CN106682650A - Mobile terminal face recognition method and system based on technology of embedded deep learning - Google Patents
Mobile terminal face recognition method and system based on technology of embedded deep learning Download PDFInfo
- Publication number
- CN106682650A CN106682650A CN201710063323.9A CN201710063323A CN106682650A CN 106682650 A CN106682650 A CN 106682650A CN 201710063323 A CN201710063323 A CN 201710063323A CN 106682650 A CN106682650 A CN 106682650A
- Authority
- CN
- China
- Prior art keywords
- face
- mobile terminal
- portrait
- feature
- target person
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/60—Protecting data
- G06F21/602—Providing cryptographic facilities or services
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/757—Matching configurations of points or features
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Software Systems (AREA)
- Computer Security & Cryptography (AREA)
- General Engineering & Computer Science (AREA)
- Computer Hardware Design (AREA)
- Human Computer Interaction (AREA)
- Bioethics (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention belongs to the field of image recognition, and specifically provides a mobile terminal face recognition method and system based on the technology of embedded deep learning. The invention aims at solving problems in the prior art that a human face recognition program is complex, an occupied memory is large, the recognition speed is small, the recognition precision is low and a hand-held mobile terminal for human face recognition is limited by a network signal. In order to solve the above problems, the system comprises a human face detection module for obtaining a face region of a target person; a feature extraction module which is used for obtaining the face features of the target person according to the obtained face region; and a feature comparison module which is used for calculating the similarity between the obtained face features of the target person and preset face features, wherein the human face detection module, the feature extraction module and the feature comparison module are disposed on a mobile terminal. Therefore, the system enables the mobile terminal to be able to judge the identity of the target person quickly and accurately under the condition that there is no network signal.
Description
Technical field
The invention belongs to field of image recognition, specifically provides a kind of mobile terminal people based on embedded deep learning technology
Face recognition method and system.
Background technology
In public safety field, during such as public security cadres and police patrol or be on duty, staff is live, large-scale in accident
Need quickly to carry out identity validation to a suspect in the places such as site of activity, such as judge a suspect be whether fugitive personnel,
Person appealing for help seeks gap and stirs up trouble personnel etc..Therefore a kind of identification system that may move, be convenient for carrying is needed badly.
The method for generally adopting at present is that a suspect is taken pictures by hand-held mobile terminal, then by wireless
Network enters photo upload to server end, the photo for being shot mobile terminal by server end with the photo in portrait data base
Matching result is fed back to hand-held mobile terminal by row matching, last server end by wireless network again.But this mode is past
Toward network signal is limited to, when network signal is poor or does not have network signal, the speed that photo upload and matching result are downloaded
Can be greatly diminished.And existing mobile terminal face identification system uses conventional methods extraction face characteristic, actually should
It is not high with middle accuracy rate.Unprecedented raising is had based on the recognition accuracy of the face identification method of deep learning, however, the party
Method has that memory consumption is big, recognition speed is slow.
Correspondingly, this area needs a kind of new face identification method to solve the above problems.
The content of the invention
In order to solve the problems referred to above of the prior art, recognition of face program complexity in solution prior art has been it, has accounted for
According to internal memory it is big, recognition speed is slow, recognition accuracy is low and the hand-held mobile terminal that face is identified is limited to network letter
Number problem, the invention provides a kind of mobile terminal face identification system based on embedded deep learning technology, including people
Face detection module, characteristic extracting module, feature comparing module, portrait DBM, the characteristic extracting module, based on depth
Face feature in the face image of the target person that the human face recognition model of study is extracted；The portrait DBM, bag
Include the quick indexing tree of portrait characteristic and portrait characteristic；The feature comparing module, is configured to feature extraction
The face feature that module is extracted, based on the quick indexing tree in the portrait DBM Rapid matching is carried out, and is exported
Matching result.
In the optimal technical scheme of said system, the characteristic extracting module is configured to the face of deep learning
Identification model, the face image data of the target person exported to face detection module carries out carrying for the face feature of target person
Take.
In the optimal technical scheme of said system, the quick indexing tree is built based on KD-Tree.
In the optimal technical scheme of said system, the face detection module, the characteristic extracting module and the spy
Levy comparing module to be arranged on mobile terminal.
In the optimal technical scheme of said system, the portrait DBM includes being arranged at recognition of face server
On it is the first as DBM and the second portrait DBM for being arranged on the mobile terminal, and described second
Portrait DBM can download portrait characteristic and portrait characteristic from the first picture DBM
Quick indexing tree.
