CN103778409A - Human face identification method based on human face characteristic data mining and device - Google Patents
Human face identification method based on human face characteristic data mining and device Download PDFInfo
- Publication number
- CN103778409A CN103778409A CN201410002094.6A CN201410002094A CN103778409A CN 103778409 A CN103778409 A CN 103778409A CN 201410002094 A CN201410002094 A CN 201410002094A CN 103778409 A CN103778409 A CN 103778409A
- Authority
- CN
- China
- Prior art keywords
- face
- data mining
- characteristic
- value
- class
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention relates to a human face identification method based on a human face characteristic data mining technology and a device. Through the special system structure design of a human face registration module, a human face identification module and a system updating module and a specific algorithm, effective intra-class registration human face characteristic information volume is increased, thus the method and the device adapt to the change of human face pose and facial expression. Through automatic system updating, identification rejection rate caused by the fact that an environment and a human face change with time is reduced. For the characteristic differences between types of intra-class human faces, the respective classification threshold is mined. The delay rate caused by the fact that the same classification threshold is used and intra-class human face differences are ignored and the identification rejection rate are reduced. The identification efficiency of a system is improved.
Description
Technical field
The present invention discloses a kind of face recognition technology, and particularly a kind of face identification method and device based on face characteristic data mining technology, belongs to recognition of face field.
Background technology
Face recognition technology is a kind of biometrics identification technology that utilizes face characteristic information to carry out identity discriminating, have contactless collection, can hidden operation, the strong and image acquisition low cost and other advantages of convenient and swift, powerful trace ability afterwards, interactivity, be widely used in the fields such as information security, video monitoring, criminal detection, public safety, social amusement.Existing face identification system generally comprises two kinds of mode of operations, be face registration mode and recognition of face pattern, it adopts 6 steps conventionally, is respectively: face detection, face key point location, people's face shape and texture normalization, feature extraction, Feature Dimension Reduction, template comparison (classification).
Along with the popularization of application in practice, problem and difficult point that recognition of face exists also highlight gradually.Under insecure collection environment, user's face performance is ever-changing, and its complexity is far beyond the picture obtaining in standard evaluation and test.Block and can cause face characteristic to change, can increase the difficulty of recognition of face, increase reject rate; The variation of illumination, attitude and expression, also can cause the face characteristic under a people's varying environment widely different, has reduced discrimination, also likely occurs wrong identification.Therefore, the variation that how to solve the face characteristic causing due to factors such as illumination, attitude, expression and time variations has improved reject rate and misclassification rate and has reduced discrimination, and have higher arithmetic speed becomes problem in the urgent need to address simultaneously.
In order to allow face register the more change information of plurality of human faces feature, improving owing to blocking, the identification problem brought of the variation of the face characteristic that causes of the factor such as illumination, attitude and expression shape change, prior art (application number: 200710163907) provide a kind of same people the registration appraisal procedure of identical photo.
Please refer to Fig. 1, at step S101, first input image to be registered.Then, carrying out face at step S102 detects.Step S103 carries out eyes location to detecting the human face region obtaining.Step S104 is human face region segmentation.Step S105 is face feature extraction unit.S106 is assessment unit.S107 is storage unit, i.e. face characteristic registration.
Incorporated by reference to reference to shown in Fig. 2, first step S201 judges whether the people under image to be registered registered in face identification system; If judge that the people under image to be registered does not register in face identification system, carries out step S202 registration; If judge that the people under image to be registered registered in face identification system, step S203 calculates the similarity between image to be registered and enrolled images, forms all image comparison of cycle criterion complete with step 204; The maximal value of similarity is calculated in step S205 sequence, and step S206 compares the maximal value of similarity and predetermined threshold value, to judge whether that treating registered images carries out step S202 registration.
From Fig. 1 and in conjunction with shown in Fig. 2, the above-mentioned registration assessment unit of prior art is only effective to identical photo in registered class, and to be tending towards the mistake identification problem of principle of similar and existing face recognition technology existence still unresolved for facial image between similar class and because attitude, illumination, expression and time such as change at the face characteristic between the class of bringing.
Prior art (number of patent application: 201210195701.6) for merging the face identification method of sparse Preserving map and multi-class attribute Bagging algorithm.In the time realizing attribute Bagging algorithm, the angle converting with Radon attribute as an example builds training sample, and carries out the training of base sorter thereon.Test at Yale face database, everyone chooses 6 pictures as training set, and remaining 5 as test set, best identified rate is iteration 20 times, and discrimination is 81%.Test at AR face database, everyone chooses 7 pictures as training set, and remaining 7 as test set, best identified rate is iteration 10 times, and discrimination is 98.29%.
