CN110222571A - Black eye intelligent determination method, device and computer readable storage medium - Google Patents
Black eye intelligent determination method, device and computer readable storage medium Download PDFInfo
- Publication number
- CN110222571A CN110222571A CN201910370451.7A CN201910370451A CN110222571A CN 110222571 A CN110222571 A CN 110222571A CN 201910370451 A CN201910370451 A CN 201910370451A CN 110222571 A CN110222571 A CN 110222571A
- Authority
- CN
- China
- Prior art keywords
- sample collection
- black eye
- data
- negative sample
- black
- 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.)
- Granted
Links
Classifications
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
- G06V40/18—Eye characteristics, e.g. of the iris
- G06V40/197—Matching; Classification
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Ophthalmology & Optometry (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of artificial intelligence technologys, disclose a kind of black eye intelligent determination method, it include: to receive face image set and tally set, and the face picture collection is pre-processed, and using the tally set as the input value of the loss function of black eye judgment models, after the data of the face image set are carried out the operation of direction gradient histogram, it is input to the support vector machines module of the black eye judgment models, the support vector machines module enters data into after carrying out just classification to face image set to the convolutional neural networks module retraining of the black eye judgment models;The test set for receiving user judges whether there is black eye for the black eye judgment models are input to after test set progress direction gradient histogram operation, and exports result.The present invention also proposes a kind of black eye intelligent judging device and a kind of computer readable storage medium.Accurately black eye intelligent judgment function may be implemented in the present invention.
Description
Technical field
It is black the present invention relates to that can be judged automatically after field of artificial intelligence, more particularly to a kind of input based on human face data
The method, apparatus and computer readable storage medium of eyelet.
Background technique
Black eye is due to staying up late, and mood swing is big, and asthenopia, aging cause eye part skin vascular flow speed excessively slow
Slow formed remains in a standstill, and tissue is for hypoxgia, and metabolic waste accumulation is excessive in blood vessel, causes eye pigmentation.Age bigger people,
Subcutaneous fat around eyes becomes thinner, so black eye just becomes apparent from.With the presence of many people's black eyes in today's society
Without knowing, it is therefore desirable to judge with the presence or absence of black eye, however, having much for black-eyed accurately identify also
Problem, if application scenarios majority is more complicated, target shadow caused by the variation of the local dynamic station of background, uneven illumination etc. can be to knowledge
Not Zeng Jia difficulty possess posture feature abundant in addition, face is non-rigid targets, different posture locating for same face,
Often difference is very big in detection and identification.
Summary of the invention
The present invention provides a kind of black eye intelligent determination method, device and computer readable storage medium, main purpose
It is to show accurately black eye judging result to user when user judges whether face has black eye.
To achieve the above object, a kind of black eye intelligent determination method provided by the invention, comprising:
Data receiver layer receives face image set, and the face image set is divided into positive sample collection by tally set and bears
Sample set, the positive sample collection and the negative sample collection are carried out include gray processing, binaryzation and noise reduction pretreatment operation after,
The positive sample collection of the pretreatment completion and negative sample collection and the tally set are input to data analysis layer;
Data analysis layer receives the face image set that pretreatment is completed, using the tally set as black eye judgment models
The input value of loss function is input to after the data of the positive sample collection and negative sample collection are carried out the operation of direction gradient histogram
The support vector machines module of the black eye judgment models, the support vector machines module is to the negative sample collection and the positive sample
This collection carries out just classification, and is input to the black eye based on the data that the tally set extracts the just classification error and judges mould
The convolutional neural networks module retraining of type after the convolutional neural networks module is trained, exports training tally set, and will
The trained label is input to the loss function, and the loss function is calculated in conjunction with the trained tally set and the tally set
Output valve, when the output valve is less than preset threshold, the black eye judgment models exit training;
Data in the test set are mapped to higher-dimension sky based on nonlinear mapping method by the test set for receiving user
Between, it is by the black eye judgment models judgement is input to after the test set progress direction gradient histogram operation of the mapping completion
It is no to have black eye, and export result.
Optionally, the data in the positive sample collection be include black-eyed facial image, the number in the negative sample collection
It does not include black-eyed facial image according to being.
Optionally, the self-adapting image denoising filter method are as follows:
G (x, y)=η (x, y)+f (x, y)
Wherein, (x, y) indicate image slices vegetarian refreshments coordinate, f (x, y) be based on self-adapting image denoising filter method to it is described just
Sample set and the negative sample collection carry out the output data after noise reduction process, and η (x, y) is noise, and g (x, y) is the positive sample
Collection and the negative sample collection,For the noise population variance of the positive sample collection and the negative sample collection,For (x, y)
Pixel grey scale mean value,For the pixel grey scale variance of (x, y), L indicates current pixel point coordinate.
Optionally, the convolutional neural networks module includes input layer, convolutional layer, output layer;
The convolutional layer includes convolution operation, pondization operation and activation operation;
The convolution operation are as follows:
Wherein ω ' is output data, and ω is the data of the just classification error, and k is the size of convolution kernel, and s is convolution behaviour
The stride of work, p are data padding matrix;
The activation operation are as follows:
Wherein y is the output valve of the activation operation, and e is nonterminating and non-recurring decimal.
