CN108763373A - Research on face image retrieval and device - Google Patents
Research on face image retrieval and device Download PDFInfo
- Publication number
- CN108763373A CN108763373A CN201810476005.XA CN201810476005A CN108763373A CN 108763373 A CN108763373 A CN 108763373A CN 201810476005 A CN201810476005 A CN 201810476005A CN 108763373 A CN108763373 A CN 108763373A
- Authority
- CN
- China
- Prior art keywords
- image
- feature
- retrieved
- model
- facial
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Processing Or Creating Images (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
A kind of Research on face image retrieval of present invention offer and device, are applied to image retrieval server.The Feature Selection Model and the Model Fusion characteristics of image corresponding to this feature extraction model that image retrieval server storage is useful for the blending image feature merged by multiple target image characteristics in extraction target facial image, the method includes：Feature based extraction model carries out feature extraction to facial image to be retrieved, obtains the blending image feature to be retrieved in facial image to be retrieved；The blending image feature to be retrieved extracted is compared with Model Fusion characteristics of image, and judges whether facial image to be retrieved meets Feature Selection Model according to comparison result；Output meets the facial image to be retrieved of Feature Selection Model as the successful facial image of retrieval.The retrieval cost of implementation of the method is low, and retrieval workload is small, quickly accurately can be concentrated from a large amount of face image data and retrieve the facial image for including multiple target image characteristics.
Description
Technical field
The present invention relates to image retrieval technologies fields, in particular to a kind of Research on face image retrieval and device.
Background technology
In image retrieval procedure, it will usually encounter that need to filter out from a large amount of image all include multiple mesh
The case where image of logo image feature, for example, it is to wear glasses and close one's eyes to retrieve target image characteristics in plurality of human faces image of comforming
The facial image of eyeball.For such situation, the prior art can only be retrieved for single target characteristics of image, be generally required
Retrieval personnel artificially carry out feature mark to facial image, with it is corresponding retrieve include multiple target image characteristics face
Image.But this retrieval mode can make include multiple target image characteristics facial image different images screen feature under answered
With, retrieval personnel will carry out multiple repeat mark for same facial image, and the retrieval of whole image retrieving is of high cost,
The heavy workload of retrieval personnel often can not have higher retrieval effect because of the limitation of the experience of retrieval mark personnel and knowledge
Rate and retrieval precision.
Invention content
In order to overcome above-mentioned deficiency in the prior art, the purpose of the present invention is to provide a kind of Research on face image retrieval
And device, the retrieval cost of implementation of the Research on face image retrieval is low, and retrieval workload is small, can be quickly and accurately from big
The image data concentration of amount retrieves the facial image for including multiple target image characteristics.
For method, the present invention provides a kind of Research on face image retrieval compared with embodiment, is applied to image retrieval service
Device, described image retrieval server, which is stored with, to be merged by multiple target image characteristics for extracting in target facial image
The Feature Selection Model and the Model Fusion characteristics of image corresponding to this feature extraction model of blending image feature, the method
Including：
Feature extraction is carried out to facial image to be retrieved based on the Feature Selection Model, obtains the face figure to be retrieved
The blending image feature to be retrieved corresponding with this feature extraction model as in；
The blending image feature to be retrieved Model Fusion image corresponding with the Feature Selection Model that will be extracted
Feature is compared, and judges whether the facial image to be retrieved meets the Feature Selection Model according to comparison result；
Output meets the facial image to be retrieved of the Feature Selection Model as the successful facial image of retrieval.
Optionally, in embodiments of the present invention, above-mentioned image retrieval server is also stored with default similarity threshold, described
The blending image feature to be retrieved Model Fusion characteristics of image corresponding with the Feature Selection Model extracted is carried out
It compares, judges that the step of whether facial image to be retrieved meets the Feature Selection Model includes according to comparison result：
Calculate the characteristic similarity between the blending image feature to be retrieved and the Model Fusion characteristics of image；
The characteristic similarity being calculated is compared with the default similarity threshold, and in the feature phase
When being greater than or equal to the default similarity threshold like degree, judge that the corresponding facial image to be retrieved meets the feature and carries
Modulus type.
Optionally, in embodiments of the present invention, it is described based on the Feature Selection Model to facial image to be retrieved into
Before the step of row feature extraction, the method further includes：
Multiple sample facial images are obtained, and with multiple described sample facial images to recognition of face convolutional neural networks mould
Type is trained, and obtains the Feature Selection Model, wherein including by the multiple target in every sample facial image
The blending image feature that multi-features obtain.