Prior art in the test result in AR storehouse significantly better than the recognition result in YALE storehouse, illustrate that prior art is more applicable for there is single expression or the situation of illumination, and lower with the situation discrimination of illumination variation for there is expression simultaneously, this is also the often lower universal phenomenon of recognition efficiency in the time that expression and illumination exist simultaneously of existing face recognition products.Simultaneously, the repeatedly iteration of prior art easily causes model over-fitting, after test face corresponding to registration face feature database changes to some extent, form the rising of situation rate and misclassification rate, lack generalization ability, and the increase of recognition time that the repeatedly iteration of prior art is brought and the reduction of efficiency, be unfavorable for technology practical application.
Summary of the invention
For the above-mentioned problem existing in face recognition application of the prior art of mentioning, the invention provides a kind of face identification method and device based on face characteristic data mining technology, it is by special system structure design and algorithm, realize and increase registration face characteristic information amount in effective class, better human face posture and the expression shape change of adapting to, and the automatic renewal of system can reduce the reject rate that environment and face temporal evolution bring, also excavate classification thresholds separately for the feature difference of all kinds of faces between class, reduced use same classification thresholds and between ignore class face difference produce the rate of causing delay and reject rate, the recognition efficiency of raising system.
The technical scheme that the present invention solves its technical matters employing is: a kind of face identification device based on face characteristic data mining, this device comprises face Registering modules, face recognition module and system update module, wherein
Described face Registering modules utilizes image local dual mode characteristic similarity matching value, judge the identical image in class and give up according to matching value, calculate core principle component analysis feature for non-identical photo, and use fuzzy FDA to extract face characteristic to core principle component analysis feature, recycling data mining technology is asked for such people's classification thresholds, and records this registrant's log-in password;
Described face recognition module, by computed segmentation image core principle component analysis feature, and use fuzzy FDA to extract face characteristic to core principle component analysis feature, the Euclidean distance of face characteristic in the current face that calculating is extracted again and storehouse, and each classification thresholds of this value and registry is compared to judgement;
Described system update module is continuous 3 all unidentified successes of current identification recognition of face, and identification mark is in claimed range and be same class people, confirm identification by log-in password, Password Input is correct, identify successfully the go forward side by side feature extraction of pedestrian's face and data mining and obtain all kinds of people's of registry classification thresholds, and upgrade face database.
The face identification method based on face characteristic data mining that adopts above-mentioned device, the method is
Described face Registering modules comprises that face location, identical face are removed, face characteristic extracts and data mining;
Described face recognition module comprises face location, and face characteristic extracts and recognition of face;
Described face database system update module comprises system update judgement, and face characteristic extracts and data mining.
The technical scheme that the present invention solves its technical matters employing further comprises:
The human face region that described face is orientated as detecting carries out nose location, according to nose locating information, face is partitioned into 4 human face regions such as face eyebrow image, face eye image, face nose image and face face image.
Described identical human face photo is removed the LBP feature for extracting respectively the human face region that face positioning step cuts apart, and the hamming distance of face in current face local binary patterns feature and current face database will be calculated, if hamming distance is less than setting threshold, is judged to be identical photo and removes.
Described face characteristic is extracted as for non-identical photo and calculates core principle component analysis feature, and use fuzzy FLDA to obtain projection coefficient to core principle component analysis feature, and extract face characteristic,
Described recognition of face is to calculate the Euclidean distance of certain face characteristic in the current face that extracts and storehouse, and this face class of this value and registry is obtained to classification thresholds compares judgement.
Described data mining comprises the steps:
First, set up face characteristic classification thresholds data mining model,
Thresholdi is classification thresholds corresponding to every class people, i is the face classification number of face database, a and b are respectively gain coefficient and the index coefficient a value 0 ~ 6 of model, b value 0.5 ~ 2, in the Xi class that to be each face characteristic in similar identify with everyone except self in face database, the most similar identification mark forms;
Then, obtain model and enter ginseng, utilize Euclidean distance as distance measure, calculate in each face characteristic in similar and face database everyone except self and identify, the most similar identification mark in the class of at every turn identifying is formed to one-dimension array Xi;
Finally, obtain Data Mining Classification threshold value, change the value of a and b, obtain manifold classification threshold value, FAR with in FRR curve map, get FAR and FRR below area and b value corresponding to minimum image, and in this figure, get FRR/FAR and approach 1/5 o'clock a value in corresponding threshold value most, with this (a, b) value in conjunction with the Xi of each face class, obtains the Classification and Identification threshold value of all kinds of faces in substitution data mining model.
