CN109190532A - It is a kind of based on cloud side fusion face identification method, apparatus and system - Google Patents
It is a kind of based on cloud side fusion face identification method, apparatus and system Download PDFInfo
- Publication number
- CN109190532A CN109190532A CN201810957725.8A CN201810957725A CN109190532A CN 109190532 A CN109190532 A CN 109190532A CN 201810957725 A CN201810957725 A CN 201810957725A CN 109190532 A CN109190532 A CN 109190532A
- Authority
- CN
- China
- Prior art keywords
- face
- image
- cloud
- fusion
- obtains
- 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
- G06V40/161—Detection; Localisation; Normalisation
- G06V40/166—Detection; Localisation; Normalisation using acquisition arrangements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
Abstract
The present invention provides a kind of face identification methods based on the fusion of cloud side, apparatus and system, it is related to field of artificial intelligence, applied to marginal end, comprising: carry out Face datection respectively to from the multiple image in the video flowing that image capture device is sent, obtain multiple facial images；Face tracking is carried out to multiple described facial images respectively, obtains multiple face image sets, includes the face of the same person in the facial image of each face image set；For each face image set, face quality evaluation is carried out to the facial image in the face image set, obtains the target facial image for meeting default face quality requirement；The corresponding target facial image of each face image set is sent to cloud, so that the cloud carries out recognition of face, the heavy burden of transmission network is alleviated, solves the problems, such as data redundancy, improve the utilization rate and validity of cloud data.
Description
Technical field
The present invention relates to field of artificial intelligence, more particularly, to a kind of face identification method based on the fusion of cloud side,
Apparatus and system.
Background technique
The intelligence degree of existing acquisition equipment (such as high-definition camera) is lower at present, and major function is data acquisition, will
Collected data return server storage by network transmission, provide support for case investigation in special circumstances, evidence obtaining.
However, such as directly being transmitted, then due to high-definition camera acquired image image resolution ratio with higher
The video data that transmission network can be brought heavy burden, and be transmitted back to server end has a redundancy of height, 99% with
On video data be all invalid data.
Summary of the invention
In view of this, the purpose of the present invention is to provide a kind of face identification method based on the fusion of cloud side, device and being
Acquired image is directly transmitted back to video data caused by server end with high redundancy to alleviate the prior art by system
The problem of, mitigate the heavy burden of transmission network.
In a first aspect, being applied to edge the embodiment of the invention provides a kind of face identification method based on the fusion of cloud side
End, comprising:
Face datection is carried out respectively to from the multiple image in the video flowing that image capture device is sent, and obtains multiple faces
Image；
Face tracking is carried out to multiple described facial images respectively, obtains multiple face image sets, each face
It include the face of the same person in the facial image of image collection；
For each face image set, face quality is carried out to the facial image in the face image set and is commented
Estimate, obtains the target facial image for meeting default face quality requirement；
The corresponding target facial image of each face image set is sent to cloud, so that the cloud carries out face knowledge
Not.
With reference to first aspect, the embodiment of the invention provides the first possible embodiments of first aspect, wherein institute
It states and carries out Face datection respectively to from the multiple image in the video flowing that image capture device is sent, obtain multiple facial images,
Include:
Receive the video flowing sent from image capture device；
Using it is multi-layer biaxially oriented stream neural network to it is described acquisition equipment provide the video flowing in each frame data into
Row Face datection obtains multiple facial images.
With reference to first aspect, the embodiment of the invention provides second of possible embodiments of first aspect, wherein institute
Stating multi-layer biaxially oriented stream neural network includes forward stream neural network and reverse flow neural network；
The forward stream neural network includes: multiple positive network layers, and the forward stream neural network is with the more of video flowing
Frame image is input, exports top characteristic pattern；
The reverse flow neural network is carried out based on the output of the top characteristic pattern and each positive network layer
Fusion treatment obtains multiple fusion feature figures；Face datection is carried out in the fusion feature figure, obtains multiple facial images.
With reference to first aspect, the embodiment of the invention provides the third possible embodiments of first aspect, wherein institute
Stating reverse flow neural network includes: at least one reversed cascade fused layer, the fused layer also with the corresponding positive net
Network layers connection；
The fused layer carries out up-sampling treatment to the top characteristic pattern, obtains up-sampling characteristic pattern；By it is described just
The output of fused layer described in output and the up-sampling characteristic pattern or upper level to network layer is merged, and is merged
Characteristic pattern；Face datection is carried out in the fusion feature figure, obtains multiple facial images.
With reference to first aspect, the embodiment of the invention provides the 4th kind of possible embodiments of first aspect, wherein every
In a positive network layer respectively include: convolutional layer and pond layer；
The frame image that the convolutional layer will receive carries out feature extraction to the frame image using nonlinear activation, obtains
The fisrt feature figure is exported to the fisrt feature figure comprising feature vector, and to the pond layer；
The pond layer receives the fisrt feature figure, and compresses to the fisrt feature figure, obtains second feature
Figure.
With reference to first aspect, the embodiment of the invention provides the 5th kind of possible embodiments of first aspect, wherein institute
It states and face tracking is carried out respectively to multiple described facial images, obtain multiple face image sets, comprising:
If detecting same face in the multiple image of the video flowing, using the maximum under Kalman filtering constraint
Overlap mode tracks the face, obtains the face image set of the face.