Optionally, the algorithm of support vector machine includes Nonlinear Mapping and constraint solving;
The Nonlinear Mapping are as follows:
Wherein,Indicate the gradient direction noxkata feature (xi,xj) Nonlinear Mapping inner product
It calculates, κ (xi,xj) it is the gradient direction noxkata feature (xi,xj) nonlinear mapping function;
The constraint solving are as follows:
Wherein, αi>=0, i=1,2 ... m
Wherein, m is the quantity of the gradient direction noxkata feature, αi, αjFor the constraint solving Lagrangian number multiply because
Son, yi, yjFor the label of the positive negative sample, s.t is constraint condition.
In addition, to achieve the above object, the present invention also provides a kind of black eye intelligent judging device, which includes storage
Device and processor are stored with the black eye intelligent decision program that can be run on the processor in the memory, described black
Eyelet intelligent decision program realizes following steps when being executed by the processor:
Data receiver layer receives face image set, and the face image set is divided into positive sample collection by tally set and bears
Sample set, the positive sample collection and the negative sample collection are carried out include gray processing, binaryzation and noise reduction pretreatment operation after,
The positive sample collection of the pretreatment completion and negative sample collection and the tally set are input to data analysis layer;
Data analysis layer receives the face image set that pretreatment is completed, using the tally set as black eye judgment models
The input value of loss function is input to after the data of the positive sample collection and negative sample collection are carried out the operation of direction gradient histogram
The support vector machines module of the black eye judgment models, the support vector machines module is to the negative sample collection and the positive sample
This collection carries out just classification, and is input to the black eye based on the data that the tally set extracts the just classification error and judges mould
The convolutional neural networks module retraining of type after the convolutional neural networks module is trained, exports training tally set, and will
The trained label is input to the loss function, and the loss function is calculated in conjunction with the trained tally set and the tally set
Output valve, when the output valve is less than preset threshold, the black eye judgment models exit training;
Data in the test set are mapped to higher-dimension sky based on nonlinear mapping method by the test set for receiving user
Between, it is by the black eye judgment models judgement is input to after the test set progress direction gradient histogram operation of the mapping completion
It is no to have black eye, and export result.
Optionally, the self-adapting image denoising filter method are as follows:
G (x, y)=η (x, y)+f (x, y)
Wherein, (x, y) indicate image slices vegetarian refreshments coordinate, f (x, y) be based on self-adapting image denoising filter method to it is described just
Sample set and the negative sample collection carry out the output data after noise reduction process, and η (x, y) is noise, and g (x, y) is the positive sample
Collection and the negative sample collection,For the noise population variance of the positive sample collection and the negative sample collection,For (x, y)
Pixel grey scale mean value,For the pixel grey scale variance of (x, y), L indicates current pixel point coordinate.
Optionally, the convolutional neural networks module includes input layer, convolutional layer, output layer;
The convolutional layer includes convolution operation, pondization operation and activation operation;
The convolution operation are as follows:
Wherein ω ' is output data, and ω is the data of the just classification error, and k is the size of convolution kernel, and s is convolution behaviour
The stride of work, p are data padding matrix;
The activation operation are as follows:
Wherein y is the output valve of the activation operation, and e is nonterminating and non-recurring decimal.
Optionally, the algorithm of support vector machine includes Nonlinear Mapping and constraint solving;
The Nonlinear Mapping are as follows:
Wherein,Indicate the gradient direction noxkata feature (xi,xj) Nonlinear Mapping inner product
It calculates, κ (xi,xj) it is the gradient direction noxkata feature (xi,xj) nonlinear mapping function;
The constraint solving are as follows:
Wherein, αi>=0, i=1,2 ... m
Wherein, m is the quantity of the gradient direction noxkata feature, αi, αjFor the constraint solving Lagrangian number multiply because
Son, yi, yjFor the label of the positive negative sample, s.t is constraint condition.
In addition, to achieve the above object, it is described computer-readable the present invention also provides a kind of computer readable storage medium
Black eye intelligent decision program is stored on storage medium, the black eye intelligent decision program can be handled by one or more
Device executes, the step of to realize black eye intelligent determination method as described above.
Black eye intelligent determination method, device and computer readable storage medium proposed by the present invention, data receiver layer connect
Face image set is received, the face image set is divided into positive sample collection and negative sample collection by tally set, by the positive sample
Collection and negative sample collection and the tally set are input to data analysis layer;Data analysis layer receives the facial image that pretreatment is completed
Collection, using the tally set as the input value of the loss function of black eye judgment models, by the positive sample collection and negative sample collection
It is input to the support vector machines module of the black eye judgment models, the support vector machines module carries out just classification, and is based on
The data that the tally set extracts the just classification error are input to the convolutional neural networks module of the black eye judgment models
Retraining, until black eye judgment models exit training;The test set of user is received, is based on nonlinear mapping method for the survey
Data in examination collection map to higher dimensional space, are input to after the test set that the mapping is completed is carried out the operation of direction gradient histogram
The black eye judgment models judge whether there is black eye, and export result.Supported since service efficiency of the present invention is higher to
Amount machine model and convolutional neural networks model, and early period reduces making an uproar for influence model judgement based on a variety of data preprocessing methods
Sound, therefore accurately black eye intelligent judgment function may be implemented in the present invention.
Detailed description of the invention
Fig. 1 is the flow diagram for the black eye intelligent determination method that one embodiment of the invention provides;
Fig. 2 is the schematic diagram of internal structure for the black eye intelligent judging device that one embodiment of the invention provides;
The module of black eye intelligent decision program in the black eye intelligent judging device that Fig. 3 provides for one embodiment of the invention
Schematic diagram.
The embodiments will be further described with reference to the accompanying drawings for the realization, the function and the advantages of the object of the present invention.