Optionally, in embodiments of the present invention, the above method further includes：
Every sample facial image institute is extracted from multiple described sample facial images based on the Feature Selection Model
Corresponding blending image feature, and average value processing is carried out to multiple blending image features, obtain the Feature Selection Model
Corresponding Model Fusion characteristics of image.
Optionally, in embodiments of the present invention, the above method further includes：
Default similarity threshold is configured.
For device, the embodiment of the present invention provides a kind of facial image retrieval device, is applied to image retrieval server,
Described image retrieval server, which is stored with, to be melted for extracting in target facial image by what multiple target image characteristics merged
The Feature Selection Model and the Model Fusion characteristics of image corresponding to this feature extraction model for closing characteristics of image, described device packet
It includes：
Characteristic extracting module is obtained for carrying out feature extraction to facial image to be retrieved based on the Feature Selection Model
To blending image feature to be retrieved corresponding with this feature extraction model in the facial image to be retrieved；
Feature judgment module, for by the blending image feature to be retrieved extracted and the Feature Selection Model pair
The Model Fusion characteristics of image answered is compared, and it is described according to comparison result to judge whether the facial image to be retrieved meets
Feature Selection Model；
Image output module meets the facial image to be retrieved of the Feature Selection Model as retrieving successfully for exporting
Facial image.
Optionally, in embodiments of the present invention, above-mentioned image retrieval server is also stored with default similarity threshold, described
The blending image feature to be retrieved Model Fusion corresponding with the Feature Selection Model that feature judgment module will extract
Characteristics of image is compared, and judges whether the facial image to be retrieved meets the Feature Selection Model according to comparison result
Mode includes：
Calculate the characteristic similarity between the blending image feature to be retrieved and the Model Fusion characteristics of image；
The characteristic similarity being calculated is compared with the default similarity threshold, and in the feature phase
When being greater than or equal to the default similarity threshold like degree, judge that the corresponding facial image to be retrieved meets the feature and carries
Modulus type.
Optionally, in embodiments of the present invention, above-mentioned apparatus further includes：
Model training module, for obtaining multiple sample facial images, and with multiple described sample facial images to face
Identification convolutional neural networks model is trained, and the Feature Selection Model is obtained, wherein being wrapped in every sample facial image
Include the blending image feature merged by the multiple target image characteristics.
Optionally, in embodiments of the present invention, above-mentioned apparatus further includes：
Feature configuration module extracts often for being based on the Feature Selection Model from multiple described sample facial images
The blending image feature corresponding to sample facial image is opened, and average value processings are carried out to multiple blending image features, is obtained
The corresponding Model Fusion characteristics of image of the Feature Selection Model.
Optionally, in embodiments of the present invention, above-mentioned apparatus further includes：
Threshold value configuration module, for being configured to default similarity threshold.
In terms of existing technologies, Research on face image retrieval and device provided in an embodiment of the present invention have with following
Beneficial effect：The retrieval cost of implementation of the Research on face image retrieval is low, and retrieval workload is small, can be quickly and accurately from big
The image data concentration of amount retrieves the facial image for including multiple target image characteristics.The method takes applied to image retrieval
It is engaged in device, is stored in the image retrieval server and merges to obtain by multiple target image characteristics for extracting in target facial image
Blending image feature Feature Selection Model and the Model Fusion characteristics of image corresponding to this feature extraction model.First, institute
It states method and feature extraction is carried out to facial image to be retrieved by the Feature Selection Model, obtain the facial image to be retrieved
In blending image feature to be retrieved corresponding with this feature extraction model.Then, the method is described to be retrieved by what is extracted
Blending image feature Model Fusion characteristics of image corresponding with the Feature Selection Model is compared, and is sentenced according to comparison result
Whether the facial image to be retrieved that breaks meets the Feature Selection Model.Finally, the method output meets the feature and carries
The facial image to be retrieved of modulus type is as the successful facial image of retrieval.The method is by being used to extract target facial image
In the Feature Selection Model of blending image feature that is merged by multiple target image characteristics in face image set to be retrieved
Each facial image carries out feature extraction, judges whether each facial image is to meet this feature extraction according to the fusion feature extracted
Model, and export when image meets this feature extraction model the facial image be retrieval successfully with the multiple target image
The corresponding facial image of feature, to quickly accurately be retrieved from a large amount of image data concentration including multiple target images
The facial image of feature reduces retrieval workload to reduce retrieval cost of implementation.