With reference to first aspect, the embodiment of the invention provides the 6th kind of possible embodiments of first aspect, wherein institute
It states for each face image set, face quality evaluation is carried out to the facial image in the face image set, is obtained
To the target facial image for meeting default face quality requirement, comprising:
Obtain true resolution, length, width and the face area of each picture frame；
Face quality evaluation is carried out to the facial image in described image frame, true resolution is selected to be greater than default resolution ratio
Threshold value, length are greater than pre-set length threshold, width is greater than predetermined width threshold value, and face area is greater than default face area threshold
Picture frame as target facial image export, obtain the target facial image for meeting default face quality requirement.
Second aspect, the embodiment of the present invention also provide a kind of face identification device based on the fusion of cloud side, are applied to edge
End, comprising:
Face detection module, for carrying out face respectively to from the multiple image in the video flowing that image capture device is sent
Detection, obtains multiple facial images；
Face tracking module obtains multiple facial images for carrying out face tracking respectively to multiple described facial images
Gather, includes the face of the same person in the facial image of each face image set；
Quality assessment modules, for being directed to each face image set, to the face in the face image set
Image carries out face quality evaluation, obtains the target facial image for meeting default face quality requirement；
Sending module, for sending the corresponding target facial image of each face image set to cloud, so as to the cloud
End carries out recognition of face.
The third aspect, the embodiment of the present invention also provide a kind of face identification system based on the fusion of cloud side, wherein the system
System includes: marginal end described in cloud and first aspect and second aspect；
The cloud, the target facial image sent for receiving the marginal end, and it is stored in pre-set image memory block
Domain；
When receiving face to be retrieved, searched in the pre-set image storage region identical as the face to be retrieved
Facial image, obtain face recognition result.
Fourth aspect, the embodiment of the present invention also provide a kind of meter of non-volatile program code that can be performed with processor
Calculation machine readable medium, wherein said program code makes the processor execute method described in first aspect.
The embodiment of the present invention brings following the utility model has the advantages that marginal end is to the video flowing sent from image capture device first
In multiple image carry out Face datection respectively, obtain multiple facial images；Then multiple described facial images are carried out respectively
Face tracking obtains multiple face image sets, includes the same person's in the facial image of each face image set
Face；It is directed to each face image set again, face quality is carried out to the facial image in the face image set and is commented
Estimate, obtains the target facial image for meeting default face quality requirement；It is corresponding that each face image set finally is sent to cloud
Target facial image meet compared with image capture device is transmitted directly to the mode in cloud due to only sending in video flowing
The target facial image of default face quality requirement, without sending the facial image for being unsatisfactory for default face quality requirement, so
The transmission that video streaming image data can be reduced alleviates transmission network heavy burden, meanwhile, cloud is less what is received
When carrying out recognition of face in target facial image, the utilization rate and validity of cloud data can be improved, to efficiently solve
The problem of data redundancy.
Other features and advantages of the present invention will illustrate in the following description, also, partly become from specification
It obtains it is clear that understand through the implementation of the invention.The objectives and other advantages of the invention are in specification, claims
And specifically noted structure is achieved and obtained in attached drawing.
To enable the above objects, features and advantages of the present invention to be clearer and more comprehensible, preferred embodiment is cited below particularly, and cooperate
Appended attached drawing, is described in detail below.
Detailed description of the invention
It, below will be to specific in order to illustrate more clearly of the specific embodiment of the invention or technical solution in the prior art
Embodiment or attached drawing needed to be used in the description of the prior art be briefly described, it should be apparent that, it is described below
Attached drawing is some embodiments of the present invention, for those of ordinary skill in the art, before not making the creative labor
It puts, is also possible to obtain other drawings based on these drawings.
Fig. 1 is a kind of structure of marginal end based on the fusion of cloud side provided in an embodiment of the present invention；
Fig. 2 is a kind of face identification method based on the fusion of cloud side provided in an embodiment of the present invention；
Fig. 3 is a kind of multi-layer biaxially oriented stream neural network structure provided in an embodiment of the present invention；
Fig. 4 is a kind of face identification device based on the fusion of cloud side provided in an embodiment of the present invention；
Fig. 5 is a kind of face identification system based on the fusion of cloud side provided in an embodiment of the present invention.
Icon: 11- processor；12- memory；13- bus；14- communication interface；41- face detection module；42- face
Tracking module；43- quality assessment modules；44- sending module；51- marginal end；The cloud 52-.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with attached drawing to the present invention
Technical solution be clearly and completely described, it is clear that described embodiments are some of the embodiments of the present invention, rather than
Whole embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art are not making creative work premise
Under every other embodiment obtained, shall fall within the protection scope of the present invention.
Collected data are returned server storage by network transmission by current existing acquisition equipment, due to high-definition camera
Acquired image image resolution ratio with higher is such as directly transmitted, then can bring heavy burden to transmission network,
And the video data for being transmitted back to server end has the redundancy of height, 99% or more video data is all invalid data, base
In this, it is provided in an embodiment of the present invention it is a kind of based on cloud side fusion face identification method, apparatus and system, transmission can be mitigated
The heavy burden of network improves the utilization rate and validity of cloud data, solves the problems, such as data redundancy.