Specific embodiment
It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not intended to limit the present invention.
The present invention provides a kind of black eye intelligent determination method.Shown in referring to Fig.1, provided for one embodiment of the invention black
The flow diagram of eyelet intelligent determination method.This method can be executed by device, which can be by software and/or hard
Part is realized.
In the present embodiment, black eye intelligent determination method includes:
S1, data receiver layer receive face image set, and the face image set is divided into positive sample collection by tally set
With negative sample collection, to the positive sample collection and the negative sample collection carry out include gray processing, binaryzation and noise reduction pretreatment grasp
After work, the positive sample collection of the pretreatment completion and negative sample collection and the tally set are input to data analysis layer.
The reception of present pre-ferred embodiments data receiver layer includes the facial image of positive sample collection, negative sample collection, tally set
Collect, the data in the positive sample collection all include black-eyed face, and the data in the negative sample collection all include not black
The face of eyelet.
Gray processing described in present pre-ferred embodiments operation be by the data in the positive sample collection and the negative sample from
Rgb format switchs to black-white-gray format, and further, the gray processing operation uses rule of three, i.e., switchs to institute according to such as minor function
State black-white-gray format:
0.30*R+0.59*G+0.11*B
Binarization operation described in present pre-ferred embodiments includes first given threshold, the pixel in the black-white-gray format
When greater than the threshold value, the pixel becomes 255, when the pixel in the black-white-gray format is less than the threshold value, the picture
Element becomes 0, i.e., the described black and white format indicates that the pixel value of the positive sample collection and the negative sample collection is 0 or 255.
Noise reduction described in present pre-ferred embodiments is based on self-adapting image denoising filter method to the black and white formatted data
Carry out noise reduction process, the self-adapting image denoising filter method are as follows:
G (x, y)=η (x, y)+d (x, y)
Wherein, (x, y) indicates that image slices vegetarian refreshments coordinate, f (x, y) are based on self-adapting image denoising filter method to described black
White formatted data carries out the output data after noise reduction process, and η (x, y) is noise, and g (x, y) is the black and white formatted data,For
The noise population variance of the black and white formatted data,For the pixel grey scale mean value of (x, y),For the pixel of (x, y)
Gray variance, L indicate current pixel point coordinate.
S2, data analysis layer receive the face image set that pretreatment is completed, and judge mould using the tally set as black eye
The input value of the loss function of type, it is defeated after the data of the positive sample collection and negative sample collection are carried out the operation of direction gradient histogram
Enter to the support vector machines module of the black eye judgment models, the support vector machines module is to the negative sample collection and described
Positive sample collection carries out just classification, and is input to the black eye based on the data that the tally set extracts the just classification error and sentences
The convolutional neural networks module retraining of disconnected model after the convolutional neural networks module is trained, exports training tally set,
And the trained label is input to the loss function, the loss function is in conjunction with the trained tally set and the tally set
Output valve is calculated, when the output valve is less than preset threshold, the black eye judgment models exit training.
It is defeated after the data progress direction gradient histogram operation of present pre-ferred embodiments, the positive sample collection and negative sample collection
Enter to black eye judgment models, gradient magnitude and ladder including calculating each pixel (x, y) of data in the face image set
Direction value is spent, and using the gradient magnitude as the first component, the gradient direction value forms gradient matrix as second component,
Data in the gradient matrix are divided into multiple fritters, and the gradient magnitude for being added each fritter is added with gradient direction value
Value, and the additive value connected to form gradient direction noxkata feature and be input to black eye judgment models.
The black eye judgment models are based on algorithm of support vector machine to the negative sample collection by present pre-ferred embodiments
It is trained with the positive sample collection, exits instruction when the loss function value in the algorithm of support vector machine is less than threshold value
Practice.The algorithm of support vector machine includes Nonlinear Mapping and constraint solving;
The Nonlinear Mapping are as follows:
Wherein,Indicate the gradient direction noxkata feature (xi,xj) Nonlinear Mapping inner product
It calculates, κ (xi,xj) it is the gradient direction noxkata feature (xi,xj) nonlinear mapping function;
The constraint solving are as follows:
Wherein, αi>=0, i=1,2 ... m
Wherein, m is the quantity of the gradient direction noxkata feature, αi, αjFor the constraint solving Lagrangian number multiply because
Son, yi, yjFor the label of the positive negative sample, s.t is constraint condition.The loss function is least square method, the loss letter
Numerical value is L (e):
Wherein, e is the trained values of the black eye judgment models and the error amount of the tally set, and k is the positive sample
The total quantity of collection and the negative sample collection, yiFor the tally set, y 'iFor the trained values, the threshold value is traditionally arranged to be
0.01。
Present pre-ferred embodiments, the convolutional neural networks module include input layer, convolutional layer, output layer;
The convolutional layer includes convolution operation, pondization operation and activation operation;
The convolution operation are as follows:
Wherein ω ' is output data, and ω is the data of the just classification error, and k is the size of convolution kernel, and s is convolution behaviour
The stride of work, p are data padding matrix;
The activation operation are as follows:
Wherein y is the output valve of the activation operation, and e is nonterminating and non-recurring decimal.
Data in the test set are mapped to higher-dimension based on nonlinear mapping method by S3, the test set for receiving user
Space is input to the black eye judgment models judgement after the test set that the mapping is completed is carried out the operation of direction gradient histogram
Whether there is black eye, and exports result.