To enable the above objects, features and advantages of the present invention to be clearer and more comprehensible, present pre-ferred embodiments cited below particularly,
And coordinate appended attached drawing, it is described in detail below.
Description of the drawings
In order to illustrate the technical solution of the embodiments of the present invention more clearly, below will be to needed in the embodiment attached
Figure is briefly described, it should be understood that the following drawings illustrates only certain embodiments of the present invention, therefore is not construed as pair
The restriction of the claims in the present invention protection domain, for those of ordinary skill in the art, what is do not made the creative labor
Under the premise of, it can also be obtained according to these attached drawings other relevant attached drawings.
Fig. 1 is a kind of block diagram of image retrieval server provided in an embodiment of the present invention.
Fig. 2 is the first flow diagram of Research on face image retrieval provided in an embodiment of the present invention.
Fig. 3 is second of flow diagram of Research on face image retrieval provided in an embodiment of the present invention.
Fig. 4 is the first block diagram that facial image retrieves device shown in Fig. 1 provided in an embodiment of the present invention.
Fig. 5 is second of block diagram that facial image retrieves device shown in Fig. 1 provided in an embodiment of the present invention.
Icon：10- image retrieval servers；11- memories；12- processors；13- communication units；100- facial images are examined
Rope device；110- characteristic extracting modules；120- feature judgment modules；130- image output modules；140- model training modules；
150- feature configuration modules；160- threshold value configuration modules.
Specific implementation mode
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.The present invention being usually described and illustrated herein in the accompanying drawings is implemented
The component of example can be arranged and be designed with a variety of different configurations.
Therefore, below the detailed description of the embodiment of the present invention to providing in the accompanying drawings be not intended to limit it is claimed
The scope of the present invention, but be merely representative of the present invention selected embodiment.Based on the embodiments of the present invention, this field is common
The every other embodiment that technical staff is obtained without creative efforts belongs to the model that the present invention protects
It encloses.
It should be noted that：Similar label and letter indicate similar terms in following attached drawing, therefore, once a certain Xiang Yi
It is defined, then it further need not be defined and explained in subsequent attached drawing in a attached drawing.
In the description of the present invention, it should be noted that term " first ", " second ", " third " etc. are only used for distinguishing and retouch
It states, is not understood to indicate or imply relative importance.
Below in conjunction with the accompanying drawings, it elaborates to some embodiments of the present invention.In the absence of conflict, following
Feature in embodiment and embodiment can be combined with each other.
Fig. 1 is please referred to, is a kind of block diagram of image retrieval server 10 provided in an embodiment of the present invention.In this hair
In bright embodiment, described image retrieval server 10 from a large amount of face image datas concentration for quickly accurately filtering out institute
Some includes the facial image of multiple target image characteristics, wherein the target image characteristics are successfully retrieved when being image retrieval
The common images feature that obtained face images all have, the target image characteristics can be the features worn glasses,
It can be the feature of eye closing eyeball, can also be that the feature of dew tooth, specific characteristics of image situation can be carried out according to Search Requirement
Different configurations.
In the present embodiment, described image retrieval server 10 includes facial image retrieval device 100, memory 11, place
Manage device 12 and communication unit 13.The memory 11, processor 12 and 13 each element of communication unit between each other directly or
Ground connection is electrically connected, to realize the transmission or interaction of data.For example, these elements can pass through one or more communication between each other
Bus or signal wire, which are realized, to be electrically connected.Facial image retrieval device 100 include it is at least one can be with software or firmware
(firmware) form is stored in the software function module in the memory 11, and the processor 12 is stored in by operation
The facial image in memory 11 retrieves 100 corresponding software function module of device, to perform various functions using with
And data processing.