For convenient for understanding the present embodiment, first to a kind of based on the fusion of cloud side disclosed in the embodiment of the present invention
Face identification method describes in detail, is applied to marginal end, the marginal end includes: processor 11, memory 12, bus 13
With communication interface 14, the processor 11, communication interface 14 and memory 12 are connected by bus 13；Processor 11 is for executing
The executable module stored in memory 12, such as computer program.
Wherein, memory 12 may include high-speed random access memory (RAM, Random Access Memory),
It may further include the memory (non-volatile memory) of non-not volatibility, for example, at least a magnetic disk storage.Pass through
At least one communication interface 53 (can be wired or wireless) is realized between the system network element and at least one other network element
Communication connection, can be used internet, wide area network, local network, Metropolitan Area Network (MAN) etc..
Bus 13 can be isa bus, pci bus or eisa bus etc..The bus can be divided into address bus, data
Bus, control bus etc..Only to be indicated with a four-headed arrow convenient for indicating, in Fig. 1, it is not intended that an only bus or
A type of bus.
Wherein, memory 12 is for storing program, and the processor 11 executes the journey after receiving and executing instruction
Sequence, method performed by the device that the stream process that aforementioned any embodiment of the embodiment of the present invention discloses defines can be applied to handle
In device 11, or realized by processor 11.
Processor 11 may be a kind of IC chip, the processing capacity with signal.During realization, above-mentioned side
Each step of method can be completed by the integrated logic circuit of the hardware in processor 11 or the instruction of software form.Above-mentioned
Processor 11 can be general processor, including central processing unit (Central Processing Unit, abbreviation CPU), network
Processor (Network Processor, abbreviation NP) etc.；It can also be digital signal processor (Digital Signal
Processing, abbreviation DSP), specific integrated circuit (Application Specific Integrated Circuit, referred to as
ASIC), ready-made programmable gate array (Field-Programmable Gate Array, abbreviation FPGA) or other are programmable
Logical device, discrete gate or transistor logic, discrete hardware components.
In embodiments of the present invention, the marginal end and image capture device communicate to connect, and image capture device can be set
It sets in the shell of the marginal end, also can be set outside the shell of the marginal end, the marginal end is set with Image Acquisition
It is standby to be communicated to connect by wired mode, it can also communicate to connect wirelessly, for example, bluetooth connection etc., such as Fig. 2 institute
Show, may comprise steps of:
Step S201 carries out Face datection to from the multiple image in the video flowing that image capture device is sent respectively, obtains
To multiple facial images；
In embodiments of the present invention, image capture device can refer to camera etc., and preparatory trained multilayer can be used
Bidirectional flow neural network carries out Face datection, and facial image can refer to the figure in the face frame region in frame image comprising face
Picture.
It in this step, can be for each frame image in video flowing in order to more fully get each face
Face datection is carried out respectively；It may be the calculation resources for saving processor, in the multiple image equal intervals of video flowing
Several frame images are selected to carry out Face datection, for example, two frames of one frame of interval or interval etc., obtain multiple faces comprising face
Image, the position where facial image are specified using rectangle frame.
Illustratively, the step S201 may comprise steps of:
Receive the video flowing sent from image capture device；
In embodiments of the present invention, image capture device is required without limitation, existing acquisition equipment, which can be used as, is
The input of system.
Using it is multi-layer biaxially oriented stream neural network to it is described acquisition equipment provide the video flowing in each frame data into
Row Face datection obtains multiple facial images.
In embodiments of the present invention, the multi-layer biaxially oriented stream neural network includes forward stream neural network and reverse flow nerve
Network detects face when using multi-layer biaxially oriented stream neural network to each frame data Face datection in the video flowing
Number is related with face number present in each frame data in the video flowing, so the face of each frame detected
Number is not fixed.
Step S202 carries out face tracking to multiple described facial images respectively, obtains multiple face image sets, each
It include the face of the same person in the facial image of the face image set；
The face tracking can be using the Maximum overlap mode under Kalman filtering constraint, and can also use can obtain it
The mode of his same effect.
Illustratively, the step S202 may include:
If detecting same face in the multiple image of the video flowing, using the maximum under Kalman filtering constraint
Overlap mode tracks the face, obtains the face image set of the face.
In embodiments of the present invention, it is assumed that detection obtains multiple people specified by above-mentioned rectangle frame in each video frame
Face then tracks face and exactly judges which face is the face for belonging to the same person in successive video frames.Kalman filtering
Maximum overlap mode under constraint refers in continuous two video frames in time, all in first video frame to detect
Multiple facial images are all respectively with a rectangle frame come designated position, all multiple facial images detected in second video frame
Also will all exist in rectangle frame present in first video frame and second video frame respectively with a rectangle frame come designated position
Rectangle frame be overlapped face of the maximum facial image as the same person.The facial image of the same person is placed on same face figure
During image set closes, and the facial image of different people will be placed in different faces image collection.
Step S203 carries out the facial image in the face image set for each face image set
Face quality evaluation obtains the target facial image for meeting default face quality requirement；
Illustratively, the step S203 may comprise steps of:
Obtain true resolution, length, width and the face area of each picture frame；
The face area refers to the product of physical length and developed width that the rectangle frame of face is detected in picture frame.
