CN108960114A - Human body recognition method and device, computer readable storage medium and electronic equipment - Google Patents
Human body recognition method and device, computer readable storage medium and electronic equipment Download PDFInfo
- Publication number
- CN108960114A CN108960114A CN201810681840.7A CN201810681840A CN108960114A CN 108960114 A CN108960114 A CN 108960114A CN 201810681840 A CN201810681840 A CN 201810681840A CN 108960114 A CN108960114 A CN 108960114A
- Authority
- CN
- China
- Prior art keywords
- pedestrian
- image
- human body
- pond
- feature
- 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
- 238000000034 method Methods 0.000 title claims abstract description 66
- 239000000284 extract Substances 0.000 claims description 13
- 238000000605 extraction Methods 0.000 claims description 12
- 238000012360 testing method Methods 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 8
- 238000004590 computer program Methods 0.000 claims description 6
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 238000013135 deep learning Methods 0.000 description 13
- 238000010586 diagram Methods 0.000 description 13
- 238000004364 calculation method Methods 0.000 description 11
- 230000006870 function Effects 0.000 description 10
- 238000005516 engineering process Methods 0.000 description 9
- 238000013527 convolutional neural network Methods 0.000 description 7
- 230000008859 change Effects 0.000 description 6
- 230000006854 communication Effects 0.000 description 6
- 238000012549 training Methods 0.000 description 6
- 238000004891 communication Methods 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 4
- 238000001514 detection method Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000005021 gait Effects 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000007796 conventional method Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000005291 magnetic effect Effects 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 239000004065 semiconductor Substances 0.000 description 2
- 238000013316 zoning Methods 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 239000000470 constituent Substances 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000003014 reinforcing effect Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
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
- G06V40/20—Movements or behaviour, e.g. gesture recognition
- G06V40/23—Recognition of whole body movements, e.g. for sport training
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Human Computer Interaction (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to network technique field, a kind of human body recognition method and device, computer readable storage medium, electronic equipment are provided.Human body recognition methods includes: acquisition pedestrian image；The feature of the pedestrian image is extracted, to obtain global characteristics figure；The global characteristics figure is divided into multiple regions along preset direction, and pond is carried out to each region to obtain pond region corresponding with each region；Then the corresponding pond provincial characteristics in each pond region is sequentially connected along the preset direction, to obtain the corresponding pedestrian image feature of the pedestrian image.The present invention can be improved image recognition precision, reduce the probability of wrong identification misrecognition；In addition the human body recognition method in the present invention adapts to the weight identification mission of the pedestrian under more scenes.
Description
Technical field
The present invention relates to field of computer technology, in particular to a kind of human body recognition method, human bioequivalence device,
Computer readable storage medium and electronic equipment.
Background technique
With the fast development of science and technology, video monitoring Intellectual Analysis Technology is widely used in every field, such as
The license plate number for analyzing vehicle from a large amount of monitor video by video monitoring Intellectual Analysis Technology, the face for analyzing pedestrian are special
The biological characteristics such as sign, gait.
Pedestrian identifies that (Person Re-Identification, abbreviation ReID) is to across camera, cross-scenario prison again
The pedestrian occurred in control video is associated the technology of identification, that is, identifies whether two human body images are the same person, or one
The reference picture outside a library is given in a larger image library, and one and reference picture are then found in library according to reference picture
Other images of common identity.Recognition methods mostly uses greatly the manual feature of field of image processing to carry out model to traditional pedestrian again
Training, such as SIFT, HOG, and identified again in conjunction with classical machine learning regression model to carry out pedestrian, but traditional knowledge again
Other precision is lower, and the probability of mistake identification misrecognition is larger, and in addition usage scenario is single, and the identification again that can not be adapted under more scenes is appointed
Business.
Therefore this field needs to seek the new human body recognition method of one kind to improve the precision that pedestrian identifies again.
It should be noted that information is only used for reinforcing the reason to background of the invention disclosed in above-mentioned background technology part
Solution, therefore may include the information not constituted to the prior art known to persons of ordinary skill in the art.
Summary of the invention
The purpose of the present invention is to provide a kind of human body recognition methods and device, computer readable storage medium, electronics to set
It is standby, and then accuracy of identification is improved, wrong identification probability of misrecognition is reduced, while improving and carrying out the energy that pedestrian identifies again under more scenes
Power.
Other characteristics and advantages of the invention will be apparent from by the following detailed description, or partially by the present invention
Practice and acquistion.