Data in the test set are mapped to higher-dimension sky based on nonlinear mapping method by present pre-ferred embodiments
Between, the data mapping uses the nonlinear mapping method of the support vector machines.
Invention also provides a kind of black eye intelligent judging device.Referring to shown in Fig. 2, provided for one embodiment of the invention black
The schematic diagram of internal structure of eyelet intelligent judging device.
In the present embodiment, the black eye intelligent judging device 1 can be PC (Personal Computer, personal electricity
Brain) or terminal devices such as smart phone, tablet computer, portable computer, it is also possible to a kind of server etc..The black eye
It encloses intelligent judging device 1 and includes at least memory 11, processor 12, communication bus 13 and network interface 14.
Wherein, memory 11 include at least a type of readable storage medium storing program for executing, the readable storage medium storing program for executing include flash memory,
Hard disk, multimedia card, card-type memory (for example, SD or DX memory etc.), magnetic storage, disk, CD etc..Memory 11
It can be the internal storage unit of black eye intelligent judging device 1, such as black eye intelligent decision dress in some embodiments
Set 1 hard disk.Memory 11 is also possible to the External memory equipment of black eye intelligent judging device 1 in further embodiments,
Such as the plug-in type hard disk being equipped in black eye intelligent judging device 1, intelligent memory card (Smart Media Card, SMC), peace
Digital (Secure Digital, SD) card, flash card (Flash Card) etc..Further, memory 11 can also be wrapped both
The internal storage unit for including black eye intelligent judging device 1 also includes External memory equipment.Memory 11 can be not only used for depositing
Storage is installed on the application software and Various types of data of black eye intelligent judging device 1, such as the generation of black eye intelligent decision program 01
Code etc., can be also used for temporarily storing the data that has exported or will export.
Processor 12 can be in some embodiments a central processing unit (Central Processing Unit,
CPU), controller, microcontroller, microprocessor or other data processing chips, the program for being stored in run memory 11
Code or processing data, such as execute black eye intelligent decision program 01 etc..
Communication bus 13 is for realizing the connection communication between these components.
Network interface 14 optionally may include standard wireline interface and wireless interface (such as WI-FI interface), be commonly used in
Communication connection is established between the device 1 and other electronic equipments.
Optionally, which can also include user interface, and user interface may include display (Display), input
Unit such as keyboard (Keyboard), optional user interface can also include standard wireline interface and wireless interface.It is optional
Ground, in some embodiments, display can be light-emitting diode display, liquid crystal display, touch-control liquid crystal display and OLED
(Organic Light-Emitting Diode, Organic Light Emitting Diode) touches device etc..Wherein, display can also be appropriate
Referred to as display screen or display unit, for be shown in the information handled in black eye intelligent judging device 1 and for show can
Depending on the user interface changed.
Fig. 2 illustrates only the black eye intelligent judging device with component 11-14 and black eye intelligent decision program 01
1, it will be appreciated by persons skilled in the art that structure shown in fig. 1 does not constitute the limit to black eye intelligent judging device 1
It is fixed, it may include perhaps combining certain components or different component layouts than illustrating less perhaps more components.
In 1 embodiment of device shown in Fig. 2, black eye intelligent decision program 01 is stored in memory 11;Processor
Following steps are realized when the black eye intelligent decision program 01 stored in 12 execution memories 11:
Step 1: data receiver layer receives face image set, the face image set is divided the sample that is positive by tally set
This collection and negative sample collection, the positive sample collection and the negative sample collection are carried out include gray processing, binaryzation and noise reduction pre- place
After reason operation, the positive sample collection of the pretreatment completion and negative sample collection and the tally set are input to data analysis layer.
Present pre-ferred embodiments, the reception of data receiver layer include the facial image of positive sample collection, negative sample collection, tally set
Collect, the data in the positive sample collection all include black-eyed face, and the data in the negative sample collection all include not black
The face of eyelet.
Present pre-ferred embodiments, gray processing operation be by the data in the positive sample collection and the negative sample from
Rgb format switchs to black-white-gray format, and further, the gray processing operation uses rule of three, i.e., switchs to institute according to such as minor function
State black-white-gray format:
0.30*R+0.59*G+0.11*B
Binarization operation described in present pre-ferred embodiments includes first given threshold, the pixel in the black-white-gray format
When greater than the threshold value, the pixel becomes 255, when the pixel in the black-white-gray format is less than the threshold value, the picture
Element becomes 0, i.e., the described black and white format indicates that the pixel value of the positive sample collection and the negative sample collection is 0 or 255.
Present pre-ferred embodiments, the noise reduction are based on self-adapting image denoising filter method to the black and white formatted data
Carry out noise reduction process, the self-adapting image denoising filter method are as follows:
G (x, y)=η (x, y)+f (x, y)
Wherein, (x, y) indicates that image slices vegetarian refreshments coordinate, f (x, y) are based on self-adapting image denoising filter method to described black
White formatted data carries out the output data after noise reduction process, and η (x, y) is noise, and g (x, y) is the black and white formatted data,
For the noise population variance of the black and white formatted data,For the pixel grey scale mean value of (x, y),For the picture of (x, y)
Plain gray variance, L indicate current pixel point coordinate.