In the optimal technical scheme of said system, the system also includes the data being arranged on recognition of face server
Encrypting module and setting data decryption module on mobile terminals, the data encryption module is used for the first picture number
It is encrypted according to the quick indexing tree of the portrait characteristic in library module and portrait characteristic, the data deciphering mould
Block is used to enter the quick indexing tree of the portrait characteristic in the second portrait DBM and portrait characteristic
Row decryption.
In the optimal technical scheme of said system, the face detection module, the characteristic extracting module, the feature
Comparing module and the portrait DBM are also disposed on recognition of face server, and the face detection module can
The image information of target person on mobile terminal is obtained by wired or wireless mode.
In the optimal technical scheme of said system, the face detection module is used for the target person of acquisition for mobile terminal
The image of thing carries out Face datection and calibration, and then obtains the face area of target person so that the characteristic extracting module energy
Enough face features that target person is obtained according to the face area of the target person.
On the other hand, present invention also offers a kind of mobile terminal recognition of face based on embedded deep learning technology
Method, the method comprising the steps of：Set up the quick indexing tree of portrait property data base and portrait property data base；
Human face recognition model based on deep learning obtains the face feature of target person；By the face feature of the target person and institute
Stating quick indexing tree carries out Rapid matching, and output matching result.
In the optimal technical scheme of said method, human face recognition model of the portrait characteristic based on deep learning
Generate.
In the optimal technical scheme of said method, the quick indexing tree of the portrait characteristic is based on KD-Tree structures
Build.
In the optimal technical scheme of said method, the human face recognition model based on deep learning obtains target person
Face feature, including：Obtain the image of target person；The facial regions of target person are obtained according to the image of the target person
Domain；Human face recognition model based on deep learning is special according to the face that the face area of the target person obtains target person
Levy.
It is described that target person is obtained according to the face area of the target person in the optimal technical scheme of said method
Face feature, including based on deep learning human face recognition model, to face detection module output target person face
View data carries out the extraction of the face feature of target person.
In the optimal technical scheme of said method, methods described also includes step：By set up portrait characteristic and
Tree is encrypted quick indexing, and is decrypted before the face feature of the target person is matched with the quick indexing tree.
It will be appreciated to those of skill in the art that in the preferred technical solution of the present invention, being arranged on recognition of face clothes
The first multiple portraits being stored with by the human face recognition model establishment based on deep learning as DBM on business device
Characteristic, and the quick indexing tree of the plurality of portrait characteristic created by KD-Tree technologies.Then data are passed through
Encrypting module is encrypted portrait characteristic and quick indexing tree, and then by being arranged on the second portrait data of mobile terminal
Module is downloaded the portrait characteristic after encryption and quick indexing tree, and by data decryption module to the people after download
As characteristic and quick indexing tree are decrypted.
Further, by arrange face detection module on mobile terminals to the face image of target person (for example according to
Piece) Face datection and calibration are carried out, and then obtain the face area of the target person.By arranging on mobile terminals and can
The characteristic extracting module communicated with face detection module obtains the face feature of target person according to the face area of target person.
It is by the feature comparing module that arrange on mobile terminals and can communicate with characteristic extracting module that the face of target person is special
Levy and matched with the quick indexing tree in the second portrait data base.Specifically, by the face feature of target person in quick rope
Draw and scanned on tree, when the phase of certain the one or more portrait feature in face feature and the quick indexing catalogue of target person
When reaching threshold value like degree, one or more portrait feature is arranged from high to low according to similarity.And then output matching result.
Therefore, by the method for the present invention, mobile terminal also can be quickly and accurately in the case of without network signal
Judge the identity of target person, and the information in the case where mobile terminal is lost in portrait data base also will not be let out
Leakage.
Description of the drawings
Fig. 1 is the system structure of the mobile terminal face identification system based on embedded deep learning technology of the present invention
Figure；
The step of Fig. 2 is the mobile terminal face identification method based on embedded deep learning technology of present invention flow process
Figure；
Fig. 3 is that the mobile terminal face identification method based on embedded deep learning technology of the present invention is built based on depth
The step of human face recognition model of study flow chart.
Reference numerals list：
1st, recognition of face server；11st, the first picture DBM；12nd, data encryption module；2nd, mobile terminal；21、
Face detection module；22nd, characteristic extracting module；23rd, feature comparing module；24th, data decryption module；25th, the second portrait data
Library module.