In practical applications, face quality evaluation is carried out to the facial image in described image frame, selects true resolution to be greater than default
Resolution threshold, length are greater than pre-set length threshold, width is greater than predetermined width threshold value, and face area is greater than default face face
The picture frame of product threshold value is exported as target facial image, obtains the target facial image for meeting default face quality requirement.Figure
Face as in has the posture, such as positive face posture etc. for meeting face recognition algorithms requirement.
In embodiments of the present invention, it is based on tracking result, it can be from resolution ratio, length, width and face area etc.
The standard of the face quality evaluation is configured, the ginseng of the camera of the default resolution threshold and image collecting device
Number is related, and illustratively, resolution ratio of camera head is higher, and default resolution value will be higher, pre-set length threshold and predetermined width threshold
Value should be all satisfied the requirement of face recognition algorithms, and the default face area threshold need to account for video frame area 80% and its
More than.Based on tracking result, facial image of the available people in successive video frames, with the movement of people, these people
There may be the variations on scale for face image, such as become larger or become smaller, the variation in angle, such as face video camera, side pair
Video camera, back to video camera etc. situations such as.In this case, the face number for the same person in successive video frames
According to, not all data are adapted to do recognition of face, for example, too small side face, back face data, then can not well into
Row identification, assessment of this patent to face quality, if meeting true resolution is greater than default resolution threshold, length greater than default
Length threshold, width are greater than predetermined width threshold value, and face area is greater than the picture frame of default face area threshold, and in image
Face there are posture this five conditions for meeting face recognition algorithms requirement in the case where human face data be considered high-quality
Human face data is measured, cloud is transmitted back.
Step S204 sends the corresponding target facial image of each face image set to cloud, so as to the cloud into
Row recognition of face.
In embodiments of the present invention, marginal end sends the target facial image to cloud by network, and send
The target facial image is the target facial image of the default face quality requirement of satisfaction in face set, is set with Image Acquisition
It is compared for cloud is transmitted directly to, reduces the transmission of video streaming image data, alleviate transmission network heavy burden, so that
When cloud carries out recognition of face in less video streaming image, the utilization rate and validity of cloud data can be improved, to have
Effect solves the problems, such as data redundancy.
In another embodiment of the present invention, the multi-layer biaxially oriented stream neural network includes forward stream neural network and reversed
Flow neural network；
The forward stream neural network includes: multiple positive network layers, and the forward stream neural network is with the more of video flowing
Frame image is input, exports top characteristic pattern；
In each positive network layer respectively include: convolutional layer and pond layer；The frame figure that the convolutional layer will receive
Picture carries out feature extraction to the frame image using nonlinear activation, obtains the fisrt feature figure comprising feature vector, and to institute
It states pond layer and exports the fisrt feature figure；The pond layer receives the fisrt feature figure, and to the fisrt feature figure into
Row compression, can reduce the feature vector that convolutional layer exports the characteristic pattern, obtain second feature figure.
In embodiments of the present invention, there are multiple " convolution, nonlinear activation, ponds " between two neighboring positive network layer
Combination operation, as shown in Figure 3, the difference of multiple forward direction network layers are that the resolution ratio of characteristic pattern is different, will pass through less convolution,
The positive network layer that nonlinear activation, pond " combination operation obtains is known as shallow-layer characteristic pattern, will be by multiple such as the F1 in Fig. 3
The positive network layer that " convolution, nonlinear activation, pond " combination operation obtains is known as further feature figure, such as the F2 in Fig. 3.It is described
The purpose of forward stream neural network is to carry out repeatedly " convolution, nonlinear activation, pond " combination operation, until exporting top
Characteristic pattern.The number that this patent combines " convolution, non-linear activation, pond " is unlimited, to the form of convolution, non-linear activation
Form, the form in pond are unlimited, and all forms for meeting same nature can be used as the component part of this patent.
The reverse flow neural network is carried out based on the output of the top characteristic pattern and each positive network layer
Fusion treatment obtains multiple fusion feature figures；Face datection is carried out in the fusion feature figure, obtains multiple facial images.
The reverse flow neural network includes: at least one reversed cascade fused layer, the fused layer also with it is corresponding
The forward direction network layer connection；The fused layer carries out up-sampling treatment to the top characteristic pattern, obtains up-sampling feature
Figure；The output of fused layer described in the output of the positive network layer and the up-sampling characteristic pattern or upper level is melted
It closes, obtains fusion feature figure；Face datection is carried out in the fusion feature figure, obtains multiple facial images.
In embodiments of the present invention, top characteristic pattern is successively carried out serving reverse flow neural network in sample, such as Fig. 3
F3 and F5 or F2 and F6 are carried out plus are grasped by the way of linearly summing it up by the operation indicated in frame, the fusion of two-way flow data
Make, so that two-way flow data is effectively fused together.The characteristics of characteristic pattern can sum up is that they are having the same
Image resolution ratio.By bidirectional flow nerve net system, characteristic pattern more with ability to express can be obtained.Each fused layer is
Referring to, there is the forward stream characteristic pattern of equal resolution and adjacent further feature figure to pass through after the characteristic pattern adduction after up-sampling
Characteristic pattern, such as the F4 and F5 in Fig. 3, this patent carries out Face datection on the two characteristic patterns, to effectively detect
Face.The face number that this patent detects on each characteristic pattern is indefinite, the face as existing for actual video frame data
Number determines.
