifier and K nearest neighbor classifier are constituted；Local classifiers are made of the Naive Bayes Classifier that several are connected in parallel；
The training of EO-1 hyperion human face recognition model and optimization module, are used for:
It obtains EO-1 hyperion facial image and as sample and marks face classification；
EO-1 hyperion facial image sample is divided into the candidate regional area equal with Naive Bayes Classifier quantity, is extracted each The local feature of a candidate's regional area, and be input to corresponding Naive Bayes Classifier and be trained；
The low frequency component of spectrum Fourier transformation in high spectrum image sample is extracted as global characteristics, training K arest neighbors classification Device；Local classifiers and EO-1 hyperion human face recognition model are separately optimized using particle swarm optimization algorithm, obtain optimizing trained completion EO-1 hyperion human face recognition model；
Test output EO-1 hyperion face class Modules, are used for: extracting the local feature of EO-1 hyperion facial image to be tested and complete Office's feature is input in the EO-1 hyperion human face recognition model that optimization training is completed, and exports EO-1 hyperion face classification.
6. a kind of small sample EO-1 hyperion face identification system as claimed in claim 5, which is characterized in that in the EO-1 hyperion people In the training of face identification model and optimization module, the process of the local feature of candidate regional area is extracted are as follows:
Candidate regional area is indicated that the four dimensions of tensor are respectively classification dimension, two spaces dimension using the tetradic With a spectral Dimensions；
Tensor is decomposed under constraint condition, obtains orthogonal factor and core coefficients tensor property, and then generate with label information Distinguishing feature as local feature.
7. a kind of small sample EO-1 hyperion face identification system as claimed in claim 6, which is characterized in that the constraint condition packet Include orthogonality and sparsity constraints condition.
8. a kind of small sample EO-1 hyperion face identification system as claimed in claim 5, which is characterized in that in the EO-1 hyperion people In the training of face identification model and optimization module, the classification separating property of different regional areas is evaluated using Fisher ratio.
9. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is held by processor It realizes when row such as the step in small sample EO-1 hyperion face identification method of any of claims 1-4.
10. a kind of computer equipment including memory, processor and stores the meter that can be run on a memory and on a processor Calculation machine program, which is characterized in that the processor is realized when executing described program as of any of claims 1-4 small Step in sample EO-1 hyperion face identification method.
CN201910641561.2A 2019-07-16 2019-07-16 A kind of small sample EO-1 hyperion face identification method and system Pending CN110502989A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910641561.2A CN110502989A (en) 2019-07-16 2019-07-16 A kind of small sample EO-1 hyperion face identification method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910641561.2A CN110502989A (en) 2019-07-16 2019-07-16 A kind of small sample EO-1 hyperion face identification method and system

Publications (1)

Publication Number Publication Date
CN110502989A true CN110502989A (en) 2019-11-26

Family

ID=68586113

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910641561.2A Pending CN110502989A (en) 2019-07-16 2019-07-16 A kind of small sample EO-1 hyperion face identification method and system

Country Status (1)

Country Link
CN (1) CN110502989A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110991551A (en) * 2019-12-13 2020-04-10 北京百度网讯科技有限公司 Sample processing method, sample processing device, electronic device and storage medium
CN111444860A (en) * 2020-03-30 2020-07-24 东华大学 Expression recognition method and system
CN111680599A (en) * 2020-05-29 2020-09-18 北京百度网讯科技有限公司 Face recognition model processing method, device, equipment and storage medium
WO2023000864A1 (en) * 2021-07-19 2023-01-26 清华大学 Face recognition method and system

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103927531A (en) * 2014-05-13 2014-07-16 江苏科技大学 Human face recognition method based on local binary value and PSO BP neural network
WO2018089081A1 (en) * 2016-11-08 2018-05-17 Qualcomm Incorporated System and method associated with object verification

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103927531A (en) * 2014-05-13 2014-07-16 江苏科技大学 Human face recognition method based on local binary value and PSO BP neural network
WO2018089081A1 (en) * 2016-11-08 2018-05-17 Qualcomm Incorporated System and method associated with object verification

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
MENGMENG WU, ET AL.: "Hyperspectral Face Recognition with Patch-Based Low Rank Tensor Decomposition and PFFT Algorithm", 《SYMMETRY》 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110991551A (en) * 2019-12-13 2020-04-10 北京百度网讯科技有限公司 Sample processing method, sample processing device, electronic device and storage medium
CN110991551B (en) * 2019-12-13 2023-09-15 北京百度网讯科技有限公司 Sample processing method, device, electronic equipment and storage medium
CN111444860A (en) * 2020-03-30 2020-07-24 东华大学 Expression recognition method and system
CN111680599A (en) * 2020-05-29 2020-09-18 北京百度网讯科技有限公司 Face recognition model processing method, device, equipment and storage medium
CN111680599B (en) * 2020-05-29 2023-08-08 北京百度网讯科技有限公司 Face recognition model processing method, device, equipment and storage medium
WO2023000864A1 (en) * 2021-07-19 2023-01-26 清华大学 Face recognition method and system

Similar Documents

Publication Publication Date Title
CN106779087B (en) A kind of general-purpose machinery learning data analysis platform
CN106599797B (en) A kind of infrared face recognition method based on local parallel neural network
CN110502989A (en) A kind of small sample EO-1 hyperion face identification method and system
CN106295124B (en) The method of a variety of image detecting technique comprehensive analysis gene subgraph likelihood probability amounts
CN109582003A (en) Based on pseudo label semi-supervised kernel part Fei Sheer discriminant analysis bearing failure diagnosis
CN101763507B (en) Face recognition method and face recognition system
CN104268593A (en) Multiple-sparse-representation face recognition method for solving small sample size problem
CN102324038B (en) Plant species identification method based on digital image
CN108446716A (en) Based on FCN the PolSAR image classification methods merged are indicated with sparse-low-rank subspace
CN109508360A (en) A kind of polynary flow data space-time autocorrelation analysis method of geography based on cellular automata
CN109615014A (en) A kind of data sorting system and method based on the optimization of KL divergence
CN104866831B (en) The face recognition algorithms of characteristic weighing
Ma et al. Linear dependency modeling for classifier fusion and feature combination
CN105678261B (en) Based on the direct-push Method of Data with Adding Windows for having supervision figure
CN102509123A (en) Brain functional magnetic resonance image classification method based on complex network
CN112926645B (en) Electricity stealing detection method based on edge calculation
CN113807299B (en) Sleep stage staging method and system based on parallel frequency domain electroencephalogram signals
CN117037427B (en) Geological disaster networking monitoring and early warning system
Masood et al. Differential evolution based advised SVM for histopathalogical image analysis for skin cancer detection
CN110288028A (en) ECG detecting method, system, equipment and computer readable storage medium
CN107918773A (en) A kind of human face in-vivo detection method, device and electronic equipment
CN106682653A (en) KNLDA-based RBF neural network face recognition method
CN111639697A (en) Hyperspectral image classification method based on non-repeated sampling and prototype network
Copiaco et al. Exploring deep time-series imaging for anomaly detection of building energy consumption
CN111931670B (en) Depth image head detection and positioning method and system based on convolutional neural network

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