Specific embodiment
With reference to the accompanying drawings describing the preferred embodiment of the present invention.It will be apparent to a skilled person that this
A little embodiments are used only for the know-why of the explanation present invention, are not intended to limit protection scope of the present invention.For example, although
Two functional modules of the first picture DBM and data encryption module are only listed to recognition of face server in accompanying drawing 1,
But recognition of face server can also include other functions module, such as communication module, and those skilled in the art can be according to need
It is made adjustment, to adapt to specific application scenario, the technical scheme after adjustment will fall into the protection model of the present invention
Enclose.
As shown in figure 1, the present invention's is mainly wrapped based on the mobile terminal face identification system of embedded deep learning technology
Include recognition of face server 1 and mobile terminal 2.Recognition of face server 1 is mainly used in setting up portrait data archival, such as fugitive
The portrait data archival of personnel.Mobile terminal 2 can be downloaded portrait data archival from recognition of face server 1, and
The photo of target person can be carried out Similarity Measure with the portrait data base for downloading, and then target is judged according to result of calculation
Whether personage is fugitive personnel.It will be appreciated to those of skill in the art that mobile terminal 2 can be any bat being convenient for carrying
According to equipment, such as mobile phone, photographing unit, video camera etc., or mobile terminal 2 can also be it is any be convenient for carrying and can with take pictures
The equipment that equipment is communicated.
With continued reference to Fig. 1, recognition of face server 1 mainly includes the first picture DBM 11 and data encryption module
12.Wherein, it is the first as preserving portrait data base in DBM 11, portrait data base can be fugitive personnel storehouse,
Person appealing for help storehouse, key monitoring personnel storehouse etc..In a preferred embodiment of the invention portrait data base is fugitive personnel storehouse, or
Person those skilled in the art can also set up as needed the portrait data of other staff in the first picture DBM 11
Archives, for example, child, old man of loss etc..Preferably, portrait data base is the first picture DBM 11 by based on deep
The portrait data with multiple portrait characteristics that the human face recognition model (hereafter will be described in more detail) of degree study is set up
Storehouse.Further, portrait data base also include using KD-Tree (abbreviation of k-dimensional trees) technique construction everyone
As the quick indexing tree of characteristic.Data encryption module 12 can be by the portrait data base in the first picture DBM 11
It is encrypted, prevents portrait characteristic from being stolen by unauthorized person.
Fig. 1 is further regarded to, mobile terminal 2 mainly includes：Face detection module 21, characteristic extracting module 22, aspect ratio
To module 23, the portrait DBM 25 of data decryption module 24 and second.Flow of information between each functional module such as Fig. 1 institutes
Show.Wherein, the second portrait DBM 25 can be by wired or wireless mode from the first picture DBM 11
Portrait data base after data encryption module 12 is encrypted is downloaded.
Data decryption module 24 is used to be decrypted portrait data base encrypted in the second portrait DBM 25,
Feature comparing module 23 is called to portrait data base.Further, data decryption module 24 includes logging in submodule
Block, the login submodule is used to carry out authentication to the user of mobile terminal 2, and the only authentication in user passes through
Afterwards, data decryption module 24 just can be decrypted to portrait data base, to lose descendant as in data base in mobile terminal 2
Information will not be stolen by unauthorized person.It will be appreciated to those of skill in the art that log in submodule entering to user identity
The mode of row checking can be the those skilled in the art such as password login, recognition of face, fingerprint recognition it is conceivable that and can be real
The mode applied.
Face detection module 21 is used for the photo of the target person that mobile terminal is photographed or received carries out face inspection
Survey and calibrate, and then obtain the face area of target person.The face that characteristic extracting module 22 is obtained by face detection module 21
The face feature of portion's extracted region target person.Because of being extracted as skilled artisans appreciate that and conventional for face feature
Technological means, so do not elaborate herein.
Feature comparing module 23 can obtain the face feature of the target person of the extraction of characteristic extracting module 22, and by the mesh
The face feature of mark personage carries out Similarity Measure with the portrait data base that data decryption module 24 is decrypted.Specifically, it is special
Levy quick indexing tree of the comparing module 23 by the face feature of target person in portrait data base and carry out Similarity Measure and (search
Rope).When the similarity of certain the one or more portrait feature in the face feature and quick indexing catalogue of target person reaches threshold
During value, one or more portrait feature is arranged from high to low according to similarity.And then output matching result, entered by user
One step determines whether target person is fugitive personnel.It will be appreciated to those of skill in the art that the threshold value of the setting can basis
Test of many times is obtained.