In another embodiment of the present invention, a kind of face identification device of cloud side fusion is also provided, marginal end is applied to
51, as shown in figure 4, the apparatus may include with lower module:
Face detection module 41, for carrying out people respectively to from the multiple image in the video flowing that image capture device is sent
Face detection, obtains multiple facial images；
Face tracking module 42 obtains multiple face figures for carrying out face tracking respectively to multiple described facial images
Image set closes, and includes the face of the same person in the facial image of each face image set；
Quality assessment modules 43, for being directed to each face image set, to the people in the face image set
Face image carries out face quality evaluation, obtains the target facial image for meeting default face quality requirement；
Sending module 44, for sending the corresponding target facial image of each face image set to cloud 52, with toilet
It states cloud 52 and carries out recognition of face.
In another embodiment of the present invention, a kind of face identification system of cloud side fusion is also provided, as shown in figure 5, institute
The system of stating includes: marginal end 41 described in cloud 52 and the embodiment of the present invention.
In embodiments of the present invention, in order to mitigate the burden that data are transmitted, the validity and utilization rate of data are improved, sufficiently
" cloud 52 " and " marginal end 51 " respective advantage are utilized, this patent comprehensively considers the optimal composition and property of system, proposes to utilize
The mode of cloud side fusion, improves the performance of total system.
Be conducive to the flexibility and scalability of lifting system, so that respectively from the angle of technology using cloud side integration technology
Full decoupled conjunction between item key technology, if all people's face the relevant technologies prepositionization is placed on marginal end 51, it will cause
The problem of calculating cost prohibitive, computing resource and storage resource to marginal end 51 have higher requirement.For example, if necessary
Recognition of face is put into front end, then needs to store all potential faces to be identified on front-end camera.
The human face data that cloud 52 mainly provides user stores, such as the resident identification card front provided by the police
Portrait.Efficient search mechanism is established, so-called efficient testing mechanism refers to using searching algorithms such as Hash, it is established that high speed
Index is captured obtained human face data based on marginal end 51, is quickly compared, find identical face.This patent is to above-mentioned
Search mechanism, for face alignment mechanism without clearly limiting, all examples for meeting above-mentioned function can be used as the component part of this system.
This patent is to above-mentioned search mechanism, and without clearly limiting, all examples for meeting above-mentioned function can be used as is face alignment mechanism
The component part of system.
System described in this patent does not limit to examples detailed above, and the example that can complete identical function can be used as this system
Component part.
In another embodiment of the present invention, a kind of non-volatile program code that can be performed with processor is also provided
Computer-readable medium, said program code make the processor execute embodiment of the method the method.
The flow chart and block diagram in the drawings show the system of multiple embodiments according to the present invention, method and computer journeys
The architecture, function and operation in the cards of sequence product.In this regard, each box in flowchart or block diagram can generation
A part of one module, section or code of table, a part of the module, section or code include one or more use
The executable instruction of the logic function as defined in realizing.It should also be noted that in some implementations as replacements, being marked in box
The function of note can also occur in a different order than that indicated in the drawings.For example, two continuous boxes can actually base
Originally it is performed in parallel, they can also be executed in the opposite order sometimes, and this depends on the function involved.It is also noted that
It is the combination of each box in block diagram and or flow chart and the box in block diagram and or flow chart, can uses and execute rule
The dedicated hardware based system of fixed function or movement is realized, or can use the group of specialized hardware and computer instruction
It closes to realize.
The computer program product of the face identification method of the fusion of cloud side provided by the embodiment of the present invention, including store
The computer readable storage medium of program code, the instruction that said program code includes can be used for executing in previous methods embodiment
The method, specific implementation can be found in embodiment of the method, and details are not described herein.
It is apparent to those skilled in the art that for convenience and simplicity of description, the system of foregoing description
It with the specific work process of device, can refer to corresponding processes in the foregoing method embodiment, details are not described herein.
In addition, in the description of the embodiment of the present invention unless specifically defined or limited otherwise, term " installation ", " phase
Even ", " connection " shall be understood in a broad sense, for example, it may be being fixedly connected, may be a detachable connection, or be integrally connected；It can
To be mechanical connection, it is also possible to be electrically connected；It can be directly connected, can also can be indirectly connected through an intermediary
Connection inside two elements.For the ordinary skill in the art, above-mentioned term can be understood at this with concrete condition
Concrete meaning in invention.
It, can be with if the function is realized in the form of SFU software functional unit and when sold or used as an independent product
It is stored in a computer readable storage medium.Based on this understanding, technical solution of the present invention is substantially in other words
The part of the part that contributes to existing technology or the technical solution can be embodied in the form of software products, the meter
Calculation machine software product is stored in a storage medium, including some instructions are used so that a computer equipment (can be a
People's computer, server or network equipment etc.) it performs all or part of the steps of the method described in the various embodiments of the present invention.
And storage medium above-mentioned includes: that USB flash disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), arbitrary access are deposited
The various media that can store program code such as reservoir (RAM, Random Access Memory), magnetic or disk.