According to the first aspect of the invention, a kind of human body recognition method is provided characterized by comprising obtain pedestrian's figure
Picture；The feature of the pedestrian image is extracted, to obtain global characteristics figure；The global characteristics figure is divided into along preset direction multiple
Region, and pond is carried out to each region to obtain pond region corresponding with each region；By each pond region
Corresponding pond provincial characteristics is sequentially connected along the preset direction, special to obtain the corresponding pedestrian image of the pedestrian image
Sign.
According to the second aspect of the invention, a kind of human bioequivalence device is provided characterized by comprising first obtains mould
Block, for obtaining pedestrian image；Characteristic extracting module, for extracting the feature of the pedestrian image, to obtain global characteristics figure；
Substep pond module for the global characteristics figure to be divided into multiple regions along preset direction, and carries out pond to each region
Change to obtain pond region corresponding with each region；Full link block, for by the corresponding pond in each pond region
Provincial characteristics is sequentially connected along the preset direction, to obtain the corresponding pedestrian image feature of the pedestrian image.
In some embodiments of the invention, aforementioned schemes, the device of the invention are based on further include: second obtains module,
For obtaining the original image comprising pedestrian；Human detection module, for by the original image carry out human testing, with
Obtain pedestrian image frame；Image generation module, for extracting the corresponding region of pedestrian from the pedestrian image frame, to be formed
State pedestrian image.
In some embodiments of the invention, aforementioned schemes are based on, substep pond module of the invention includes: subregion list
Member, for the global characteristics figure to be averaged piecemeal along y direction, to form multiple regions；Pond unit, for dividing
It is other that average pond is carried out to each region, to obtain multiple pond regions.
In some embodiments of the invention, aforementioned schemes, the device of the invention further include: the feature extraction mould are based on
Block, substep pond module and the full link block constitute pedestrian's identification model.
In some embodiments of the invention, aforementioned schemes, the device of the invention further include: pond feature extraction mould are based on
Block extracts the feature in each pond region, for the convolution kernel by presetting size to obtain the pond provincial characteristics.
In some embodiments of the invention, the pedestrian image includes pedestrian image and target pedestrian image to be checked, base
In aforementioned schemes, the device of the invention further include: similarity calculation module, for according to the pedestrian image to be checked it is corresponding to
Pedestrian image feature and the corresponding target pedestrian characteristics of image of the target pedestrian image are examined, it is special to calculate the target pedestrian image
The similarity of sign and the pedestrian image feature to be checked；Judgment module, for judging the pedestrian to be checked according to the similarity
Whether pedestrian in image is pedestrian in the target pedestrian image.
In some embodiments of the invention, aforementioned schemes are based on, similarity calculation module of the invention includes: distance meter
Unit is calculated, for calculating the target pedestrian characteristics of image at a distance from the pedestrian image feature to be checked；Similarity calculation list
Member, at a distance from the pedestrian image feature to be checked, determining the target pedestrian according to the target pedestrian characteristics of image
The similarity of characteristics of image and the pedestrian image feature to be checked.
In some embodiments of the invention, aforementioned schemes, the device of the invention are based on further include: arrangement module is used for
According to size of the target pedestrian characteristics of image at a distance from the pedestrian image feature to be checked, from the near to the remote arrangement it is described to
Examine pedestrian image.
In some embodiments of the invention, aforementioned schemes, the device of the invention are based on further include: third obtains module,
For obtaining the original image comprising pedestrian acquired by image capture device；Pedestrian image generation module to be checked, for institute
It states original image to carry out human testing and extract human region, to obtain the pedestrian image to be checked.
In some embodiments of the invention, aforementioned schemes, the device of the invention are based on further include: target pedestrian image is raw
At module, for receiving the human body image of user's selection, and using the human body image as the target pedestrian image.
In some embodiments of the invention, aforementioned schemes, the device of the invention are based on further include: the 4th obtains module,
For obtaining multiple human body mark images, each human body mark image corresponds to different scenes；Model training module is used for institute
It states human body mark image and is input to pedestrian's identification model, to be trained to pedestrian's identification model.
According to the third aspect of the invention we, a kind of computer-readable medium is provided, computer program is stored thereon with, institute
It states and realizes such as above-mentioned human body recognition method as described in the examples when program is executed by processor.
According to the fourth aspect of the invention, a kind of electronic equipment is provided, comprising: one or more processors；Storage dress
It sets, for storing one or more programs, when one or more of programs are executed by one or more of processors, makes
It obtains one or more of processors and realizes such as above-mentioned human body recognition methods as described in the examples.