In the present embodiment, the memory 11 can be used for storing synthesized by multiple face image sets to be retrieved it is to be retrieved
Face image data collection can be used for storage and merge to obtain by multiple target image characteristics for extracting in target facial image
Blending image feature Feature Selection Model, the Model Fusion image that can also be stored corresponding to this feature extraction model is special
Sign, wherein the blending image is characterized as being provided simultaneously with the feature of the multiple target image characteristics, the Model Fusion image
Be characterized in based on be used for training generate this feature extraction model face images get with this feature extraction model
Corresponding blending image feature.The memory 11 may be, but not limited to, random access memory (Random Access
Memory, RAM), read-only memory (Read Only Memory, ROM), programmable read only memory (Programmable
Read-Only Memory, PROM), Erasable Programmable Read Only Memory EPROM (Erasable Programmable Read-Only
Memory, EPROM), electrically erasable programmable read-only memory (Electric Erasable Programmable Read-
Only Memory, EEPROM) etc..Wherein, memory 11 can be also used for storing various application programs, and the processor 12 exists
It receives after executing instruction, executes the application program.Further, the software program in above-mentioned memory 11 and module be also
May include operating system, may include it is various for manage system task (such as memory management, storage device control, power supply pipe
Reason etc.) component software and/or driving, and can be in communication with each other with various hardware or component software, to provide other software group
The running environment of part.
In the present embodiment, the processor 12 can be a kind of IC chip with signal handling capacity.Its
Described in processor 12 can be general processor, including central processing unit (Central Processing Unit, CPU), net
Network processor (Network Processor, NP) etc..It may be implemented or execute the disclosed each side in the embodiment of the present invention
Method, step and logic diagram.General processor can be microprocessor or the processor can also be any conventional processing
Device etc..
In the present embodiment, the communication unit 13 is used to establish described image retrieval server 10 and other by network
Communication connection between external equipment, and carried out data transmission by the network, wherein described image retrieval server 10 can
Facial image to be retrieved is obtained from other described external equipments by the communication unit 13, or from a large amount of face figures to be retrieved
By what is retrieved include that the face images of multiple target image characteristics are sent to the external equipment as in.
In the present embodiment, described image retrieval server 10 is examined by the facial image being stored in the memory 11
Rope device 100 quickly accurately retrieves all people for including multiple target image characteristics from a large amount of facial images to be retrieved
Face image reduces retrieval workload to reduce retrieval cost of implementation.
It is understood that structure shown in FIG. 1 is only a kind of structural schematic diagram of image retrieval server 10, the figure
It may also include than shown in Fig. 1 more either less components as retrieval server 10 or match with different from shown in Fig. 1
It sets.Hardware, software, or its combination realization may be used in each component shown in Fig. 1.
Fig. 2 is please referred to, is the first flow diagram of Research on face image retrieval provided in an embodiment of the present invention.At this
In inventive embodiments, the Research on face image retrieval is applied to image retrieval server 10 shown in Fig. 1, wherein the figure
As being stored in retrieval server 10 for extracting the fusion merged by multiple target image characteristics in target facial image
The Feature Selection Model and the Model Fusion characteristics of image corresponding to this feature extraction model of characteristics of image, below to shown in Fig. 2
Research on face image retrieval detailed process and step be described in detail.
In embodiments of the present invention, the Research on face image retrieval includes the following steps：
Step S210, feature based extraction model carry out feature extraction to facial image to be retrieved, obtain described to be retrieved
Blending image feature to be retrieved corresponding with this feature extraction model in facial image.
In the present embodiment, the blending image to be retrieved is characterized as in facial image to be retrieved for judging that this is to be retrieved
Whether facial image is the fusion feature for retrieving object.The Feature Selection Model pair that described image retrieval server 10 passes through storage
Face image data to be retrieved concentrates each facial image to be retrieved to carry out feature extraction, obtains in each facial image to be retrieved and is somebody's turn to do
The corresponding blending image feature to be retrieved of Feature Selection Model.
Step S220, the blending image feature to be retrieved model corresponding with the Feature Selection Model that will be extracted
Blending image feature is compared, and judges whether the facial image to be retrieved meets the feature extraction according to comparison result
Model.
In the present embodiment, described image retrieval server 10 is also stored with for judging whether facial image to be retrieved accords with
Close the default similarity threshold of Feature Selection Model.The fusion figure to be retrieved that described image retrieval server 10 will extract
As feature Model Fusion characteristics of image corresponding with the Feature Selection Model is compared, and described in being judged according to comparison result
The step of whether facial image to be retrieved meets the Feature Selection Model include：
Calculate the characteristic similarity between the blending image feature to be retrieved and the Model Fusion characteristics of image；
The characteristic similarity being calculated is compared with the default similarity threshold, and in the feature phase
When being greater than or equal to the default similarity threshold like degree, judge that the corresponding facial image to be retrieved meets the feature and carries
Modulus type.