The invention has the beneficial effects as follows: the present invention not only can increase registration face characteristic information amount in effective class, better human face posture and the expression shape change of adapting to, and the automatic renewal of system can reduce the reject rate that environment and face temporal evolution bring, also excavate classification thresholds separately for the feature difference of all kinds of faces between class, reduce and used same classification thresholds and the rate of causing delay and reject rate that between ignore class, face difference produces, improved the recognition efficiency of system.
Below in conjunction with the drawings and specific embodiments, the present invention will be further described.
Accompanying drawing explanation
Fig. 1 is the process flow diagram of existing face registration process.
Fig. 2 is the process flow diagram of assessment unit.
Fig. 3 is system flowchart of the present invention.
Fig. 4 is that the present invention does not comprise the test result figure of data mining part in AR storehouse.
Fig. 5 is the FAR(false acceptance rate that Fig. 4 is corresponding) and FRR(false rejection rate) figure.
Fig. 6 is the FAR(false acceptance rate of the present invention in AR storehouse experimental result) and FRR(false rejection rate) figure.
Embodiment
The present embodiment is the preferred embodiment for the present invention, and other all its principles are identical with the present embodiment or approximate with basic structure, all within protection domain of the present invention.
The face identification device based on face characteristic data mining in the present invention is made up of face registration, recognition of face and 3 modules of system update, wherein, face Registering modules utilizes image local dual mode characteristic similarity matching value, judges the identical image in class and gives up according to matching value.Calculate core principle component analysis (KPCA) feature for non-identical photo, and use fuzzy FDA to extract face characteristic to KPCA feature, recycling data mining technology is asked for such people's classification thresholds, and records this registrant's log-in password.Face recognition module, by computed segmentation image core principle component analysis (KPCA) feature, and use fuzzy FDA to extract face characteristic to KPCA feature, then calculate the Euclidean distance of face characteristic in the current face that extracts and storehouse, and each classification thresholds of this value and registry is compared to judgement.System update module is continuous 3 all unidentified successes of current identification recognition of face, and identification mark is in claimed range and be same class people, confirms to identify by log-in password.Password Input is correct, identifies successfully the go forward side by side feature extraction of pedestrian's face and data mining and obtains all kinds of people's of registry classification thresholds, and upgrade face database.The present invention not only can increase registration face characteristic information amount in effective class, better human face posture and the expression shape change of adapting to, and the automatic renewal of system can reduce the reject rate that environment and face temporal evolution bring, also excavate classification thresholds separately for the feature difference of all kinds of faces between class, reduce and used same classification thresholds and the rate of causing delay and reject rate that between ignore class, face difference produces, improved the recognition efficiency of system.
In the present invention, protect a kind of face identification method based on face characteristic data mining, wherein, face Registering modules comprises, the steps such as face location, identical face removal, face characteristic extraction and data mining simultaneously:
The first step, face location.First the human face region detecting is carried out to nose location, according to nose locating information, face is partitioned into 4 human face regions such as face eyebrow image, face eye image, face nose image and face face image;
Second step, identical human face photo is removed.First extract respectively the LBP feature of the human face region that face positioning step cuts apart, and will calculate the hamming distance of face in current face local binary patterns (LBP, Local Binary Pattern) feature and current face database.If hamming distance is less than setting threshold, is judged to be identical photo and removes;
The 3rd step, face characteristic extracts.Calculate core principle component analysis (KPCA) feature for non-identical photo, use fuzzy FLDA to obtain projection coefficient to KPCA feature, and extract face characteristic.
The 4th step, data mining.
In the present embodiment, face characteristic data mining technology comprises following characteristics:
First, set up face characteristic classification thresholds data mining model.
Thresholdi is classification thresholds corresponding to every class people, the face classification number that i is face database, and a and b are respectively gain coefficient and the index coefficient a value 0 ~ 6 of model, b value 0.5 ~ 2.In the Xi class that to be each face characteristic in similar identify with everyone except self in face database, the most similar identification mark forms.
Then, obtain model and enter ginseng.Utilize Euclidean distance as distance measure, calculate in each face characteristic in similar and face database everyone except self and identify, the most similar identification mark in the class of at every turn identifying is formed to one-dimension array Xi.
Finally, obtain Data Mining Classification threshold value.Change the value of a and b, obtain manifold classification threshold value, FAR with in FRR curve map, get FAR and FRR below area and b value corresponding to minimum image, and in this figure, get FRR/FAR and approach 1/5 o'clock a value in corresponding threshold value most, with this (a, b) value in conjunction with the Xi of each face class, obtains the Classification and Identification threshold value of all kinds of faces in substitution data mining model
.