It will be appreciated to those of skill in the art that a recognition of face server in a preferred embodiment of the invention
1 can serve multiple stage mobile terminal 1.
In addition, those skilled in the art can be with omitted data encrypting module 12 and data decryption module 24, by people
Face detection module 21, characteristic extracting module 22, the portrait DBM 25 of feature comparing module 23 and second are provided entirely in people
On face identification server 1.The photo of target person is sent to into face by wired or wireless mode by mobile terminal 2 to know
Other server 1, is then processed and is matched by recognition of face server 1 to the photo of target person, and will be processed, be matched knot
Fruit is sent to mobile terminal 2.Meanwhile, the comparison result on mobile terminal 2 can also send to recognition of face server and carry out
Storage and management.
With reference to above-mentioned mobile terminal face identification system to the present invention based on embedded deep learning technology
Mobile terminal face identification method is described in detail.
As shown in Fig. 2 the present invention's is mainly wrapped based on the mobile terminal face identification method of embedded deep learning technology
Include：Step S100, sets up based on the human face recognition model of deep learning；Step S200, sets up portrait data base；Step S300,
Obtain the facial zone of target person；Step S400, according to the facial zone of target person the face feature of target person is obtained；
Step S500, by the face feature in the face feature of target person and portrait data base Similarity Measure is carried out.
As shown in figure 3, step S100 also includes：Step S101, prepares training sample；Step S102, builds and is based on depth
The human face recognition model of study；Step S103, is trained by the human face recognition model based on deep learning；Step S104,
Human face recognition model of the compression based on deep learning.
Specifically, in step S101, face image datas more than 100,000 people is first collected, the face image data includes
Everyone 1~3 certificate photo and 5~10 scenes is taken pictures.Further, there is the age of 1~10 year in everyone photo
Span, and the human face photo at scene is in many factors rings such as different illumination, different angles, different expressions, different resolutions
Shoot under border.Skilled addressee readily understands that, each individual of sample (people) is gathered multiple have the age across
The photo of degree, can be carried out according to related algorithm (such as the human face recognition model based on deep learning of the present invention) to individual of sample
Facial information calculates that the appearance after such as calculating individual of sample for many years (such as 5 years, 10 years, 15 years) according to existing photographic intelligence is kept away
Cause the phenomenon that can not be identified to individual of sample having exempted from the recent photo because lacking individual specimen；Further to each
Individual individual of sample (people) gathers multiple photos for shooting under various circumstances, it is possible to increase individual of sample the match is successful rate, keeps away
The shooting photo that environmental factorss when having exempted from because taking pictures are caused is not matched that and None- identified sample with the photo in Sample Storehouse
The phenomenon of body.
Specifically, in step s 102, building one includes ten Ge Juan basic units, five down-sampling layers, ten non-linear biographies
Layer, a full articulamentum and a neural network structure for returning layer and a contrast loss layer are broadcast, the network structure is base
In the human face recognition model of deep learning, specifically, connection one is non-linear behind each volume basic unit for the structure of the neutral net
Propagation layer, a pond layer is connected per two groups after " volume basic unit → nonlinear propagation layer ", and above-mentioned repetition connects afterwards for 5 times and connects entirely
Layer is connect, again respectively connection returns layer and contrast loss layer after full articulamentum.
Specifically, in step s 103, by each training sample through Face datection, scaling normalization and data augmentation
Afterwards, together with corresponding class label (those skilled in the art can be set as needed) input build based on depth
In the human face recognition model of habit, more than 5,000,000 iteration of training, network parameter is obtained.Skilled artisans appreciate that
Be, due to Face datection, scaling normalization and data augmentation be people in the art it will be appreciated that and conventional technological means, institute
No longer to elaborate herein.
Specifically, in step S104, quantify to share the modes such as weights to based on depth by model beta pruning, weights
The human face recognition model of habit is compressed so that after compression is the ten of original size based on the human face recognition model of deep learning
/ mono-, and then can effectively reduce the computing resource based on the human face recognition model of deep learning so that based on deep learning
Human face recognition model can be embedded on mobile terminal.It will be appreciated to those of skill in the art that will by model beta pruning
The absolute value of parameter removes less than the side of certain threshold value in deep learning network；Weights quantify shared weights：By to each
The weights of layer are clustered (such as clustered using k-means), the shared identical weights of same cluster (cluster)；Adopt in addition
Replaced storing concrete numerical value with the method for the index for storing weights, and then the memory space of model can be greatly reduced.