Wherein, COS distance similarity calculating method, local sensitivity Hash can be used in described image retrieval server 10
(LSH) at least one of method, product quantization (PQ) method method calculate blending image feature to be retrieved and melt with the model
Close the characteristic similarity between characteristics of image.Described image retrieval server 10 is by the characteristic similarity being calculated and institute
It states default similarity threshold to be compared, and judges when the characteristic similarity is greater than or equal to the default similarity threshold
Facial image to be retrieved corresponding to this feature similarity meets the Feature Selection Model.
Step S230, output meet the facial image to be retrieved of the Feature Selection Model as the successful face figure of retrieval
Picture.
In the present embodiment, characteristic similarity is greater than or equal to the default similarity by described image retrieval server 10
The facial image to be retrieved of threshold value be determined as include multiple target image characteristics the face for meeting the Feature Selection Model
Image, and the facial image output to be retrieved for meeting the Feature Selection Model is used as and retrieves successful facial image,
The facial image exported at this time be and meanwhile include the multiple target image characteristics facial image.
Fig. 3 is please referred to, is second of flow diagram of Research on face image retrieval provided in an embodiment of the present invention.At this
In inventive embodiments, before step S210, the Research on face image retrieval further includes step S208 and step S209.
Step S208 obtains multiple sample facial images, and with multiple described sample facial images to recognition of face convolution
Neural network model is trained, and obtains the Feature Selection Model.
In the present embodiment, the sample facial image be simultaneously include the multiple target image characteristics be used for instruct
The facial image of Feature Selection Model is got, there are merged by the target image characteristics in the sample facial image at this time
Obtained blending image feature.Described image retrieval server 10 knows face by multiple the sample facial images that will be got
The mode that other convolutional neural networks model is trained, obtains the Feature Selection Model.
Step S209 extracts every sample people based on the Feature Selection Model from multiple described sample facial images
Blending image feature corresponding to face image, and average value processing is carried out to multiple blending image features, obtain the feature
The corresponding Model Fusion characteristics of image of extraction model.
In the present embodiment, described image retrieval server 10 is extracted by the Feature Selection Model from training this feature
The blending image extracted in multiple the above-mentioned sample facial images used when model corresponding to every sample facial image is special
Sign, and average value processing is carried out to multiple blending image features, obtain corresponding model corresponding with the Feature Selection Model
Blending image feature.
In embodiments of the present invention, before step S208, the Research on face image retrieval further includes step S207.
Step S207 configures default similarity threshold.
In the present embodiment, described image retrieval server 10 can be by way of configuring default similarity threshold
Facial image search criteria are marked off, wherein the default similarity threshold can be 70%, can also be 85%, can also be
95%, specific numerical value can carry out different configurations according to demand.
Please refer to Fig. 4, be shown in Fig. 1 provided in an embodiment of the present invention facial image retrieval device 100 the first
Block diagram.In embodiments of the present invention, the facial image retrieval device 100 is sentenced including characteristic extracting module 110, feature
Disconnected module 120 and image output module 130.
The characteristic extracting module 110 carries out feature extraction for feature based extraction model to facial image to be retrieved,
Obtain blending image feature to be retrieved corresponding with this feature extraction model in the facial image to be retrieved.
In the present embodiment, the characteristic extracting module 110 can execute step S210 shown in Fig. 2, specifically hold
Row process is referred to above to the detailed description of step S210.
The feature judgment module 120, for putting forward the blending image feature to be retrieved extracted with the feature
The corresponding Model Fusion characteristics of image of modulus type is compared, and whether judges the facial image to be retrieved according to comparison result
Meet the Feature Selection Model.
In the present embodiment, described image retrieval server 10 is also stored with for judging whether facial image to be retrieved accords with
Close the default similarity threshold of Feature Selection Model.The fusion figure to be retrieved that the feature judgment module 120 will extract
As feature Model Fusion characteristics of image corresponding with the Feature Selection Model is compared, according to comparison result judge described in wait for
Whether retrieval facial image, which meets the mode of the Feature Selection Model, includes：
Calculate the characteristic similarity between the blending image feature to be retrieved and the Model Fusion characteristics of image；
The characteristic similarity being calculated is compared with the default similarity threshold, and in the feature phase
When being greater than or equal to the default similarity threshold like degree, judge that the corresponding facial image to be retrieved meets the feature and carries
Modulus type.
Described image output module 130, for exporting the facial image conduct to be retrieved for meeting the Feature Selection Model
Retrieve successful facial image.