Face recognition module in the present embodiment comprises face location, and face characteristic extracts and recognition of face, and it is specific as follows:
The first step, face location.First the human face region detecting is carried out to nose location, according to nose locating information, face is partitioned into 4 human face regions such as face eyebrow image, face eye image, face nose image and face face image;
Second step, face characteristic extracts.Computed segmentation image core principle component analysis (KPCA) feature, and adopt KPCA feature to use the fuzzy FDA projection coefficient of described Registering modules to extract face characteristic.
The 3rd step, calculates the Euclidean distance of certain face characteristic in the current face that extracts and storehouse, and this face class of this value and registry is obtained to classification thresholds compares judgement.
In the present embodiment, face database system update module comprises system update judgement, and face characteristic extracts and data mining, specific as follows:
The first step, system update is judged.When system and device is in face recognition module operational process, continuous 3 all unidentified successes of current identification recognition of face, and recognition result is that identification mark is greater than the same class people of setting threshold, needs to input log-in password and confirms identification.Password Input is correct, identifies successfully, and can enter system update.
Second step, face characteristic extracts.Adopt current identification recognition of face continuous 3 times all unidentified successfully the most similar current human face region substitute in 3 identification human face region in the most dissimilar storehouse, recalculate core principle component analysis (KPCA) feature of human face region in storehouse, use fuzzy FLDA to obtain projection coefficient to KPCA feature, and extract face characteristic.
The 3rd step, data mining.
In the present embodiment, face characteristic data mining technology comprises following characteristics:
First, set up face characteristic classification thresholds data mining model.
Thresholdi is classification thresholds corresponding to every class people, the face classification number that i is face database, and a and b are respectively gain coefficient and the index coefficient a value 0 ~ 6 of model, b value 0.5 ~ 2.In the Xi class that to be each face characteristic in similar identify with everyone except self in face database, the most similar identification mark forms.
Then, obtain model and enter ginseng.Utilize Euclidean distance as distance measure, calculate in each face characteristic in similar and face database everyone except self and identify, the most similar identification mark in the class of at every turn identifying is formed to one-dimension array Xi.
Finally, obtain Data Mining Classification threshold value.Change the value of a and b, obtain manifold classification threshold value, FAR with in FRR curve map, get FAR and FRR below area and b value corresponding to minimum image, and in this figure, get FRR/FAR and approach 1/5 o'clock a value in corresponding threshold value most, with this (a, b) value in conjunction with the Xi of each face class, obtains the Classification and Identification threshold value of all kinds of faces in substitution data mining model
.
For above-mentioned purpose of the present invention, feature and advantage can be become apparent more, below in conjunction with the drawings and specific embodiments, the present invention is further detailed explanation.
Please refer to accompanying drawing 3, the invention provides a kind of face identification method and device based on face characteristic data mining technology, it comprises: face Registering modules, face recognition module and face database system update module.
Described face Registering modules, comprises, face positioning step, and identical face is removed step, face characteristic extraction step and data mining step etc.:
Step S301, face location.First the human face region detecting is carried out to nose location, according to nose locating information, face is partitioned into 4 human face regions such as face eyebrow image, face eye image, face nose image and face face image;
Step S302, identical human face photo is removed.First extract respectively the LBP feature of the human face region that face positioning step cuts apart, and will calculate the hamming distance of face in current face local binary patterns (LBP, Local Binary Pattern) feature and current face database.If hamming distance is less than setting threshold, is judged to be identical photo and removes;
Step S303, face characteristic extracts.Calculate core principle component analysis (KPCA) feature for non-identical photo, use fuzzy FLDA to obtain projection coefficient to KPCA feature, and extract face characteristic.
Step S304, data mining.Described face characteristic data mining technology comprises following characteristics:
First, set up face characteristic classification thresholds data mining model.
Thresholdi is classification thresholds corresponding to every class people, the face classification number that i is face database, and a and b are respectively gain coefficient and the index coefficient a value 0 ~ 6 of model, b value 0.5 ~ 2.In the Xi class that to be each face characteristic in similar identify with everyone except self in face database, the most similar identification mark forms.
Then, obtain model and enter ginseng.Utilize Euclidean distance as distance measure, calculate in each face characteristic in similar and face database everyone except self and identify, the most similar identification mark in the class of at every turn identifying is formed to one-dimension array Xi.