Step 2: data analysis layer receives the face image set that pretreatment is completed, sentence using the tally set as black eye
The data of the positive sample collection and negative sample collection are carried out the operation of direction gradient histogram by the input value of the loss function of disconnected model
Afterwards, it is input to the support vector machines module of the black eye judgment models, the support vector machines module is to the negative sample collection
With the positive sample collection carry out just classification, and based on the tally set extract it is described just classification error data be input to it is described black
The convolutional neural networks module retraining of eyelet judgment models, after the convolutional neural networks module is trained, output training
Tally set, and the trained label is input to the loss function, the loss function is in conjunction with the trained tally set and institute
It states tally set and calculates output valve, when the output valve is less than preset threshold, the black eye judgment models exit training.
It is defeated after the data progress direction gradient histogram operation of present pre-ferred embodiments, the positive sample collection and negative sample collection
Enter to black eye judgment models, gradient magnitude and ladder including calculating each pixel (x, y) of data in the face image set
Direction value is spent, and using the gradient magnitude as the first component, the gradient direction value forms gradient matrix as second component,
Data in the gradient matrix are divided into multiple fritters, and the gradient magnitude for being added each fritter is added with gradient direction value
Value, and the additive value connected to form gradient direction noxkata feature and be input to black eye judgment models.
The black eye judgment models are based on algorithm of support vector machine to the negative sample collection by present pre-ferred embodiments
It is trained with the positive sample collection, exits instruction when the loss function value in the algorithm of support vector machine is less than threshold value
Practice.The algorithm of support vector machine includes Nonlinear Mapping and constraint solving;
The Nonlinear Mapping are as follows:
Wherein,Indicate the gradient direction noxkata feature (xi,xj) Nonlinear Mapping inner product
It calculates, κ (xi,xj) it is the gradient direction noxkata feature (xi,xj) nonlinear mapping function;
The constraint solving are as follows:
Wherein, αi>=0, i=1,2 ... m
Wherein, m is the quantity of the gradient direction noxkata feature, αi, αjFor the constraint solving Lagrangian number multiply because
Son, yi, yjFor the label of the positive negative sample, s.t is constraint condition.The loss function is least square method, the loss letter
Numerical value is L (e):
Wherein, e is the trained values of the black eye judgment models and the error amount of the tally set, and k is the positive sample
The total quantity of collection and the negative sample collection, yiFor the tally set, y 'iFor the trained values, the threshold value is traditionally arranged to be
0.01。
Convolutional neural networks module described in present pre-ferred embodiments includes input layer, convolutional layer, output layer;
The convolutional layer includes convolution operation, pondization operation and activation operation;
The convolution operation are as follows:
Wherein ω ' is output data, and ω is the data of the just classification error, and k is the size of convolution kernel, and s is convolution behaviour
The stride of work, p are data padding matrix;
The activation operation are as follows:
Wherein y is the output valve of the activation operation, and e is nonterminating and non-recurring decimal.
Step 3: receiving the test set of user, the data in the test set are mapped to based on nonlinear mapping method
Higher dimensional space is input to the black eye judgment models after the test set that the mapping is completed is carried out the operation of direction gradient histogram
Black eye is judged whether there is, and exports result.
Data in the test set are mapped to higher-dimension sky based on nonlinear mapping method by present pre-ferred embodiments
Between, the data mapping uses the nonlinear mapping method of the support vector machines.
Optionally, in other embodiments, black eye intelligent decision program can also be divided into one or more mould
Block, one or more module are stored in memory 11, and (the present embodiment is processor by one or more processors
12) performed to complete the present invention, the so-called module of the present invention is the series of computation machine program for referring to complete specific function
Instruction segment, for describing implementation procedure of the black eye intelligent decision program in black eye intelligent judging device.
It is the black eye intelligent decision in one embodiment of black eye intelligent judging device of the present invention for example, referring to shown in Fig. 3
The program module schematic diagram of program, in the embodiment, the black eye intelligent decision program can be divided into data reception
Block 10, model training module 20, black eye judgment module 30 be illustratively:
The data reception module 10 is used for: receiving face image set, the face image set is divided by tally set
Be positive sample set and negative sample collection, carries out including gray processing, binaryzation and noise reduction to the positive sample collection and the negative sample collection
Pretreatment operation after, positive sample collection and negative sample collection and the tally set that the pretreatment is completed are input at data
Manage layer.
The model training module 20 is used for: the face image set that pretreatment is completed is received, using the tally set as black
The data of the positive sample collection and negative sample collection are carried out direction gradient histogram by the input value of the loss function of eyelet judgment models
After operation, it is input to the support vector machines module of the black eye judgment models, the support vector machines module is to the negative sample
This collection and the positive sample collection carry out just classification, and are input to institute based on the data that the tally set extracts the just classification error
The convolutional neural networks module retraining for stating black eye judgment models, after the convolutional neural networks module is trained, output
Training tally set, and the trained label is input to the loss function, the loss function is in conjunction with the trained tally set
Output valve is calculated with the tally set, when the output valve is less than preset threshold, the black eye judgment models exit training.
The black eye judgment module 30 is used for: being received the test set of user, is based on nonlinear mapping method for the survey
Data in examination collection map to higher dimensional space, are input to after the test set that the mapping is completed is carried out the operation of direction gradient histogram
The black eye judgment models judge whether there is black eye, and export result.
The program modules such as above-mentioned data reception module 10, model training module 20, black eye judgment module 30 are performed
Functions or operations step and the above-described embodiment realized are substantially the same, and details are not described herein.