According to the human body recognition method in this example embodiment, after server obtains pedestrian image, first to pedestrian image
Feature extraction is carried out to obtain global characteristics figure；Then global characteristics figure is divided into multiple regions, the area Bing Duige along preset direction
Domain pond forms pond corresponding with each region region；Finally by the corresponding pond provincial characteristics in each pond region according to default side
To connection, to obtain pedestrian image feature corresponding with pedestrian image.The present invention is by being divided into multiple regions for global characteristics figure
To carry out substep pond, then the pond provincial characteristics in each pond region is sequentially connected and obtains the corresponding feature of pedestrian image,
The accuracy of identification for improving pedestrian image reduces the probability of wrong identification misrecognition.
The present invention is it should be understood that above general description and following detailed description is only exemplary and explanatory
, the present invention can not be limited.
Detailed description of the invention
The drawings herein are incorporated into the specification and forms part of this specification, and shows and meets implementation of the invention
Example, and be used to explain the principle of the present invention together with specification.It should be evident that the accompanying drawings in the following description is only the present invention
Some embodiments for those of ordinary skill in the art without creative efforts, can also basis
These attached drawings obtain other attached drawings.
Fig. 1 is shown can be using the embodiment of the present invention for human body recognition method or the example for human bioequivalence device
The schematic diagram of property system architecture；
Fig. 2 shows the structural schematic diagrams of the computer system of the electronic equipment suitable for being used to realize the embodiment of the present invention；
Fig. 3 shows the human body recognition method flow chart in one embodiment of the invention；
Fig. 4 A-4C shows the global characteristics figure in one embodiment of the invention and shows along the structure that preset direction is divided into multiple regions
It is intended to；
Fig. 5 shows in one embodiment of the invention dense piece in dense convolutional neural networks of structural schematic diagram；
Fig. 6 shows the flow chart of human body recognition method in one embodiment of the invention；
Fig. 7 shows the flow chart that pedestrian in one embodiment of the invention identifies again；
Fig. 8 shows the structural schematic diagram of human bioequivalence device in one embodiment of the invention.
Specific embodiment
Example embodiment is described more fully with reference to the drawings.However, example embodiment can be with a variety of shapes
Formula is implemented, and is not understood as limited to example set forth herein；On the contrary, thesing embodiments are provided so that the present invention will more
Fully and completely, and by the design of example embodiment comprehensively it is communicated to those skilled in the art.
In addition, described feature, structure or characteristic can be incorporated in one or more implementations in any suitable manner
In example.In the following description, many details are provided to provide and fully understand to the embodiment of the present invention.However,
It will be appreciated by persons skilled in the art that technical solution of the present invention can be practiced without one or more in specific detail,
Or it can be using other methods, constituent element, device, step etc..In other cases, it is not shown in detail or describes known side
Method, device, realization or operation are to avoid fuzzy each aspect of the present invention.
Block diagram shown in the drawings is only functional entity, not necessarily must be corresponding with physically separate entity.
I.e., it is possible to realize these functional entitys using software form, or realized in one or more hardware modules or integrated circuit
These functional entitys, or these functional entitys are realized in heterogeneous networks and/or processor device and/or microcontroller device.
Flow chart shown in the drawings is merely illustrative, it is not necessary to including all content and operation/step,
It is not required to execute by described sequence.For example, some operation/steps can also decompose, and some operation/steps can close
And or part merge, therefore the sequence actually executed is possible to change according to the actual situation.
Fig. 1 is shown can be using the human body recognition method of the embodiment of the present invention or the exemplary system of human bioequivalence device
The schematic diagram of framework 100.
As shown in Figure 1, system architecture 100 may include terminal device 101, network 102 and server 103.Network 102 is used
To provide the medium of communication link between terminal device 101 and server 103.Network 102 may include various connection types,
Such as wired, wireless communication link or fiber optic cables etc..
It should be understood that the number of terminal device, network and server in Fig. 1 is only schematical.According to realization need
It wants, can have any number of terminal device, network and server.For example server 103 can be multiple server compositions
Server cluster etc..
User can be used terminal device 101 and be interacted by network 102 with server 103, to receive or send message etc..
Terminal device 101 can be the various electronic equipments with display screen, including but not limited to smart phone, tablet computer, portable
Formula computer and desktop computer etc..