Finally, obtain Data Mining Classification threshold value.Change the value of a and b, obtain manifold classification threshold value, FAR with in FRR curve map, get FAR and FRR below area and b value corresponding to minimum image, and in this figure, get FRR/FAR and approach 1/5 o'clock a value in corresponding threshold value most, with this (a, b) value in conjunction with the Xi of each face class, obtains the Classification and Identification threshold value of all kinds of faces in substitution data mining model
.
Face recognition module comprises face location, and face characteristic extracts and recognition of face:
Step S305, face location.First the human face region detecting is carried out to nose location, according to nose locating information, face is partitioned into 4 human face regions such as face eyebrow image, face eye image, face nose image and face face image;
Step S306, face characteristic extracts.Computed segmentation image core principle component analysis (KPCA) feature, and adopt KPCA feature to use the fuzzy FDA projection coefficient of described Registering modules to extract face characteristic.
Step S307, calculates the Euclidean distance of certain face characteristic in the current face that extracts and storehouse, and this face class of this value and registry is obtained to classification thresholds compares judgement.
Described face database system update module comprises system update judgement, and face characteristic extracts and data mining:
Step S308, system update is judged.When system and device is in face recognition module operational process, continuous 3 all unidentified successes of current identification recognition of face, and recognition result is that identification mark is greater than the same class people of setting threshold, needs to input log-in password and confirms identification.Password Input is correct, identifies successfully, and can enter system update.
Step S303, face characteristic extracts.Adopt current identification recognition of face continuous 3 times all unidentified successfully the most similar current human face region substitute in 3 identification human face region in the most dissimilar storehouse, recalculate core principle component analysis (KPCA) feature of human face region in storehouse, use fuzzy FLDA to obtain projection coefficient to KPCA feature, and extract face characteristic.
Step S304, data mining.Described face characteristic data mining technology comprises following characteristics:
First, set up face characteristic classification thresholds data mining model.
Thresholdi is classification thresholds corresponding to every class people, the face classification number that i is face database, and a and b are respectively gain coefficient and the index coefficient a value 0 ~ 6 of model, b value 0.5 ~ 2.In the Xi class that to be each face characteristic in similar identify with everyone except self in face database, the most similar identification mark forms.
Then, obtain model and enter ginseng.Utilize Euclidean distance as distance measure, calculate in each face characteristic in similar and face database everyone except self and identify, the most similar identification mark in the class of at every turn identifying is formed to one-dimension array Xi.
Finally, obtain Data Mining Classification threshold value.Change the value of a and b, obtain manifold classification threshold value, FAR with in FRR curve map, get FAR and FRR below area and b value corresponding to minimum image, and in this figure, get FRR/FAR and approach 1/5 o'clock a value in corresponding threshold value most, with this (a, b) value in conjunction with the Xi of each face class, obtains the Classification and Identification threshold value of all kinds of faces in substitution data mining model
.
With an instantiation, the present invention is made an explanation below, in order to verify the inventive method and device, in Yale face database and AR face data, test.Wherein Yale database has comprised 15 volunteers, and every volunteer has 11 pictures, and totally 165 pictures, comprise illumination, the conversion of expression and attitude.AR database comprises 100 volunteers, and every volunteer has 26 pictures, and we therefrom get the picture of everyone 14 have no occluders of 50 volunteers, and totally 700, this storehouse also comprises illumination, the variation of expression and attitude.Everyone chooses 6 pictures as training set wherein to carry out laboratory at Yale face database, and remaining 5 as test set.In the time that AR face database is tested, everyone chooses front 7 pictures of same time period collection as training set, and 7 of a rear time period as test set.Table 1 has provided the discrimination of the present invention at different face databases.Table one statistics is not considered reject rate.Fig. 4 is that the inventive method does not comprise the test result of data mining part in AR storehouse.This figure illustrates the attribute difference such as texture of every face due to self, and its decipherment distance gap is larger, and single judgment threshold may bring reject rate or the raising of misclassification rate.Fig. 5 is the FAR(false acceptance rate that Fig. 4 is corresponding) and FRR(false rejection rate) figure.When this figure is presented at FAR and is 4.857%, FRR is 3.429%.Fig. 6 is the FAR(false acceptance rate of the inventive method in AR storehouse experimental result) and FRR(false rejection rate) figure.As can be seen from Figure 6, in the time that FAR is 4.857%, FRR is 0.5714%, has obviously reduced reject rate, has improved the recognition efficiency of system.