In addition, the embodiment of the present invention also proposes a kind of computer readable storage medium, the computer readable storage medium
On be stored with black eye intelligent decision program, the black eye intelligent decision program can be executed by one or more processors, with
Realize following operation:
Face image set is received, the face image set is divided into positive sample collection and negative sample collection by tally set, right
The positive sample collection and the negative sample collection carry out include gray processing, binaryzation and noise reduction pretreatment operation after, will be described pre-
It handles the positive sample collection completed and negative sample collection and the tally set is input to data analysis layer;
The face image set that pretreatment is completed is received, using the tally set as the loss function of black eye judgment models
Input value is input to the black eye after the data of the positive sample collection and negative sample collection are carried out the operation of direction gradient histogram
The support vector machines module of judgment models, the support vector machines module carry out just the negative sample collection and the positive sample collection
Classification, and it is refreshing based on the convolution that the data that the tally set extracts the just classification error are input to the black eye judgment models
Through network module retraining, after the convolutional neural networks module is trained, training tally set is exported, and the training is marked
Label are input to the loss function, and the loss function calculates output valve in conjunction with the trained tally set and the tally set, when
When the output valve is less than preset threshold, the black eye judgment models exit training;
Data in the test set are mapped to higher-dimension sky based on nonlinear mapping method by the test set for receiving user
Between, it is by the black eye judgment models judgement is input to after the test set progress direction gradient histogram operation of the mapping completion
It is no to have black eye, and export result.
Computer readable storage medium specific embodiment of the present invention and above-mentioned black eye intelligent judging device and method are each
Embodiment is essentially identical, does not make tired state herein.
It should be noted that the serial number of the above embodiments of the invention is only for description, do not represent the advantages or disadvantages of the embodiments.And
The terms "include", "comprise" herein or any other variant thereof is intended to cover non-exclusive inclusion, so that packet
Process, device, article or the method for including a series of elements not only include those elements, but also including being not explicitly listed
Other element, or further include for this process, device, article or the intrinsic element of method.Do not limiting more
In the case where, the element that is limited by sentence "including a ...", it is not excluded that including process, device, the article of the element
Or there is also other identical elements in method.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment side
Method can be realized by means of software and necessary general hardware platform, naturally it is also possible to by hardware, but in many cases
The former is more preferably embodiment.Based on this understanding, technical solution of the present invention substantially in other words does the prior art
The part contributed out can be embodied in the form of software products, which is stored in one as described above
In storage medium (such as ROM/RAM, magnetic disk, CD), including some instructions are used so that terminal device (it can be mobile phone,
Computer, server or network equipment etc.) execute method described in each embodiment of the present invention.
The above is only a preferred embodiment of the present invention, is not intended to limit the scope of the invention, all to utilize this hair
Equivalent structure or equivalent flow shift made by bright specification and accompanying drawing content is applied directly or indirectly in other relevant skills
Art field, is included within the scope of the present invention.
Claims (10)
1. a kind of black eye intelligent determination method, which is characterized in that the described method includes:
Data receiver layer receives face image set, and the face image set is divided into positive sample collection and negative sample by tally set
Collection, the positive sample collection and the negative sample collection are carried out include gray processing, binaryzation and noise reduction pretreatment operation after, by institute
It states the positive sample collection for pre-processing completion and negative sample collection and the tally set is input to data analysis layer;
Data analysis layer receives the face image set that pretreatment is completed, using the tally set as the loss of black eye judgment models
The input value of function after the data of the positive sample collection and negative sample collection are carried out the operation of direction gradient histogram, is input to described
The support vector machines module of black eye judgment models, the support vector machines module is to the negative sample collection and the positive sample collection
Just classification is carried out, and is input to the black eye judgment models based on the data that the tally set extracts the just classification error
Convolutional neural networks module retraining after the convolutional neural networks module is trained, exports training tally set, and will be described
Training label is input to the loss function, and the loss function is calculated in conjunction with the trained tally set and the tally set and exported
Value, when the output valve is less than preset threshold, the black eye judgment models exit training;
Data in the test set are mapped to higher dimensional space based on nonlinear mapping method by the test set for receiving user, will
The black eye judgment models are input to after the test set progress direction gradient histogram operation that the mapping is completed to judge whether there is
Black eye, and export result.
2. black eye intelligent determination method as described in claim 1, which is characterized in that the data in the positive sample collection are packets
Include black-eyed facial image, the data in the negative sample collection be do not include black-eyed facial image.
3. the black eye intelligent determination method as described in claim 1 to 2, which is characterized in that the self-adapting image denoising filter
Wave method are as follows:
G (x, y)=η (x, y)+f (x, y)
Wherein, (x, y) indicates that image slices vegetarian refreshments coordinate, f (x, y) are based on self-adapting image denoising filter method to the positive sample
Collection and the negative sample collection carry out noise reduction process after output data, η (x, y) be noise, g (x, y) be the positive sample collection and
The negative sample collection,For the noise population variance of the positive sample collection and the negative sample collection,For the pixel of (x, y)
Gray average,For the pixel grey scale variance of (x, y), L indicates current pixel point coordinate.
4. black eye intelligent determination method as claimed in claim 1, which is characterized in that the convolutional neural networks module includes defeated
Enter layer, convolutional layer, output layer;
The convolutional layer includes convolution operation, pondization operation and activation operation;
The convolution operation are as follows:
Wherein ω ' is output data, and ω is the data of the just classification error, and k is the size of convolution kernel, and s is convolution operation
Stride, p are data padding matrix;
The activation operation are as follows:
Wherein y is the output valve of the activation operation, and e is nonterminating and non-recurring decimal.