Server 103 can be to provide the server of various services.Such as handling pedestrian image, user
By the input equipment that is connect with terminal device 101 in terminal device 101 selection target pedestrian image, and target pedestrian is schemed
As being sent to server 103, server 103 is to target pedestrian's image zooming-out feature, substep pondization and connects each area Chi Hua entirely
Characteristic of field forms target pedestrian characteristics of image；Server 103 is obtained by multiple image capture devices positioned at different location simultaneously
Multiple original images comprising pedestrian are taken, form pedestrian image frame by carrying out human testing respectively to original image, and from row
Human body parts are extracted in people's frames images forms pedestrian image to be measured；Then feature, substep pondization are extracted simultaneously to pedestrian image to be measured
Full connection forms pedestrian image feature to be measured, and multiple pedestrian image features to be measured and corresponding human body parts are stored in server
In 103, image library is formed.After server 103 obtains target pedestrian characteristics of image, by calculating target pedestrian characteristics of image and spy
The distance of each pedestrian image feature to be measured in library is levied, to judge the similarity of the two, and then according to similarity determination and target line
The immediate pedestrian image feature to be measured of people's characteristics of image, and the corresponding pedestrian image to be measured of the pedestrian image feature to be measured is sent out
Terminal device 101 is sent to for reference.
Fig. 2 shows the structural representations of the computer system of the electronic equipment suitable for being used to realize the embodiment in the present invention
Figure.
It should be noted that Fig. 2 shows the computer system 200 of electronic equipment be only an example, should not be to this hair
The function and use scope of bright embodiment bring any restrictions.
As shown in Fig. 2, computer system 200 includes central processing unit (CPU) 201, it can be read-only according to being stored in
Program in memory (ROM) 202 or be loaded into the program in random access storage device (RAM) 203 from storage section 208 and
Execute various movements appropriate and processing.In RAM 203, it is also stored with various programs and data needed for system operatio.CPU
201, ROM 202 and RAM 203 is connected with each other by bus 204.Input/output (I/O) interface 205 is also connected to bus
204。
I/O interface 205 is connected to lower component: the importation 206 including keyboard, mouse etc.；It is penetrated including such as cathode
The output par, c 207 of spool (CRT), liquid crystal display (LCD) etc. and loudspeaker etc.；Storage section 208 including hard disk etc.；
And the communications portion 209 of the network interface card including LAN card, modem etc..Communications portion 209 via such as because
The network of spy's net executes communication process.Driver 210 is also connected to I/O interface 205 as needed.Detachable media 211, such as
Disk, CD, magneto-optic disk, semiconductor memory etc. are mounted on as needed on driver 210, in order to read from thereon
Computer program be mounted into storage section 208 as needed.
Particularly, according to an embodiment of the invention, may be implemented as computer below with reference to the process of flow chart description
Software program.For example, the embodiment of the present invention includes a kind of computer program product comprising be carried on computer-readable medium
On computer program, which includes the program code for method shown in execution flow chart.In such reality
It applies in example, which can be downloaded and installed from network by communications portion 209, and/or from detachable media
211 are mounted.When the computer program is executed by central processing unit (CPU) 201, executes and limited in the system of the application
Various functions.
It should be noted that computer-readable medium shown in the present invention can be computer-readable signal media or meter
Calculation machine readable storage medium storing program for executing either the two any combination.Computer readable storage medium for example can be --- but not
Be limited to --- electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor system, device or device, or any above combination.Meter
The more specific example of calculation machine readable storage medium storing program for executing can include but is not limited to: have the electrical connection, just of one or more conducting wires
Taking formula computer disk, hard disk, random access storage device (RAM), read-only memory (ROM), erasable type may be programmed read-only storage
Device (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), light storage device, magnetic memory device,
Or above-mentioned any appropriate combination.In the present invention, computer readable storage medium can be it is any include or storage journey
The tangible medium of sequence, the program can be commanded execution system, device or device use or in connection.And at this
In invention, computer-readable signal media may include in a base band or as carrier wave a part propagate data-signal,
Wherein carry computer-readable program code.The data-signal of this propagation can take various forms, including but unlimited
In electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be that computer can
Any computer-readable medium other than storage medium is read, which can send, propagates or transmit and be used for
By the use of instruction execution system, device or device or program in connection.Include on computer-readable medium
Program code can transmit with any suitable medium, including but not limited to: wireless, electric wire, optical cable, RF etc. are above-mentioned
Any appropriate combination.
Flow chart and block diagram in attached drawing are illustrated according to the system of various embodiments of the invention, method and computer journey
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, program segment or code of table, a part of above-mentioned module, program segment or code include one or more
Executable instruction for implementing the specified logical function.It should also be noted that in some implementations as replacements, institute in box
The function of mark can also occur in a different order than that indicated in the drawings.For example, two boxes succeedingly indicated are practical
On can be basically executed in parallel, they can also be executed in the opposite order sometimes, and this depends on the function involved.Also it wants
It is noted that the combination of each box in block diagram or flow chart and the box in block diagram or flow chart, can use and execute rule
The dedicated hardware based systems of fixed functions or operations is realized, or can use the group of specialized hardware and computer instruction
It closes to realize.