The discrimination of the present invention of the different face databases of table 1.
| Face database | Yale | AR |
| Discrimination | 98.3% | 94.6% |
The present invention not only can increase registration face characteristic information amount in effective class, better human face posture and the expression shape change of adapting to, and the automatic renewal of system can reduce the reject rate that environment and face temporal evolution bring, also excavate classification thresholds separately for the feature difference of all kinds of faces between class, reduce and used same classification thresholds and the rate of causing delay and reject rate that between ignore class, face difference produces, improved the recognition efficiency of system.
Claims (6)
1. the face identification device based on face characteristic data mining, is characterized in that: described device comprises face Registering modules, face recognition module and system update module, wherein
Described face Registering modules utilizes image local dual mode characteristic similarity matching value, judge the identical image in class and give up according to matching value, calculate core principle component analysis feature for non-identical photo, and use fuzzy FDA to extract face characteristic to core principle component analysis feature, recycling data mining technology is asked for such people's classification thresholds, and records this registrant's log-in password;
Described face recognition module, by computed segmentation image core principle component analysis feature, and use fuzzy FDA to extract face characteristic to core principle component analysis feature, the Euclidean distance of face characteristic in the current face that calculating is extracted again and storehouse, and each classification thresholds of this value and registry is compared to judgement;
Described system update module is continuous 3 all unidentified successes of current identification recognition of face, and identification mark is in claimed range and be same class people, confirm identification by log-in password, Password Input is correct, identify successfully the go forward side by side feature extraction of pedestrian's face and data mining and obtain all kinds of people's of registry classification thresholds, and upgrade face database.
2. the face identification method based on face characteristic data mining that adopts the face identification device based on face characteristic data mining as claimed in claim 1, is characterized in that: described method is
Described face Registering modules comprises that face location, identical face are removed, face characteristic extracts and data mining;
Described face recognition module comprises face location, and face characteristic extracts and recognition of face;
Described face database system update module comprises system update judgement, and face characteristic extracts and data mining.
3. method according to claim 2, it is characterized in that: the human face region that described face is orientated as detecting carries out nose location, according to nose locating information, face is partitioned into 4 human face regions such as face eyebrow image, face eye image, face nose image and face face image.
4. method according to claim 2, it is characterized in that: described identical human face photo is removed the LBP feature for extracting respectively the human face region that face positioning step cuts apart, and the hamming distance of face in current face local binary patterns feature and current face database will be calculated, if hamming distance is less than setting threshold, is judged to be identical photo and removes.
5. method according to claim 2, is characterized in that: described face characteristic is extracted as for non-identical photo and calculates core principle component analysis feature, use fuzzy FLDA to obtain projection coefficient to core principle component analysis feature, and extract face characteristic,
Method according to claim 2, is characterized in that: described recognition of face is to calculate the Euclidean distance of certain face characteristic in the current face that extracts and storehouse, and this face class of this value and registry is obtained to classification thresholds compares judgement.
6. method according to claim 2, is characterized in that: described data mining comprises the steps:
First, set up face characteristic classification thresholds data mining model,
Thresholdi is classification thresholds corresponding to every class people, i is the face classification number of face database, a and b are respectively gain coefficient and the index coefficient a value 0 ~ 6 of model, b value 0.5 ~ 2, in the Xi class that to be each face characteristic in similar identify with everyone except self in face database, the most similar identification mark forms;
Then, obtain model and enter ginseng, utilize Euclidean distance as distance measure, calculate in each face characteristic in similar and face database everyone except self and identify, the most similar identification mark in the class of at every turn identifying is formed to one-dimension array Xi;