Specifically, in step s 200, first, the human face recognition model founder based on deep learning, should as data base
Portrait data base includes multiple portrait characteristics, and those skilled in the art can be selected the portrait characteristic as needed
Fixed, for example, the portrait characteristic is the facial feature information of fugitive personnel.Then, set up the plurality of by KD-Tree technologies
The quick indexing tree of portrait characteristic, and quick indexing tree is stored to portrait data base.Finally, by portrait data stock
Store up into the first picture DBM 11.
Further, portrait data base is encrypted by data encryption module 12.Then, by the second portrait data base
The portrait data base that module 25 encrypts data encryption module 12 is from the first as downloading to mobile terminal 2 in DBM 11
On.Finally, it is decrypted by 24 pairs of portrait database trees downloaded on mobile terminal 2 of data decryption module.
Specifically, in step S300, the photo of mobile terminal 2 is obtained target person by face detection module 21
Detected and calibrated the face area for further obtaining target person.
Specifically, in step S400, feature is carried out to the face area of target person by characteristic extracting module 22 and is carried
Take and then obtain the face feature of target person.
Specifically, in step S500, feature comparing module 23 is by the human face recognition model based on deep learning by mesh
Scan on the face feature of mark personage quick indexing tree after decryption.When the face feature and quick indexing of target person
When the similarity of one or more portrait feature of certain in catalogue reaches threshold value, by one or more portrait feature according to similar
Degree is arranged from high to low.And then output matching result, further determine that whether target person is fugitive personnel by user.Most simultaneously
Matching result is stored to mobile terminal 2, or when mobile terminal 2 and recognition of face server 1 can be communicated, will
Result of calculation is uploaded to recognition of face server 1.
It will be appreciated to those of skill in the art that the mobile terminal 2 of the present invention also can be right in the case where not networking
Target person carries out recognition of face so that the mobile terminal 2 of the present invention is no longer limited by network signal, can be by more rings
Used in border.The present invention is processed the facial characteristics of target person by the human face recognition model based on deep learning, and
By to portrait Database quick indexing tree, not only increasing the accuracy of recognition of face, recognition of face is also improved
Speed.The present invention substantially reduces internal memory occupancy volume by will be compressed based on the human face recognition model of deep learning.This
It is bright also to be made one as the data in data base is in transmission and the quilt of mobile terminal 2 by data encryption module 12 and data decryption module 24
In the case of loss, will not be stolen, it is ensured that the safety of information in portrait data base.
So far, technical scheme is described already in connection with preferred implementation shown in the drawings, but, this area
Technical staff is it is easily understood that protection scope of the present invention is expressly not limited to these specific embodiments.Without departing from this
On the premise of the principle of invention, those skilled in the art can make the change or replacement of equivalent to correlation technique feature, these
Technical scheme after changing or replacing it is fallen within protection scope of the present invention.
Claims (13)
1. a kind of mobile terminal face identification system based on embedded deep learning technology, including face detection module, feature
Extraction module, feature comparing module, portrait DBM, it is characterised in that
The characteristic extracting module, the face in the face image of the target person that the human face recognition model based on deep learning is extracted
Portion's feature；
The portrait DBM, including the quick indexing tree of portrait characteristic and portrait characteristic；
The feature comparing module, is configured to the face feature for being extracted characteristic extracting module, based on the portrait data base
Quick indexing tree in module carries out Rapid matching, and output matching result.
2. the mobile terminal face identification system based on embedded deep learning technology according to claim 1, its feature
It is that the characteristic extracting module is configured to the human face recognition model of deep learning, the mesh to face detection module output
The face image data of mark personage carries out the extraction of the face feature of target person.
3. the mobile terminal face identification system based on embedded deep learning technology according to claim 2, its feature
It is that the quick indexing tree is built based on KD-Tree.
4. the mobile terminal face identification system based on embedded deep learning technology according to claim 3, its feature
It is that the face detection module, the characteristic extracting module and the feature comparing module are arranged on mobile terminal.