In the present embodiment, described image output module 130 can execute step S230 shown in Fig. 2, specifically hold
Row process is referred to above to the detailed description of step S230.
Fig. 5 is please referred to, is second of facial image retrieval device 100 shown in Fig. 1 provided in an embodiment of the present invention
Block diagram.In embodiments of the present invention, facial image retrieval device 100 can also include model training module 140,
Feature configuration module 150 and threshold value configuration module 160.
The model training module 140, for obtaining multiple sample facial images, and with multiple described sample facial images
Recognition of face convolutional neural networks model is trained, the Feature Selection Model is obtained.
In the present embodiment, include to merge to obtain by the multiple target image characteristics in every sample facial image
Blending image feature, the model training module 140 can execute step S208 shown in Fig. 3, specific implementation procedure
It is referred to above to the detailed description of step S208.
The feature configuration module 150, for being based on the Feature Selection Model from multiple described sample facial images
The blending image feature corresponding to every sample facial image is extracted, and multiple blending image features are carried out at mean value
Reason, obtains the corresponding Model Fusion characteristics of image of the Feature Selection Model.
In the present embodiment, the feature configuration module 150 can execute step S209 shown in Fig. 3, specifically hold
Row process is referred to above to the detailed description of step S209.
The threshold value configuration module 160, for being configured to default similarity threshold.
In the present embodiment, the threshold value configuration module 160 can execute step S207 shown in Fig. 3, specifically hold
Row process is referred to above to the detailed description of step S207.
In conclusion in Research on face image retrieval provided in an embodiment of the present invention and device, the facial image inspection
The retrieval cost of implementation of Suo Fangfa is low, and retrieval workload is small, quickly accurately can concentrate and retrieve from a large amount of image data
Go out the facial image for including multiple target image characteristics.The method is applied to image retrieval server, the image retrieval service
The spy for extracting the blending image feature merged by multiple target image characteristics in target facial image is stored in device
Levy the Model Fusion characteristics of image corresponding to extraction model and this feature extraction model.First, the method passes through the feature
Extraction model to facial image to be retrieved carry out feature extraction, obtain in the facial image to be retrieved with this feature extraction model
Corresponding blending image feature to be retrieved.Then, the method by the blending image feature to be retrieved extracted with it is described
The corresponding Model Fusion characteristics of image of Feature Selection Model is compared, and judges the face figure to be retrieved according to comparison result
Seem no to meet the Feature Selection Model.Finally, the method output meets the face to be retrieved of the Feature Selection Model
Image is as the successful facial image of retrieval.The method is by being used to extract in target facial image by multiple target images spy
The Feature Selection Model for the blending image feature that sign fusion obtains carries out feature to each facial image in face image set to be retrieved
Extraction, judges whether each facial image is to meet this feature extraction model, and meet in image according to the fusion feature extracted
It is retrieval successfully face figure corresponding with the multiple target image characteristics that the facial image is exported when this feature extraction model
Picture, to quickly accurately be concentrated from a large amount of image data and retrieve the facial image for including multiple target image characteristics,
Cost of implementation is retrieved to reduce, reduces retrieval workload.
The foregoing is only a preferred embodiment of the present invention, is not intended to restrict the invention, for the skill of this field
For art personnel, the invention may be variously modified and varied.All within the spirits and principles of the present invention, any made by repair
Change, equivalent replacement, improvement etc., should all be included in the protection scope of the present invention.
Claims (10)
1. a kind of Research on face image retrieval, which is characterized in that be applied to image retrieval server, described image retrieval server
It is stored with and is carried for extracting the feature of the blending image feature merged by multiple target image characteristics in target facial image
Model Fusion characteristics of image corresponding to modulus type and this feature extraction model, the method includes：
Feature extraction is carried out to facial image to be retrieved based on the Feature Selection Model, is obtained in the facial image to be retrieved
Blending image feature to be retrieved corresponding with this feature extraction model；
The blending image feature to be retrieved Model Fusion characteristics of image corresponding with the Feature Selection Model that will be extracted
It is compared, and judges whether the facial image to be retrieved meets the Feature Selection Model according to comparison result；
Output meets the facial image to be retrieved of the Feature Selection Model as the successful facial image of retrieval.