5. black eye intelligent determination method as described in claim 1, which is characterized in that the algorithm of support vector machine includes non-
Linear Mapping and constraint solving;
The Nonlinear Mapping are as follows:
κ(xi,xj)=< θ (xi),θ(xj)>
Wherein, < θ (xi),θ(xj) > indicate the gradient direction noxkata feature (xi,xj) Nonlinear Mapping inner product calculate, κ (xi,
xj) it is the gradient direction noxkata feature (xi,xj) nonlinear mapping function;
The constraint solving are as follows:
Wherein, αi>=0, i=1,2 ... m
Wherein, m is the quantity of the gradient direction noxkata feature, αi, αjMultiply the factor for the Lagrangian number of the constraint solving,
yi, yjFor the label of the positive negative sample, s.t is constraint condition.
6. a kind of black eye intelligent judging device, which is characterized in that described device includes memory and processor, the memory
On be stored with the black eye intelligent decision program that can be run on the processor, the black eye intelligent decision program is described
Processor realizes following steps when executing:
Data receiver layer receives face image set, and the face image set is divided into positive sample collection and negative sample by tally set
Collection, the positive sample collection and the negative sample collection are carried out include gray processing, binaryzation and noise reduction pretreatment operation after, by institute
It states the positive sample collection for pre-processing completion and negative sample collection and the tally set is input to data analysis layer;
Data analysis layer receives the face image set that pretreatment is completed, using the tally set as the loss of black eye judgment models
The input value of function after the data of the positive sample collection and negative sample collection are carried out the operation of direction gradient histogram, is input to described
The support vector machines module of black eye judgment models, the support vector machines module is to the negative sample collection and the positive sample collection
Just classification is carried out, and is input to the black eye judgment models based on the data that the tally set extracts the just classification error
Convolutional neural networks module retraining after the convolutional neural networks module is trained, exports training tally set, and will be described
Training label is input to the loss function, and the loss function is calculated in conjunction with the trained tally set and the tally set and exported
Value, when the output valve is less than preset threshold, the black eye judgment models exit training;
Data in the test set are mapped to higher dimensional space based on nonlinear mapping method by the test set for receiving user, will
The black eye judgment models are input to after the test set progress direction gradient histogram operation that the mapping is completed to judge whether there is
Black eye, and export result.
7. black eye intelligent judging device as claimed in claim 6, which is characterized in that the data in the positive sample collection are packets
Include black-eyed facial image, the data in the negative sample collection be do not include black-eyed facial image.
8. black eye intelligent judging device as claimed in claims 6 or 7, which is characterized in that the self-adapting image denoising filter
Wave method are as follows:
G (x, y)=η (x, y)+f (x, y)
Wherein, (x, y) indicates that image slices vegetarian refreshments coordinate, f (x, y) are based on self-adapting image denoising filter method to the positive sample
Collection and the negative sample collection carry out noise reduction process after output data, η (x, y) be noise, g (x, y) be the positive sample collection and
The negative sample collection,For the noise population variance of the positive sample collection and the negative sample collection,For the pixel of (x, y)
Gray average,For the pixel grey scale variance of (x, y), L indicates current pixel point coordinate.
9. black eye intelligent judging device as claimed in claim 6, which is characterized in that the convolutional neural networks module includes
Input layer, convolutional layer, output layer;
The convolutional layer includes convolution operation, pondization operation and activation operation;
The convolution operation are as follows:
Wherein ω ' is output data, and ω is the data of the just classification error, and k is the size of convolution kernel, and s is convolution operation
Stride, p are data padding matrix;
The activation operation are as follows:
Wherein y is the output valve of the activation operation, and e is nonterminating and non-recurring decimal.
10. a kind of computer readable storage medium, which is characterized in that be stored with black eye on the computer readable storage medium
Intelligent decision program, the black eye intelligent decision program can be executed by one or more processor, to realize as right is wanted
Described in asking any one of 1 to 5 the step of black eye intelligent determination method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910370451.7A CN110222571B (en) | 2019-05-06 | 2019-05-06 | Intelligent judgment method and device for black eye and computer readable storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910370451.7A CN110222571B (en) | 2019-05-06 | 2019-05-06 | Intelligent judgment method and device for black eye and computer readable storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110222571A true CN110222571A (en) | 2019-09-10 |
CN110222571B CN110222571B (en) | 2023-04-07 |
Family
ID=67820327
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910370451.7A Active CN110222571B (en) | 2019-05-06 | 2019-05-06 | Intelligent judgment method and device for black eye and computer readable storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110222571B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110853047A (en) * | 2019-10-12 | 2020-02-28 | 平安科技(深圳)有限公司 | Intelligent image segmentation and classification method and device and computer readable storage medium |