5. the mobile terminal face identification system based on embedded deep learning technology according to claim 4, its feature
It is that the portrait DBM includes being arranged at the first as DBM and being arranged on recognition of face server
The second portrait DBM on the mobile terminal, and
The second portrait DBM can download from the first picture DBM portrait characteristic and
The quick indexing tree of portrait characteristic.
6. the mobile terminal face identification system based on embedded deep learning technology according to claim 5, its feature
It is that the system also includes number of the data encryption module being arranged on recognition of face server with setting on mobile terminals
According to deciphering module,
The data encryption module is used for the portrait characteristic and portrait feature in the first picture DBM
The quick indexing tree of data is encrypted,
The data decryption module is used for the portrait characteristic and portrait feature in the second portrait DBM
The quick indexing tree of data is decrypted.
7. the mobile terminal face identification system based on embedded deep learning technology according to claim 3, its feature
It is, the face detection module, the characteristic extracting module, the feature comparing module and the portrait DBM
It is arranged on recognition of face server, and the face detection module can obtain mobile whole by wired or wireless mode
The image information of target person on end.
8. the mobile terminal recognition of face based on embedded deep learning technology according to any one of claim 1 to 7
System, it is characterised in that the face detection module is used to for the image of the target person of acquisition for mobile terminal to carry out face inspection
Survey and calibrate, and then obtain the face area of target person so that the characteristic extracting module can be according to the target person
Face area obtain target person face feature.
9. a kind of mobile terminal face identification method based on embedded deep learning technology, it is characterised in that methods described bag
Include following steps：
Set up the quick indexing tree of portrait characteristic and portrait characteristic；
Human face recognition model based on deep learning obtains the face feature of target person；
The face feature of the target person is carried out into Rapid matching, and output matching result with the quick indexing tree.
10. the mobile terminal face identification method based on embedded deep learning technology according to claim 9, its feature
It is that human face recognition model of the portrait characteristic based on deep learning is generated.
The 11. mobile terminal face identification methods based on embedded deep learning technology according to claim 10, it is special
Levy and be, the quick indexing tree of the portrait characteristic is built based on KD-Tree.
The 12. mobile terminal face identification methods based on embedded deep learning technology according to claim 11, it is special
Levy and be, the human face recognition model based on deep learning obtains the face feature of target person, including：
Obtain the image of target person；
The face area of target person is obtained according to the image of the target person；
Human face recognition model based on deep learning is special according to the face that the face area of the target person obtains target person
Levy.
The 13. mobile terminal face identification methods based on embedded deep learning technology according to claim 12, it is special
Levy and be, methods described also includes step：
The portrait characteristic of foundation and quick indexing tree are encrypted, and the target person face feature with it is described
It is decrypted before the matching of quick indexing tree.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710063323.9A CN106682650A (en) | 2017-01-26 | 2017-01-26 | Mobile terminal face recognition method and system based on technology of embedded deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710063323.9A CN106682650A (en) | 2017-01-26 | 2017-01-26 | Mobile terminal face recognition method and system based on technology of embedded deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106682650A true CN106682650A (en) | 2017-05-17 |
Family
ID=58860261
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710063323.9A Pending CN106682650A (en) | 2017-01-26 | 2017-01-26 | Mobile terminal face recognition method and system based on technology of embedded deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106682650A (en) |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107403155A (en) * | 2017-07-26 | 2017-11-28 | 北京声迅电子股份有限公司 | A kind of rapid classification sorting technique and device |