Finally, obtain Data Mining Classification threshold value, change the value of a and b, obtain manifold classification threshold value, FAR with in FRR curve map, get FAR and FRR below area and b value corresponding to minimum image, and in this figure, get FRR/FAR and approach 1/5 o'clock a value in corresponding threshold value most, with this (a, b) value in conjunction with the Xi of each face class, obtains the Classification and Identification threshold value of all kinds of faces in substitution data mining model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410002094.6A CN103778409A (en) | 2014-01-02 | 2014-01-02 | Human face identification method based on human face characteristic data mining and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410002094.6A CN103778409A (en) | 2014-01-02 | 2014-01-02 | Human face identification method based on human face characteristic data mining and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103778409A true CN103778409A (en) | 2014-05-07 |
Family
ID=50570623
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410002094.6A Pending CN103778409A (en) | 2014-01-02 | 2014-01-02 | Human face identification method based on human face characteristic data mining and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103778409A (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105117629A (en) * | 2015-08-21 | 2015-12-02 | 京东方科技集团股份有限公司 | Electronic device control method and apparatus, and electronic device |
| CN106295482A (en) * | 2015-06-11 | 2017-01-04 | 中国移动（深圳）有限公司 | The update method of a kind of face database and device |
| CN106485221A (en) * | 2016-09-30 | 2017-03-08 | 江苏四五安全科技有限公司 | A kind of method that benchmark photograph is replaced automatically according to similar concentration degree |
| CN106934328A (en) * | 2015-12-30 | 2017-07-07 | 厦门中控生物识别信息技术有限公司 | The appraisal procedure and device of a kind of Face datection algorithm |
| CN107135208A (en) * | 2017-04-21 | 2017-09-05 | 广州有意思网络科技有限公司 | A kind of register method for moving social platform |
| CN107357799A (en) * | 2017-05-17 | 2017-11-17 | 广州视源电子科技股份有限公司 | Assess the method and device of the face identification system registration upper limit |
| CN107423690A (en) * | 2017-06-26 | 2017-12-01 | 广东工业大学 | A kind of face identification method and device |
| CN108021882A (en) * | 2017-12-01 | 2018-05-11 | 宁波高新区锦众信息科技有限公司 | A kind of face identification system of the robot based on the period |
| CN108197250A (en) * | 2017-12-29 | 2018-06-22 | 深圳云天励飞技术有限公司 | Picture retrieval method, electronic equipment and storage medium |
| CN108206929A (en) * | 2016-12-16 | 2018-06-26 | 北京华泰科捷信息技术股份有限公司 | A kind of contactless personnel information acquisition device and its acquisition method |
| CN109190561A (en) * | 2018-09-04 | 2019-01-11 | 四川长虹电器股份有限公司 | Face identification method and system in a kind of video playing |
| CN109934114A (en) * | 2019-02-15 | 2019-06-25 | 重庆工商大学 | A kind of finger vena template generation and more new algorithm and system |
| CN110363150A (en) * | 2019-07-16 | 2019-10-22 | 深圳市商汤科技有限公司 | Data-updating method and device, electronic equipment and storage medium |
| CN110503030A (en) * | 2019-08-21 | 2019-11-26 | 杭州宇泛智能科技有限公司 | A method of the difficult identification face percent of pass of raising for recognition of face |
| CN111210544A (en) * | 2018-11-05 | 2020-05-29 | 赵青贺 | Door control method and device based on cloud computing |
| CN111756921A (en) * | 2020-06-01 | 2020-10-09 | Oppo（重庆）智能科技有限公司 | Face recognition method and device, terminal and readable storage medium |
| CN111859000A (en) * | 2020-06-24 | 2020-10-30 | 天津大学 | Method for constructing and updating human face feature database under deep learning model |
| CN113095110A (en) * | 2019-12-23 | 2021-07-09 | 浙江宇视科技有限公司 | Method, device, medium and electronic equipment for dynamically warehousing face data |
| CN113963392A (en) * | 2020-07-03 | 2022-01-21 | 北京君正集成电路股份有限公司 | Face recognition method based on dynamic adjustment threshold |
-
2014
- 2014-01-02 CN CN201410002094.6A patent/CN103778409A/en active Pending
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106295482A (en) * | 2015-06-11 | 2017-01-04 | 中国移动（深圳）有限公司 | The update method of a kind of face database and device |
| CN106295482B (en) * | 2015-06-11 | 2019-10-29 | 中移信息技术有限公司 | A kind of update method and device of face database |
| CN105117629A (en) * | 2015-08-21 | 2015-12-02 | 京东方科技集团股份有限公司 | Electronic device control method and apparatus, and electronic device |
| CN106934328A (en) * | 2015-12-30 | 2017-07-07 | 厦门中控生物识别信息技术有限公司 | The appraisal procedure and device of a kind of Face datection algorithm |
| CN106485221B (en) * | 2016-09-30 | 2019-09-17 | 江苏四一五安全科技有限公司 | A method of benchmark photograph is replaced according to similar concentration degree automatically |
| CN106485221A (en) * | 2016-09-30 | 2017-03-08 | 江苏四五安全科技有限公司 | A kind of method that benchmark photograph is replaced automatically according to similar concentration degree |