In the description of the present invention, it should be noted that term " center ", "upper", "lower", "left", "right", "vertical",
The orientation or positional relationship of the instructions such as "horizontal", "inner", "outside" be based on the orientation or positional relationship shown in the drawings, merely to
Convenient for description the present invention and simplify description, rather than the device or element of indication or suggestion meaning must have a particular orientation,
It is constructed and operated in a specific orientation, therefore is not considered as limiting the invention.In addition, term " first ", " second ",
" third " is used for descriptive purposes only and cannot be understood as indicating or suggesting relative importance.
Finally, it should be noted that embodiment described above, only a specific embodiment of the invention, to illustrate the present invention
Technical solution, rather than its limitations, scope of protection of the present invention is not limited thereto, although with reference to the foregoing embodiments to this hair
It is bright to be described in detail, those skilled in the art should understand that: anyone skilled in the art
In the technical scope disclosed by the present invention, it can still modify to technical solution documented by previous embodiment or can be light
It is readily conceivable that variation or equivalent replacement of some of the technical features；And these modifications, variation or replacement, do not make
The essence of corresponding technical solution is detached from the spirit and scope of technical solution of the embodiment of the present invention, should all cover in protection of the invention
Within the scope of.Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. a kind of face identification method based on the fusion of cloud side, which is characterized in that be applied to marginal end, comprising:
Face datection is carried out respectively to from the multiple image in the video flowing that image capture device is sent, and obtains multiple face figures
Picture；
Face tracking is carried out to multiple described facial images respectively, obtains multiple face image sets, each facial image
It include the face of the same person in the facial image of set；
For each face image set, face quality evaluation is carried out to the facial image in the face image set,
Obtain the target facial image for meeting default face quality requirement；
The corresponding target facial image of each face image set is sent to cloud, so that the cloud carries out recognition of face.
2. the face identification method according to claim 1 based on the fusion of cloud side, which is characterized in that described to be adopted to from image
The multiple image in video flowing that collection equipment is sent carries out Face datection respectively, obtains multiple facial images, comprising:
Receive the video flowing sent from image capture device；
The each frame data in the video flowing provided using multi-layer biaxially oriented stream neural network the acquisition equipment carry out people
Face detection, obtains multiple facial images.
3. the face identification method according to claim 2 based on the fusion of cloud side, which is characterized in that the multi-layer biaxially oriented stream
Neural network includes forward stream neural network and reverse flow neural network；
The forward stream neural network includes: multiple positive network layers, and the forward stream neural network is with the multiframe figure of video flowing
As exporting top characteristic pattern to input；
The reverse flow neural network is merged based on the output of the top characteristic pattern and each positive network layer
Processing, obtains multiple fusion feature figures；Face datection is carried out in the fusion feature figure, obtains multiple facial images.
4. the face identification method according to claim 3 based on the fusion of cloud side, which is characterized in that the reverse flow nerve
Network includes: at least one reversed cascade fused layer, and the fused layer is also connected with the corresponding positive network layer；
The fused layer carries out up-sampling treatment to the top characteristic pattern, obtains up-sampling characteristic pattern；By the positive net
The output of fused layer described in the output of network layers and the up-sampling characteristic pattern or upper level is merged, and fusion feature is obtained
Figure；Face datection is carried out in the fusion feature figure, obtains multiple facial images.
5. the face identification method according to claim 3 based on the fusion of cloud side, which is characterized in that each positive net
In network layers respectively include: convolutional layer and pond layer；
The frame image that the convolutional layer will receive carries out feature extraction to the frame image using nonlinear activation, is wrapped
Fisrt feature figure containing feature vector, and the fisrt feature figure is exported to the pond layer；
The pond layer receives the fisrt feature figure, and compresses to the fisrt feature figure, obtains second feature figure.
6. it is according to claim 1 based on cloud side fusion face identification method, which is characterized in that it is described to it is described multiple
Facial image carries out face tracking respectively, obtains multiple face image sets, comprising:
If detecting same face in the multiple image of the video flowing, using the Maximum overlap under Kalman filtering constraint
Mode tracks the face, obtains the face image set of the face.
7. the face identification method according to claim 6 based on the fusion of cloud side, which is characterized in that described to be directed to each institute
Face image set is stated, face quality evaluation is carried out to the facial image in the face image set, obtains meeting default people
The target facial image of face quality requirement, comprising:
Obtain true resolution, length, width and the face area of each picture frame；
Face quality evaluation is carried out to the facial image in described image frame, true resolution is selected to be greater than default resolution ratio threshold
Value, length are greater than pre-set length threshold, width is greater than predetermined width threshold value, and face area is greater than default face area threshold
Picture frame is exported as target facial image, obtains the target facial image for meeting default face quality requirement.
8. a kind of face identification device based on the fusion of cloud side, which is characterized in that be applied to marginal end, comprising:
Face detection module, for carrying out face inspection respectively to from the multiple image in the video flowing that image capture device is sent
It surveys, obtains multiple facial images；
Face tracking module obtains multiple face image sets for carrying out face tracking respectively to multiple described facial images,
It include the face of the same person in the facial image of each face image set；
Quality assessment modules, for being directed to each face image set, to the facial image in the face image set
Face quality evaluation is carried out, the target facial image for meeting default face quality requirement is obtained；
Sending module, for sending the corresponding target facial image of each face image set to cloud, so as to the cloud into
Row recognition of face.