Being described in unit involved in the embodiment of the present invention can be realized by way of software, can also be by hard
The mode of part realizes that described unit also can be set in the processor.Wherein, the title of these units is in certain situation
Under do not constitute restriction to the unit itself.
As on the other hand, present invention also provides a kind of computer-readable medium, which be can be
Included in electronic equipment described in above-described embodiment；It is also possible to individualism, and without in the supplying electronic equipment.
Above-mentioned computer-readable medium carries one or more program, when the electronics is set by one for said one or multiple programs
When standby execution, so that method described in electronic equipment realization as the following examples.For example, the electronic equipment can be real
Now each step as shown in figures 3 to 6.
This field in the related technology, the manual feature for generalling use field of image processing carries out model training, and ties
Classical machine learning regression model (such as linear regression) is closed to identify again to carry out pedestrian.Gradually with deep learning method
Universal, recognition methods starts gradually to improve to deep learning frame existing pedestrian again.But whether conventional method is also
It is deep learning model, the technical solution that pedestrian identifies again is substantially all to input one first to refer to human body image and a target
Human body image, the key feature for then extracting two images judge whether two images belong to the same person.
But pedestrian in the related technology is identified again there is also defect, is identified again by the above method, on the one hand, is known
Other precision is lower, and the probability of mistake identification misrecognition is larger；On the other hand, scene is single, can not adapt to the pedestrian under more scenes
Weight identification mission.
For problem present in the relevant technologies, in an embodiment of the present invention, a kind of human bioequivalence side is provided firstly
Method, with to there are the problem of optimize processing, with specific reference to shown in Fig. 3, human body recognition methods is suitable for previous embodiment
In the electronic equipment, and at least include the following steps, specifically:
Step S310: pedestrian image is obtained；
Step S320: extracting the feature of the pedestrian image, to obtain global characteristics figure；
Step S330: the global characteristics figure is divided into multiple regions along preset direction, and pond is carried out to each region
Change to obtain pond region corresponding with each region；
Step S340: the corresponding pond provincial characteristics in each pond region is sequentially connected along the preset direction, with
Obtain the corresponding pedestrian image feature of the pedestrian image.
According to the human body recognition method in this example embodiment, after server 103 obtains pedestrian image, to pedestrian image into
Row feature extraction is to obtain global characteristics figure；Then global characteristics figure is divided into multiple regions and substep pond, each region warp
Pond region is formed behind pond；The corresponding pond provincial characteristics in each pond region is finally in turn connected to form pedestrian image spy
Sign.By to global characteristics figure substep pond and connecting and to form pedestrian image feature and can be improved the precision that pedestrian identifies again, drop
The probability of low wrong identification misrecognition.
In the following, the human body recognition method in this example embodiment is further detailed.
In step s310, pedestrian image is obtained.
In this exemplary embodiment, the available original image comprising pedestrian acquired by image capture device；Then
Human testing is carried out to the pedestrian in original image, obtains the pedestrian image frame comprising pedestrian and less background；Then from pedestrian
The corresponding region of pedestrian's human body is extracted in frames images, to form the pedestrian image for only including pedestrian.In the present invention, the original of acquisition
Beginning image may include a pedestrian, also may include multiple pedestrians.It, can be in original image after obtaining original image
Each pedestrian is detected to obtain the human body rectangle frame for including pedestrian, then to the pedestrian in the human body rectangle frame comprising pedestrian
Human body edge detection is carried out, to obtain the pedestrian image for only including pedestrian, by human body and background separation.Image in the present invention is adopted
Collection equipment can be the equipment that video camera, camera, camera etc. have camera function.Moreover, it is noted that image is adopted
Different location can be located at by collecting equipment, for including the original image of pedestrian from different location, angle acquisition, in order to from pedestrian
All angles, posture, gait etc. are analyzed, and the accuracy of identification of pedestrian is improved.
In step s 320, the feature of the pedestrian image is extracted, to obtain a global characteristics figure.