| CN108206929A (en) * | 2016-12-16 | 2018-06-26 | 北京华泰科捷信息技术股份有限公司 | A kind of contactless personnel information acquisition device and its acquisition method |
| CN107135208A (en) * | 2017-04-21 | 2017-09-05 | 广州有意思网络科技有限公司 | A kind of register method for moving social platform |
| CN107135208B (en) * | 2017-04-21 | 2020-07-14 | 广州有意思网络科技有限公司 | Registration method of sports social platform |
| CN107357799A (en) * | 2017-05-17 | 2017-11-17 | 广州视源电子科技股份有限公司 | Assess the method and device of the face identification system registration upper limit |
| CN107423690A (en) * | 2017-06-26 | 2017-12-01 | 广东工业大学 | A kind of face identification method and device |
| CN108021882A (en) * | 2017-12-01 | 2018-05-11 | 宁波高新区锦众信息科技有限公司 | A kind of face identification system of the robot based on the period |
| CN108197250A (en) * | 2017-12-29 | 2018-06-22 | 深圳云天励飞技术有限公司 | Picture retrieval method, electronic equipment and storage medium |
| CN109190561A (en) * | 2018-09-04 | 2019-01-11 | 四川长虹电器股份有限公司 | Face identification method and system in a kind of video playing |
| CN111210544A (en) * | 2018-11-05 | 2020-05-29 | 赵青贺 | Door control method and device based on cloud computing |
| CN109934114A (en) * | 2019-02-15 | 2019-06-25 | 重庆工商大学 | A kind of finger vena template generation and more new algorithm and system |
| CN109934114B (en) * | 2019-02-15 | 2023-05-12 | 重庆工商大学 | Finger vein template generation and updating algorithm and system |
| CN110363150A (en) * | 2019-07-16 | 2019-10-22 | 深圳市商汤科技有限公司 | Data-updating method and device, electronic equipment and storage medium |
| CN110503030A (en) * | 2019-08-21 | 2019-11-26 | 杭州宇泛智能科技有限公司 | A method of the difficult identification face percent of pass of raising for recognition of face |
| CN113095110A (en) * | 2019-12-23 | 2021-07-09 | 浙江宇视科技有限公司 | Method, device, medium and electronic equipment for dynamically warehousing face data |
| CN113095110B (en) * | 2019-12-23 | 2024-03-08 | 浙江宇视科技有限公司 | Method, device, medium and electronic equipment for dynamically warehousing face data |
| CN111756921A (en) * | 2020-06-01 | 2020-10-09 | Oppo（重庆）智能科技有限公司 | Face recognition method and device, terminal and readable storage medium |
| CN111859000A (en) * | 2020-06-24 | 2020-10-30 | 天津大学 | Method for constructing and updating human face feature database under deep learning model |
| CN113963392A (en) * | 2020-07-03 | 2022-01-21 | 北京君正集成电路股份有限公司 | Face recognition method based on dynamic adjustment threshold |
| CN113963392B (en) * | 2020-07-03 | 2024-05-03 | 北京君正集成电路股份有限公司 | Face recognition method based on dynamic adjustment threshold |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103778409A (en) | Human face identification method based on human face characteristic data mining and device | |
| Nguyen et al. | Robust minutiae extractor: Integrating deep networks and fingerprint domain knowledge | |
| CN102945366B (en) | A kind of method and device of recognition of face | |
| CN106203294B (en) | The testimony of a witness based on face character analysis unifies auth method | |
| CN105095856B (en) | Face identification method is blocked based on mask | |
| CN103914686B (en) | A kind of face alignment authentication method and system shone based on certificate photo with collection | |
| CN109145742B (en) | Pedestrian identification method and system | |
| CN105335722A (en) | Detection system and detection method based on depth image information | |
| CN105335726B (en) | Recognition of face confidence level acquisition methods and system | |
| CN102254165B (en) | Hand back vein identification method based on fusion of structural coding features and texture coding features | |
| CN108009472A (en) | A kind of finger back arthrosis line recognition methods based on convolutional neural networks and Bayes classifier | |
| Li et al. | Fast and effective text detection | |
| CN105138968A (en) | Face authentication method and device | |
| CN102629320A (en) | Ordinal measurement statistical description face recognition method based on feature level | |
| CN104036254A (en) | Face recognition method | |
| CN102542243A (en) | LBP (Local Binary Pattern) image and block encoding-based iris feature extracting method | |
| CN104008372A (en) | Distributed face recognition method in wireless multi-media sensor network | |
| CN103488966A (en) | Intelligent mobile phone capable of identifying real-name ticket information | |
| CN102262723A (en) | Face recognition method and device | |
| Das et al. | Human face detection in color images using HSV color histogram and WLD | |
| CN104299000A (en) | Handwriting recognition method based on local fragment distribution characteristics | |
| CN107025435A (en) | A kind of face recognition processing method and system | |
| CN106295523A (en) | A kind of public arena based on SVM Pedestrian flow detection method | |
| CN106650363A (en) | Identity recognition method and system | |
| CN113158958B (en) | Traffic method and related device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20140507 |