CN111428552A (en) * | 2019-12-31 | 2020-07-17 | 深圳数联天下智能科技有限公司 | Black eye recognition method and device, computer equipment and storage medium |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2007279799A (en) * | 2006-04-03 | 2007-10-25 | Seiko Epson Corp | Image exchange device, image exchange method and image exchange processing program |
CN104850825A (en) * | 2015-04-18 | 2015-08-19 | 中国计量学院 | Facial image face score calculating method based on convolutional neural network |
CN105739127A (en) * | 2016-05-06 | 2016-07-06 | 京东方科技集团股份有限公司 | Intelligent glasses and eye fatigue recognition method |
CN106650688A (en) * | 2016-12-30 | 2017-05-10 | 公安海警学院 | Eye feature detection method, device and recognition system based on convolutional neural network |
US20180114056A1 (en) * | 2016-10-25 | 2018-04-26 | Vmaxx, Inc. | Vision Based Target Tracking that Distinguishes Facial Feature Targets |
CN107977623A (en) * | 2017-11-30 | 2018-05-01 | 睿视智觉(深圳)算法技术有限公司 | A kind of robustness human eye state determination methods |
WO2018188534A1 (en) * | 2017-04-14 | 2018-10-18 | 深圳市商汤科技有限公司 | Face image processing method and device, and electronic device |
CN108985159A (en) * | 2018-06-08 | 2018-12-11 | 平安科技(深圳)有限公司 | Human-eye model training method, eye recognition method, apparatus, equipment and medium |
WO2019033525A1 (en) * | 2017-08-17 | 2019-02-21 | 平安科技(深圳)有限公司 | Au feature recognition method, device and storage medium |
CN109376575A (en) * | 2018-08-20 | 2019-02-22 | 奇酷互联网络科技(深圳)有限公司 | Method, mobile terminal and the storage medium that human body in image is beautified |
-
2019
- 2019-05-06 CN CN201910370451.7A patent/CN110222571B/en active Active
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2007279799A (en) * | 2006-04-03 | 2007-10-25 | Seiko Epson Corp | Image exchange device, image exchange method and image exchange processing program |
CN104850825A (en) * | 2015-04-18 | 2015-08-19 | 中国计量学院 | Facial image face score calculating method based on convolutional neural network |
CN105739127A (en) * | 2016-05-06 | 2016-07-06 | 京东方科技集团股份有限公司 | Intelligent glasses and eye fatigue recognition method |
US20180114056A1 (en) * | 2016-10-25 | 2018-04-26 | Vmaxx, Inc. | Vision Based Target Tracking that Distinguishes Facial Feature Targets |
CN106650688A (en) * | 2016-12-30 | 2017-05-10 | 公安海警学院 | Eye feature detection method, device and recognition system based on convolutional neural network |
WO2018188534A1 (en) * | 2017-04-14 | 2018-10-18 | 深圳市商汤科技有限公司 | Face image processing method and device, and electronic device |
WO2019033525A1 (en) * | 2017-08-17 | 2019-02-21 | 平安科技(深圳)有限公司 | Au feature recognition method, device and storage medium |
CN107977623A (en) * | 2017-11-30 | 2018-05-01 | 睿视智觉(深圳)算法技术有限公司 | A kind of robustness human eye state determination methods |
CN108985159A (en) * | 2018-06-08 | 2018-12-11 | 平安科技(深圳)有限公司 | Human-eye model training method, eye recognition method, apparatus, equipment and medium |
CN109376575A (en) * | 2018-08-20 | 2019-02-22 | 奇酷互联网络科技(深圳)有限公司 | Method, mobile terminal and the storage medium that human body in image is beautified |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110853047A (en) * | 2019-10-12 | 2020-02-28 | 平安科技(深圳)有限公司 | Intelligent image segmentation and classification method and device and computer readable storage medium |
CN110853047B (en) * | 2019-10-12 | 2023-09-15 | 平安科技(深圳)有限公司 | Intelligent image segmentation and classification method, device and computer readable storage medium |
CN111428552A (en) * | 2019-12-31 | 2020-07-17 | 深圳数联天下智能科技有限公司 | Black eye recognition method and device, computer equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN110222571B (en) | 2023-04-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107633204B (en) | Face occlusion detection method, apparatus and storage medium | |
CN111260665B (en) | Image segmentation model training method and device | |
CN110349147B (en) | Model training method, fundus macular region lesion recognition method, device and equipment | |
CN109492612A (en) | Fall detection method and its falling detection device based on skeleton point | |
CN112162930B (en) | Control identification method, related device, equipment and storage medium | |
CN111382714B (en) | Image detection method, device, terminal and storage medium | |
CN107145833A (en) | The determination method and apparatus of human face region | |
CN108038469A (en) | Method and apparatus for detecting human body | |
CN110135421A (en) | Licence plate recognition method, device, computer equipment and computer readable storage medium | |
CN107679448A (en) | Eyeball action-analysing method, device and storage medium | |
CN110415212A (en) | Abnormal cell detection method, device and computer readable storage medium | |
CN108280455A (en) | Human body critical point detection method and apparatus, electronic equipment, program and medium | |
CN109919179A (en) | Aneurysms automatic testing method, device and computer readable storage medium | |
CN110163204A (en) | Businessman's monitoring and managing method, device and storage medium based on image recognition | |
CN109871751A (en) | Attitude appraisal procedure, device and storage medium based on facial expression recognition | |
CN108491866A (en) | Porny identification method, electronic device and readable storage medium storing program for executing | |
CN107944398A (en) | Based on depth characteristic association list diagram image set face identification method, device and medium | |
CN111126347B (en) | Human eye state identification method, device, terminal and readable storage medium | |
CN108898587A (en) | Image processing method, picture processing unit and terminal device | |
CN109255768A (en) | Image completion method, apparatus, terminal and computer readable storage medium | |
CN109543773A (en) | Image processing method, device, medium and electronic equipment | |
US20190034787A1 (en) | Relative position encoding based networks for action recognition | |
CN108961267A (en) | Image processing method, picture processing unit and terminal device | |
CN110610191A (en) | Elevator floor identification method and device and terminal equipment | |
CN112016617B (en) | Fine granularity classification method, apparatus and computer readable storage medium |
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 |