In this exemplary embodiment, the feature of pedestrian image can be extracted, global characteristics figure is formed.Extracting pedestrian image
Feature when, can extract respectively pedestrian head, upper body, the lower part of the body, foot feature, finally the characteristics of image of each section is connected
Form the feature of pedestrian image；Human body can also be equally divided into muti-piece along y direction, extract the corresponding row of each piecemeal respectively
The corresponding characteristics of image of each piecemeal is finally connected the feature to form pedestrian image by the feature of everybody body；Certainly it can also use
Other way extracts the feature of pedestrian image, and the disclosure is not specifically limited in this embodiment.It, can be with when extracting the feature of pedestrian image
Using histograms of oriented gradients (Histogram of Oriented Gradient, abbreviation HOG), Scale invariant features transform
(Scale-invariant Feature Transform, abbreviation SIFT) scheduling algorithm, be also possible to by deep learning network come
Extract pedestrian image feature, wherein deep learning network can be convolutional neural networks (Convolutional Neural
Network, abbreviation CNN), Recognition with Recurrent Neural Network (Recurrent Neural Network, abbreviation RNN), residual error network
Network models such as (Residual Networks, abbreviation ResNet).When using histograms of oriented gradients algorithm extraction pedestrian image
Feature when, pedestrian image is carried out gray processing, calculates each pixel in pedestrian image after normalization first gradient；Then will
Pedestrian image is divided into small cell factory, and counts the histogram of gradients of each cell factory, finally these set of histograms
It can be formed by the profiler of pedestrian image altogether.When using Scale invariant features transform algorithm extraction pedestrian image
When feature, building scale space, detection extreme point obtain scale invariability first；Then characteristic point is filtered and is carried out
It is accurately positioned；Then it is characterized a distribution direction value, and generates Feature Descriptor；Finally to the SIFT feature of two images to
Amount is measured using the Euclidean distance of key point feature vector as the similarity determination of key point in two images.Work as use
When deep learning network model extracts the feature of pedestrian image, pedestrian image is inputted into deep learning network model, by multiple
Convolutional layer, pond layer carry out feature extraction to pedestrian image to obtain multiple images feature, and pass through full articulamentum for multiple figures
As feature is connected to obtain the feature of pedestrian image.Since deep learning network identifies essence compared to conventional pedestrian again recognition methods
Degree is higher, therefore the present invention carries out pedestrian using deep learning network and identifies again.
In this exemplary embodiment, the feature of pedestrian image can be extracted, by a feature extraction network to obtain and row
The corresponding global characteristics figure of people's image.Different size of convolution kernel can be set in feature extraction network, to pedestrian image into
Row multilayer convolution operation, to form global characteristics figure.
In step S330, the global characteristics figure is divided into multiple regions along preset direction, and to each region into
Row pond is to obtain multiple pond regions.
It in this exemplary embodiment, can be default by characteristic pattern edge by substep pond network after obtaining global characteristics figure
Direction is divided into multiple regions, which can be X direction or y direction, is also possible to other arbitrarily set
Direction, Fig. 4 A-4C show the structural schematic diagram that global characteristics figure is divided into multiple regions along preset direction, such as Fig. 4 A-4C institute
Show, at subregion, multiple regions can be divided into according to human body composed structure, for example, according to the head of human body, upper body, under
Body, foot carry out piecemeal (as shown in Figure 4 A)；Global characteristics figure can be averaged piecemeal along horizontal axis, form multiple regions (as schemed
Shown in 4B)；Global characteristics figure can be averaged piecemeal along the longitudinal axis, be formed multiple regions (as shown in Figure 4 C)；It can also will be global
Characteristic pattern is divided according to arbitrary proportion along any direction.It is contemplated that the stature ratio of each pedestrian is different, according to people
The effect is unsatisfactory for body composed structure piecemeal, and global characteristics figure is averaged piecemeal, then facilitates at deep learning network
Reason, and treatment effect is fine.Since operation object is the global characteristics figure that pedestrian's human body is formed, feature is mainly along longitudinal axis side
Global characteristics figure is divided into multiple regions along y direction to distribution, therefore in the present invention.
In this exemplary embodiment, after global characteristics figure being divided into multiple regions along preset direction, pass through substep pond net
Each region is carried out pond by network, to obtain multiple pond regions.
In step S340, the corresponding pond provincial characteristics in each pond region is successively connected along the preset direction
It connects, to obtain the corresponding pedestrian image feature of the pedestrian image.
In this example embodiment, global characteristics figure is divided into multiple regions, and each region forms pond by pond
Change region after, can by fully-connected network by the corresponding pond provincial characteristics in each pond region according to pre- in step S330
Set direction (y direction) is sequentially connected, and to obtain the corresponding pedestrian image feature of pedestrian image, which is used for
Prediction to pedestrian's identity.
In this example embodiment, in order to keep the pedestrian image feature of full articulamentum output richer, it can use
By before the corresponding pond provincial characteristics cascade in each pond region, the convolution kernel by presetting size extracts each fully-connected network
The feature in pond region improves the dimension of pond provincial characteristics, and then improves the dimension of pedestrian image feature, that is, improves pedestrian
The precision of characteristics of image.