2. according to the method described in claim 1, it is characterized in that, described image retrieval server is also stored with default similarity
Threshold value, the blending image feature to be retrieved Model Fusion image corresponding with the Feature Selection Model that will be extracted
Feature is compared, and the step of whether facial image to be retrieved meets the Feature Selection Model judged according to comparison result
Including：
Calculate the characteristic similarity between the blending image feature to be retrieved and the Model Fusion characteristics of image；
The characteristic similarity being calculated is compared with the default similarity threshold, and in the characteristic similarity
When more than or equal to the default similarity threshold, judge that the corresponding facial image to be retrieved meets the feature extraction mould
Type.
3. method according to claim 1 or 2, which is characterized in that be based on the Feature Selection Model to be checked described
Before rope facial image carries out the step of feature extraction, the method further includes：
Obtain multiple sample facial images, and with multiple described sample facial images to recognition of face convolutional neural networks model into
Row training, obtains the Feature Selection Model, wherein including by the multiple target image in every sample facial image
The blending image feature that Fusion Features obtain.
4. according to the method described in claim 3, it is characterized in that, the method further includes：
It is extracted corresponding to every sample facial image from multiple described sample facial images based on the Feature Selection Model
Blending image feature, and average value processings are carried out to multiple blending image features, obtain the Feature Selection Model and correspond to
Model Fusion characteristics of image.
5. according to the method described in claim 4, it is characterized in that, the method further includes：
Default similarity threshold is configured.
6. a kind of facial image retrieves device, which is characterized in that be applied to image retrieval server, described image retrieval server
It is stored with and is carried for extracting the feature of the blending image feature merged by multiple target image characteristics in target facial image
Model Fusion characteristics of image corresponding to modulus type and this feature extraction model, described device include：
Characteristic extracting module obtains institute for carrying out feature extraction to facial image to be retrieved based on the Feature Selection Model
State blending image feature to be retrieved corresponding with this feature extraction model in facial image to be retrieved；
Feature judgment module, for the blending image feature to be retrieved extracted is corresponding with the Feature Selection Model
Model Fusion characteristics of image is compared, and judges whether the facial image to be retrieved meets the feature according to comparison result
Extraction model；
Image output module, for exporting the facial image to be retrieved for meeting the Feature Selection Model as the successful people of retrieval
Face image.
7. device according to claim 6, which is characterized in that described image retrieval server is also stored with default similarity
Threshold value, the feature judgment module are corresponding with the Feature Selection Model by the blending image feature to be retrieved extracted
Model Fusion characteristics of image is compared, and judges whether the facial image to be retrieved meets the feature and carry according to comparison result
The mode of modulus type includes：
Calculate the characteristic similarity between the blending image feature to be retrieved and the Model Fusion characteristics of image；
The characteristic similarity being calculated is compared with the default similarity threshold, and in the characteristic similarity
When more than or equal to the default similarity threshold, judge that the corresponding facial image to be retrieved meets the feature extraction mould
Type.
8. the device described according to claim 6 or 7, which is characterized in that described device further includes：
Model training module, for obtaining multiple sample facial images, and with multiple described sample facial images to recognition of face
Convolutional neural networks model is trained, and obtains the Feature Selection Model, wherein including in every sample facial image
The blending image feature merged by the multiple target image characteristics.
9. device according to claim 8, which is characterized in that described device further includes：
Feature configuration module extracts every sample for being based on the Feature Selection Model from multiple described sample facial images
Blending image feature corresponding to this facial image, and average value processing is carried out to multiple blending image features, it obtains described
The corresponding Model Fusion characteristics of image of Feature Selection Model.