| CN107634943A (en) * | 2017-09-08 | 2018-01-26 | 中国地质大学（武汉） | A kind of weights brief wireless sense network data compression method, equipment and storage device |
| CN107993663A (en) * | 2017-09-11 | 2018-05-04 | 北京航空航天大学 | A kind of method for recognizing sound-groove based on Android |
| CN108228696A (en) * | 2017-08-31 | 2018-06-29 | 深圳市商汤科技有限公司 | Research on face image retrieval and system, filming apparatus, computer storage media |
| CN108345847A (en) * | 2018-01-30 | 2018-07-31 | 石数字技术成都有限公司 | A kind of facial image label data generation system and method |
| CN109086789A (en) * | 2018-06-08 | 2018-12-25 | 四川斐讯信息技术有限公司 | A kind of image-recognizing method and system |
| CN109284746A (en) * | 2018-11-28 | 2019-01-29 | 江苏中利电子信息科技有限公司 | A kind of fire-fighting scene face identification system |
| CN109344731A (en) * | 2018-09-10 | 2019-02-15 | 电子科技大学 | The face identification method of lightweight neural network based |
| CN109558839A (en) * | 2018-11-29 | 2019-04-02 | 徐州立讯信息科技有限公司 | Adaptive face identification method and the equipment and system for realizing this method |
| CN110196734A (en) * | 2018-02-27 | 2019-09-03 | 上海寒武纪信息科技有限公司 | A kind of computing device and Related product |
| CN110196735A (en) * | 2018-02-27 | 2019-09-03 | 上海寒武纪信息科技有限公司 | A kind of computing device and Related product |
| WO2020024619A1 (en) * | 2018-08-01 | 2020-02-06 | Oppo广东移动通信有限公司 | Data processing method and apparatus, computer-readable storage medium and electronic device |
| CN111553320A (en) * | 2020-05-14 | 2020-08-18 | 支付宝(杭州)信息技术有限公司 | Feature extraction method for protecting personal data privacy, model training method and hardware |
| CN111783674A (en) * | 2020-07-02 | 2020-10-16 | 厦门市美亚柏科信息股份有限公司 | Face recognition method and system based on AR glasses |
| CN112052730A (en) * | 2020-07-30 | 2020-12-08 | 广州市标准化研究院 | 3D dynamic portrait recognition monitoring device and method |
| CN112149497A (en) * | 2020-08-10 | 2020-12-29 | 中标慧安信息技术股份有限公司 | Operating system safe login method based on face recognition |
| CN112989879A (en) * | 2019-12-14 | 2021-06-18 | 杨明 | Data receiving and transmitting system and method applying characteristic analysis |
| US11068707B2 (en) | 2018-10-30 | 2021-07-20 | Fujitsu Limited | Person searching method and apparatus and image processing device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104573679A (en) * | 2015-02-08 | 2015-04-29 | 天津艾思科尔科技有限公司 | Deep learning-based face recognition system in monitoring scene |
| CN105488478A (en) * | 2015-12-02 | 2016-04-13 | 深圳市商汤科技有限公司 | Face recognition system and method |
| CN105574506A (en) * | 2015-12-16 | 2016-05-11 | 深圳市商汤科技有限公司 | Intelligent face tracking system and method based on depth learning and large-scale clustering |
-
2017
- 2017-01-26 CN CN201710063323.9A patent/CN106682650A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104573679A (en) * | 2015-02-08 | 2015-04-29 | 天津艾思科尔科技有限公司 | Deep learning-based face recognition system in monitoring scene |
| CN105488478A (en) * | 2015-12-02 | 2016-04-13 | 深圳市商汤科技有限公司 | Face recognition system and method |
| CN105574506A (en) * | 2015-12-16 | 2016-05-11 | 深圳市商汤科技有限公司 | Intelligent face tracking system and method based on depth learning and large-scale clustering |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107403155A (en) * | 2017-07-26 | 2017-11-28 | 北京声迅电子股份有限公司 | A kind of rapid classification sorting technique and device |
| CN108228696A (en) * | 2017-08-31 | 2018-06-29 | 深圳市商汤科技有限公司 | Research on face image retrieval and system, filming apparatus, computer storage media |
| WO2019042230A1 (en) * | 2017-08-31 | 2019-03-07 | 深圳市商汤科技有限公司 | Method, system, photographing device, and computer storage medium for facial image search |
| CN108228696B (en) * | 2017-08-31 | 2021-03-23 | 深圳市商汤科技有限公司 | Face image retrieval method and system, shooting device and computer storage medium |
| US11182594B2 (en) | 2017-08-31 | 2021-11-23 | Shenzhen Sensetime Technology Co., Ltd. | Face image retrieval methods and systems, photographing apparatuses, and computer storage media |
| CN107634943A (en) * | 2017-09-08 | 2018-01-26 | 中国地质大学（武汉） | A kind of weights brief wireless sense network data compression method, equipment and storage device |
| CN107993663A (en) * | 2017-09-11 | 2018-05-04 | 北京航空航天大学 | A kind of method for recognizing sound-groove based on Android |
| CN108345847A (en) * | 2018-01-30 | 2018-07-31 | 石数字技术成都有限公司 | A kind of facial image label data generation system and method |
| CN110196735A (en) * | 2018-02-27 | 2019-09-03 | 上海寒武纪信息科技有限公司 | A kind of computing device and Related product |