9. a kind of face identification system based on the fusion of cloud side, which is characterized in that the system comprises: cloud and such as claim
1 to 8 any marginal end；
The cloud, the target facial image sent for receiving the marginal end, and it is stored in pre-set image storage region；
When receiving face to be retrieved, people identical with the face to be retrieved is searched in the pre-set image storage region
Face image obtains face recognition result.
10. a kind of computer-readable medium for the non-volatile program code that can be performed with processor, which is characterized in that described
Program code makes the processor execute described any the method for claim 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810957725.8A CN109190532A (en) | 2018-08-21 | 2018-08-21 | It is a kind of based on cloud side fusion face identification method, apparatus and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810957725.8A CN109190532A (en) | 2018-08-21 | 2018-08-21 | It is a kind of based on cloud side fusion face identification method, apparatus and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109190532A true CN109190532A (en) | 2019-01-11 |
Family
ID=64918845
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810957725.8A Pending CN109190532A (en) | 2018-08-21 | 2018-08-21 | It is a kind of based on cloud side fusion face identification method, apparatus and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109190532A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109918951A (en) * | 2019-03-12 | 2019-06-21 | 中国科学院信息工程研究所 | A kind of artificial intelligence process device side channel system of defense based on interlayer fusion |
| CN111079670A (en) * | 2019-12-20 | 2020-04-28 | 北京百度网讯科技有限公司 | Face recognition method, face recognition device, face recognition terminal and face recognition medium |
| CN111400047A (en) * | 2020-03-18 | 2020-07-10 | 青岛牛利智能科技有限公司 | Method for detecting and identifying human face from monitoring video stream through cloud edge cooperation |
| CN111738193A (en) * | 2020-06-29 | 2020-10-02 | 湖南国科微电子股份有限公司 | Face snapshot method and face snapshot system |
| CN111753756A (en) * | 2020-06-28 | 2020-10-09 | 浙江大华技术股份有限公司 | Object identification-based deployment alarm method and device and storage medium |
| CN111931692A (en) * | 2020-08-31 | 2020-11-13 | 青岛聚看云科技有限公司 | Display device and image recognition method |
| CN112329665A (en) * | 2020-11-10 | 2021-02-05 | 上海大学 | Face snapshot system |
| CN112733685A (en) * | 2020-12-31 | 2021-04-30 | 北京澎思科技有限公司 | Face tracking method, system and computer readable storage medium |
| CN114093004A (en) * | 2021-11-25 | 2022-02-25 | 成都智元汇信息技术股份有限公司 | Face fusion comparison method and device based on multiple cameras |
| CN115761571A (en) * | 2022-10-26 | 2023-03-07 | 北京百度网讯科技有限公司 | Video-based target retrieval method, device, equipment and storage medium |
| CN116386119A (en) * | 2023-05-09 | 2023-07-04 | 北京维艾狄尔信息科技有限公司 | Body-building footpath-based identity recognition method, body-building footpath-based identity recognition system, body-building footpath-based identity recognition terminal and storage medium |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130167139A1 (en) * | 2011-12-21 | 2013-06-27 | Yahoo! Inc. | Method and system for distributed application stack test certification |
| CN103927520A (en) * | 2014-04-14 | 2014-07-16 | 中国华戎控股有限公司 | Method for detecting human face under backlighting environment |
| CN107341517A (en) * | 2017-07-07 | 2017-11-10 | 哈尔滨工业大学 | The multiple dimensioned wisp detection method of Fusion Features between a kind of level based on deep learning |
| CN107819854A (en) * | 2017-11-14 | 2018-03-20 | 深圳市华讯方舟软件信息有限公司 | Public safety system and construction method based on cloud edge calculations |
| CN107909174A (en) * | 2017-11-29 | 2018-04-13 | 南京甄视智能科技有限公司 | Session system based on face recognition technology, method of registering, map guidance method |
| CN108171207A (en) * | 2018-01-17 | 2018-06-15 | 百度在线网络技术（北京）有限公司 | Face identification method and device based on video sequence |
| CN108228872A (en) * | 2017-07-21 | 2018-06-29 | 北京市商汤科技开发有限公司 | Facial image De-weight method and device, electronic equipment, storage medium, program |
| CN108256459A (en) * | 2018-01-10 | 2018-07-06 | 北京博睿视科技有限责任公司 | Library algorithm is built in detector gate recognition of face and face based on multiple-camera fusion automatically |
-
2018
- 2018-08-21 CN CN201810957725.8A patent/CN109190532A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130167139A1 (en) * | 2011-12-21 | 2013-06-27 | Yahoo! Inc. | Method and system for distributed application stack test certification |
| CN103927520A (en) * | 2014-04-14 | 2014-07-16 | 中国华戎控股有限公司 | Method for detecting human face under backlighting environment |
| CN107341517A (en) * | 2017-07-07 | 2017-11-10 | 哈尔滨工业大学 | The multiple dimensioned wisp detection method of Fusion Features between a kind of level based on deep learning |
| CN108228872A (en) * | 2017-07-21 | 2018-06-29 | 北京市商汤科技开发有限公司 | Facial image De-weight method and device, electronic equipment, storage medium, program |