Human body recognition method of the invention is by carrying out substep pond after extracting feature to pedestrian image, then again by each pond
Change the corresponding pond provincial characteristics in region to cascade to form pedestrian image feature.Human body recognition method through the invention can be by people
Body is divided into multiple regions, extracts the feature in each region respectively, to solve difference pedestrian identifies again when with human body attitude
Feature representation caused by changing distinguishes the larger problem of difficulty, improves the recognition accuracy that pedestrian identifies again.
In this example embodiment, this feature extract network, substep pond network and fully-connected network can group embark on journey
People's identification model.In order to improve the recognition accuracy that pedestrian identifies again, the present invention is based on dense convolutional neural networks
(DenseNet201) a group traveling together's identification model is established, and the last layer global pool layer of model is changed to substep pondization grade again
The form of connection is identified again with carrying out pedestrian.Fig. 5 shows in a kind of dense convolutional neural networks dense piece (Dense Block)
Structural schematic diagram, as shown in figure 5, dense piece 500 include 5 layers of (X0、X1、X2、X3、X4), wherein first layer X0For input layer, often
The output feature of layer passes through normalization-line rectification-convolution (BN-ReLU-Conv) operation (H1、H2、H3、H4) form a spy
Sign figure realizes the jump of characteristic pattern after the characteristic pattern of every layer of output is added to respectively in the corresponding output characteristic pattern of network layer
Jump is linked and is shared.Richer feature representation can be obtained by being multiplexed and sharing network middle layer feature weight, while right
The ultimate attainment utilization of middle layer feature can reach better effect and less parameter, that is to say, that dense convolutional neural networks tool
Have the advantages that feature representation abundant, parameter is few, calculation amount is few.Certainly, the present invention can also be using other deep learning nets
Network frame, the present invention are not specifically limited in this embodiment.
In this example embodiment, Fig. 6 shows the flow chart of human body recognition method in the present invention, as shown in fig. 6, clothes
Business device 103 receives one W × H original image, and it is global to carry out convolution algorithm output one to original image by deep learning network
Characteristic pattern；The global characteristics figure is divided into multiple regions along the y direction stripping and slicing that is averaged, and with a column vector f to each region into
The average pond of row, to obtain multiple pond regions；Then each pond region can be mentioned by the convolution kernel that size is 1 × 1
Feature is taken, so that characteristics of image is richer；Finally by full articulamentum by the corresponding pond provincial characteristics grade in each pond region
Connection forms final pedestrian image feature.
In this example embodiment, in order to have the deep learning network model in the present invention all to different scenes
There is good image recognition accuracy rate, then can will mark image conduct by the way that the pedestrian under different scenes to be labeled
Sample is trained deep learning network model.In the present invention, the forming method of image is marked specifically: first by not
Image capture device with position acquires image formation pedestrian's video sequence that several include multiple pedestrians；Then to to be marked
Pedestrian's video sequence carry out human testing and human body tracking, obtain the consecutive image sequence of same human body；Then by multiple people
Body image sequence carries out integration storage, and carries out identity validation to each sequence one by one by mark personnel, while in image sequence
Pedestrian's feature is selected in column, and clearly pedestrian image is labeled.Wherein pedestrian's feature clearly refers to non-overlapping, feature (face spy
Sign, gait, figure etc.) clearly pedestrian image.By using subtly pedestrian image labeled data training pattern, can be improved
Model identifies the accuracy rate of the pedestrian under different scenes, so that model has better generalization.
In this example embodiment, in order to be identified to pedestrian again, pedestrian image can be divided into pedestrian's figure to be checked
Picture and target pedestrian image, wherein pedestrian image to be checked can be by multiple pedestrian image feature composition characteristic libraries, different rows
People's characteristics of image corresponds to different pedestrian images.By the way that the corresponding feature of target pedestrian image is corresponding with pedestrian image to be checked
Feature compares, and obtains the feature with the highest pedestrian image to be checked of characteristic similarity of target pedestrian image, then those are waited for
Examining the pedestrian in pedestrian image and the pedestrian in target pedestrian image is largely same people.
In this example embodiment, pedestrian image to be checked can be to be set by multiple Image Acquisition that different location is arranged in
The original image comprising pedestrian of standby acquisition is formed by image by human testing and human region extraction；Target pedestrian image
It can be the pedestrian image chosen from the image library for being stored in terminal device 101 or server 103, such as user can pass through
The hardware device (such as mouse, stylus) connecting with terminal device 101 selects to need the human body of retrieval and inquisition from image library
Image is as target pedestrian image.
It, can be right respectively using human body recognition method of the invention after obtaining pedestrian image and target pedestrian image to be checked
Pedestrian image and target pedestrian image to be checked is handled, to obtain pedestrian image feature and target pedestrian's characteristics of image to be checked.