10. device according to claim 9, which is characterized in that described device further includes：
Threshold value configuration module, for being configured to default similarity threshold.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810476005.XA CN108763373A (en) | 2018-05-17 | 2018-05-17 | Research on face image retrieval and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810476005.XA CN108763373A (en) | 2018-05-17 | 2018-05-17 | Research on face image retrieval and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108763373A true CN108763373A (en) | 2018-11-06 |
Family
ID=64006897
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810476005.XA Pending CN108763373A (en) | 2018-05-17 | 2018-05-17 | Research on face image retrieval and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108763373A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110781842A (en) * | 2019-10-29 | 2020-02-11 | 深圳市商汤科技有限公司 | Image processing method and device, electronic equipment and storage medium |
| CN110795592A (en) * | 2019-10-28 | 2020-02-14 | 深圳市商汤科技有限公司 | Picture processing method, device and equipment |
| CN112445928A (en) * | 2019-08-29 | 2021-03-05 | 华为技术有限公司 | Image retrieval method and device |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104899579A (en) * | 2015-06-29 | 2015-09-09 | 小米科技有限责任公司 | Face recognition method and face recognition device |
| US20160224826A1 (en) * | 2006-12-01 | 2016-08-04 | Google Inc. | Identifying Images Using Face Recognition |
| CN106022317A (en) * | 2016-06-27 | 2016-10-12 | 北京小米移动软件有限公司 | Face identification method and apparatus |
| CN107944020A (en) * | 2017-12-11 | 2018-04-20 | 深圳云天励飞技术有限公司 | Facial image lookup method and device, computer installation and storage medium |
| CN107958244A (en) * | 2018-01-12 | 2018-04-24 | 成都视观天下科技有限公司 | A kind of face identification method and device based on the fusion of video multiframe face characteristic |
-
2018
- 2018-05-17 CN CN201810476005.XA patent/CN108763373A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160224826A1 (en) * | 2006-12-01 | 2016-08-04 | Google Inc. | Identifying Images Using Face Recognition |
| CN104899579A (en) * | 2015-06-29 | 2015-09-09 | 小米科技有限责任公司 | Face recognition method and face recognition device |
| CN106022317A (en) * | 2016-06-27 | 2016-10-12 | 北京小米移动软件有限公司 | Face identification method and apparatus |
| CN107944020A (en) * | 2017-12-11 | 2018-04-20 | 深圳云天励飞技术有限公司 | Facial image lookup method and device, computer installation and storage medium |
| CN107958244A (en) * | 2018-01-12 | 2018-04-24 | 成都视观天下科技有限公司 | A kind of face identification method and device based on the fusion of video multiframe face characteristic |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112445928A (en) * | 2019-08-29 | 2021-03-05 | 华为技术有限公司 | Image retrieval method and device |
| CN110795592A (en) * | 2019-10-28 | 2020-02-14 | 深圳市商汤科技有限公司 | Picture processing method, device and equipment |
| WO2021082505A1 (en) * | 2019-10-28 | 2021-05-06 | 深圳市商汤科技有限公司 | Picture processing method, apparatus and device, storage medium, and computer program |
| CN110781842A (en) * | 2019-10-29 | 2020-02-11 | 深圳市商汤科技有限公司 | Image processing method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111860506B (en) | Method and device for recognizing characters | |
| CN108491805B (en) | Identity authentication method and device | |
| JP6994588B2 (en) | Face feature extraction model training method, face feature extraction method, equipment, equipment and storage medium | |
| CN110472510A (en) | Based on infrared and visual picture electrical equipment fault detection method and assessment equipment | |
| CN108763373A (en) | Research on face image retrieval and device | |
| CN107609598A (en) | Image authentication model training method, device and readable storage medium storing program for executing | |
| CN109614517A (en) | Classification method, device, equipment and the storage medium of video | |
| CN111814775B (en) | Target object abnormal behavior identification method, device, terminal and storage medium | |
| CN108108711B (en) | Face control method, electronic device and storage medium | |
| CN109344864B (en) | Image processing method and device for dense object | |
| CN111091109A (en) | Method, system and equipment for predicting age and gender based on face image | |
| CN109522775A (en) | Face character detection method, device and electronic equipment | |
| CN110442712A (en) | Determination method, apparatus, server and the text of risk try system | |
| CN105956469A (en) | Method and device for identifying file security | |
| CN108460346A (en) | Fingerprint identification method and device | |
| CN109740553B (en) | Image semantic segmentation data screening method and system based on recognition | |
| US7679636B1 (en) | Apparatus and method for routing telecommunication calls using visual caller information | |
| WO2023040146A1 (en) | Behavior recognition method and apparatus based on image fusion, and electronic device and medium | |
| US20240185409A1 (en) | Systems and Methods for Troubleshooting Equipment Installation Using Artificial Intelligence | |
| CN107885637B (en) | A kind of server exception detection method and device | |
| CN110276315A (en) | Airport monitoring method, apparatus and system | |
| CN111310837A (en) | Vehicle refitting recognition method, device, system, medium and equipment | |
| CN110490118A (en) | Image processing method and device | |
| CN106960188B (en) | Weather image classification method and device | |
| CN111783677A (en) | Face recognition method, face recognition device, server and computer readable medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181106 |
|
| RJ01 | Rejection of invention patent application after publication |