| CN107819854A (en) * | 2017-11-14 | 2018-03-20 | 深圳市华讯方舟软件信息有限公司 | Public safety system and construction method based on cloud edge calculations |
| CN107909174A (en) * | 2017-11-29 | 2018-04-13 | 南京甄视智能科技有限公司 | Session system based on face recognition technology, method of registering, map guidance method |
| CN108256459A (en) * | 2018-01-10 | 2018-07-06 | 北京博睿视科技有限责任公司 | Library algorithm is built in detector gate recognition of face and face based on multiple-camera fusion automatically |
| CN108171207A (en) * | 2018-01-17 | 2018-06-15 | 百度在线网络技术（北京）有限公司 | Face identification method and device based on video sequence |
Non-Patent Citations (1)
| Title |
|---|
| T.LIN等: "Feature Pyramid Networks for Object Detection", 《2017 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109918951A (en) * | 2019-03-12 | 2019-06-21 | 中国科学院信息工程研究所 | A kind of artificial intelligence process device side channel system of defense based on interlayer fusion |
| CN111079670B (en) * | 2019-12-20 | 2023-11-03 | 北京百度网讯科技有限公司 | Face recognition method, device, terminal and medium |
| CN111079670A (en) * | 2019-12-20 | 2020-04-28 | 北京百度网讯科技有限公司 | Face recognition method, face recognition device, face recognition terminal and face recognition medium |
| CN111400047A (en) * | 2020-03-18 | 2020-07-10 | 青岛牛利智能科技有限公司 | Method for detecting and identifying human face from monitoring video stream through cloud edge cooperation |
| CN111753756A (en) * | 2020-06-28 | 2020-10-09 | 浙江大华技术股份有限公司 | Object identification-based deployment alarm method and device and storage medium |
| CN111738193A (en) * | 2020-06-29 | 2020-10-02 | 湖南国科微电子股份有限公司 | Face snapshot method and face snapshot system |
| CN111931692A (en) * | 2020-08-31 | 2020-11-13 | 青岛聚看云科技有限公司 | Display device and image recognition method |
| CN112329665A (en) * | 2020-11-10 | 2021-02-05 | 上海大学 | Face snapshot system |
| CN112329665B (en) * | 2020-11-10 | 2022-05-17 | 上海大学 | Face snapshot system |
| CN112733685A (en) * | 2020-12-31 | 2021-04-30 | 北京澎思科技有限公司 | Face tracking method, system and computer readable storage medium |
| CN114093004A (en) * | 2021-11-25 | 2022-02-25 | 成都智元汇信息技术股份有限公司 | Face fusion comparison method and device based on multiple cameras |
| CN115761571A (en) * | 2022-10-26 | 2023-03-07 | 北京百度网讯科技有限公司 | Video-based target retrieval method, device, equipment and storage medium |
| CN116386119A (en) * | 2023-05-09 | 2023-07-04 | 北京维艾狄尔信息科技有限公司 | Body-building footpath-based identity recognition method, body-building footpath-based identity recognition system, body-building footpath-based identity recognition terminal and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109190532A (en) | It is a kind of based on cloud side fusion face identification method, apparatus and system | |
| CN108460362B (en) | System and method for detecting human body part | |
| CN110807385B (en) | Target detection method, target detection device, electronic equipment and storage medium | |
| CN109697434B (en) | Behavior recognition method and device and storage medium | |
| CN104217202B (en) | Information identifying method, equipment and system | |
| Loy et al. | An adaptive fusion architecture for target tracking | |
| CN102096805B (en) | Apparatus and method for registering plurality of facial images for face recognition | |
| CN107578453A (en) | Compressed image processing method, apparatus, electronic equipment and computer-readable medium | |
| CN104408428B (en) | A kind of processing method and processing device of same scene photo | |
| CN107944382B (en) | Method for tracking target, device and electronic equipment | |
| EP3080779A1 (en) | Efficient facial landmark tracking using online shape regression method | |
| CN110163041A (en) | Video pedestrian recognition methods, device and storage medium again | |
| CN110399788A (en) | AU detection method, device, electronic equipment and the storage medium of image | |
| CN110264495A (en) | A kind of method for tracking target and device | |
| CN108664849A (en) | The detection device of event, method and image processing equipment in video | |
| CN108921131B (en) | Method and device for generating face detection model and three-dimensional face image | |
| CN109783680B (en) | Image pushing method, image acquisition device and image processing system | |
| CN109447006A (en) | Image processing method, device, equipment and storage medium | |
| CN110263699A (en) | Method of video image processing, device, equipment and storage medium | |
| CN111654694B (en) | Quality evaluation method and device of image processing algorithm and electronic equipment | |
| CN111626163A (en) | Human face living body detection method and device and computer equipment | |
| CN104769611A (en) | Method and apparatus for summarization based on facial expressions | |
| CN109040587A (en) | It captures processing method, device, capture mechanism, equipment and storage medium | |
| Wang et al. | Paul: Procrustean autoencoder for unsupervised lifting | |
| CN109359689B (en) | Data identification method and device |
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: 20190111 |