| CN110196734A (en) * | 2018-02-27 | 2019-09-03 | 上海寒武纪信息科技有限公司 | A kind of computing device and Related product |
| CN109086789A (en) * | 2018-06-08 | 2018-12-25 | 四川斐讯信息技术有限公司 | A kind of image-recognizing method and system |
| US11373445B2 (en) | 2018-08-01 | 2022-06-28 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Method and apparatus for processing data, and computer readable storage medium |
| WO2020024619A1 (en) * | 2018-08-01 | 2020-02-06 | Oppo广东移动通信有限公司 | Data processing method and apparatus, computer-readable storage medium and electronic device |
| CN109344731A (en) * | 2018-09-10 | 2019-02-15 | 电子科技大学 | The face identification method of lightweight neural network based |
| US11068707B2 (en) | 2018-10-30 | 2021-07-20 | Fujitsu Limited | Person searching method and apparatus and image processing device |
| CN109284746A (en) * | 2018-11-28 | 2019-01-29 | 江苏中利电子信息科技有限公司 | A kind of fire-fighting scene face identification system |
| CN109558839A (en) * | 2018-11-29 | 2019-04-02 | 徐州立讯信息科技有限公司 | Adaptive face identification method and the equipment and system for realizing this method |
| CN112989879A (en) * | 2019-12-14 | 2021-06-18 | 杨明 | Data receiving and transmitting system and method applying characteristic analysis |
| CN111553320A (en) * | 2020-05-14 | 2020-08-18 | 支付宝(杭州)信息技术有限公司 | Feature extraction method for protecting personal data privacy, model training method and hardware |
| CN111553320B (en) * | 2020-05-14 | 2021-12-21 | 支付宝(杭州)信息技术有限公司 | Feature extraction method for protecting personal data privacy, model training method and hardware |
| CN111783674A (en) * | 2020-07-02 | 2020-10-16 | 厦门市美亚柏科信息股份有限公司 | Face recognition method and system based on AR glasses |
| CN112052730A (en) * | 2020-07-30 | 2020-12-08 | 广州市标准化研究院 | 3D dynamic portrait recognition monitoring device and method |
| CN112052730B (en) * | 2020-07-30 | 2024-03-29 | 广州市标准化研究院 | 3D dynamic portrait identification monitoring equipment and method |
| CN112149497A (en) * | 2020-08-10 | 2020-12-29 | 中标慧安信息技术股份有限公司 | Operating system safe login method based on face recognition |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106682650A (en) | Mobile terminal face recognition method and system based on technology of embedded deep learning | |
| CN105719188B (en) | The anti-method cheated of settlement of insurance claim and server are realized based on plurality of pictures uniformity | |
| CN107358146B (en) | Method for processing video frequency, device and storage medium | |
| CN107451510A (en) | Biopsy method and In vivo detection system | |
| CN104205865B (en) | Method and apparatus for certification video content | |
| US20200143155A1 (en) | High Definition Camera and Image Recognition System for Criminal Identification | |
| CN106156702A (en) | Identity identifying method and equipment | |
| CN104636764B (en) | A kind of image latent writing analysis method and its device | |
| Tian et al. | Daxing smartphone identification dataset | |
| CN111104833A (en) | Method and apparatus for in vivo examination, storage medium, and electronic device | |
| CN106791655A (en) | A kind of method for processing video frequency and device | |
| CN106961434A (en) | One kind carries out fingerprint modeling for wireless device and knows method for distinguishing | |
| CN112801054B (en) | Face recognition model processing method, face recognition method and device | |
| CN110460502A (en) | Application rs traffic recognition methods under VPN based on distribution characteristics random forest | |
| CN107092821A (en) | A kind of distributed face authentication information generating method, authentication method and device | |
| CN107169458A (en) | Data processing method, device and storage medium | |
| CN103020655B (en) | A kind of remote identity authentication method based on single training image per person | |
| CN109815813A (en) | Image processing method and Related product | |
| CN103886235A (en) | Face image biological key generating method | |
| CN110298240A (en) | A kind of user vehicle recognition methods, device, system and storage medium | |
| Li et al. | Forensic analysis of digital images over smart devices and online social networks | |
| CN112507986B (en) | Multi-channel human face in-vivo detection method and device based on neural network | |
| CN113642639B (en) | Living body detection method, living body detection device, living body detection equipment and storage medium | |
| CN108683901B (en) | Data processing method, MEC server and computer readable storage medium | |
| Su et al. | Secure image storage system based on compressed sensing and 2D-SLLIM in cloud environment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170517 |