It, can be by calculating mesh in order to judge whether the pedestrian in target pedestrian image and the pedestrian in pedestrian image to be checked are same people
The similarity of pedestrian image feature and pedestrian image feature to be checked is marked, and whether the pedestrian in two images is judged according to similarity
For same people.It may be same people if similarity is high；It is not same people if similarity is low.
In this example embodiment, can by calculate target pedestrian characteristics of image and pedestrian image feature to be checked away from
From similarity is obtained, which can be Euclidean distance or COS distance, naturally it is also possible to be other distances, apart from smaller, phase
Higher like spending, the probability that pedestrian is same people in two images is also bigger；Vice versa.
In this example embodiment, target pedestrian characteristics of image is being obtained at a distance from multiple pedestrian image features to be checked
Afterwards, can by those pedestrian images to be checked according to pedestrian image feature to be checked and the distance of target pedestrian's characteristics of image from the near to the distant
Arrangement, come foremost is exactly the highest pedestrian image to be checked of similarity, and coming last is exactly minimum to be checked of similarity
Pedestrian image.For the ease of user's observation, pedestrian image to be checked successively can be sent to terminal device 101 in sequence and carried out
It shows.
As another embodiment of the present invention, the multiple original graphs comprising pedestrian that image capture device can be acquired
As human body recognition method according to the invention is handled, obtain multiple pedestrian image features, and by pedestrian image feature with it is right
The pedestrian image answered is stored in server 103, forms feature database.When user selects to need to retrieve in terminal device 101 to look into
After the target pedestrian image of inquiry, server 103 handles target pedestrian image human body recognition method according to the invention, obtains
Target pedestrian characteristics of image is obtained, then calculates each of target pedestrian characteristics of image and feature database pedestrian image feature again
Distance, to judge the pedestrian image feature to be checked that there is high similarity with target pedestrian characteristics of image, and by those pedestrians to be checked
The corresponding pedestrian image to be checked of characteristics of image is sent to terminal device 101.
It is described in detail below in conjunction with a concrete application scene of the Fig. 7 to the embodiment of the present invention, in the application scenarios
In, image capture device can be set at mall entrance, pedestrian's weight is carried out based on the image capture device acquired image
Identification.Fig. 7 shows the flow chart that pedestrian identifies again, as shown in fig. 7, in step s 701, by the way that mall entrance is arranged in
Image capture device acquisition includes the original image of pedestrian；In step S702, the pedestrian in the original image of acquisition is carried out
Human testing obtains a rectangle frame comprising pedestrian's human body, that is, forms pedestrian image frame；In step S703, pedestrian is schemed
Human body in frame is detected, obtain only comprising pedestrian do not include background a pedestrian image, then to pedestrian image at
Reason obtains pedestrian image feature, while pedestrian image feature being stored in feature database with corresponding pedestrian image；In step
In S704, the high pedestrian image feature to be checked of similarity is retrieved in feature database according to target pedestrian's characteristics of image, and according to phase
It is ranked up like degree height；In step S705, search result is shown according to the sequence of similarity from high to low.
Although the present invention is not it is worth noting that, above-mentioned application scenarios are identified to the pedestrian in market
It is confined to the application scenarios, human body recognition method of the invention can also be applied to the scenes such as meeting, party, and the present invention is herein not
It repeats again.
The device of the invention embodiment introduced below can be used for executing the above-mentioned human body recognition method of the present invention.For
Undisclosed details in apparatus of the present invention embodiment please refers to the embodiment of the above-mentioned human body recognition method of the present invention.
Fig. 8 shows a kind of structural schematic diagram of human bioequivalence device.Referring to shown in Fig. 8, human bioequivalence device 800 can be with
It include: the first acquisition module 801, characteristic extracting module 802, substep pond module 803 and full link block 804.
Specifically, first module 801 is obtained, for obtaining pedestrian image；Characteristic extracting module 802, it is described for extracting
The feature of pedestrian image, to obtain global characteristics figure；Substep pond module 803, for the global characteristics figure is square along presetting
To being divided into multiple regions, and pond is carried out to each region to obtain pond region corresponding with each region；Full connection
Module 804, for the corresponding pond provincial characteristics in each pond region to be sequentially connected along the preset direction, to obtain
State the corresponding pedestrian image feature of pedestrian image.
In this example embodiment, human bioequivalence device 800 further includes the second acquisition module 805, human detection module
806 and image generation module 807.
Specifically, second module 805 is obtained, for obtaining the original image comprising pedestrian；Human detection module 806 is used
In by the original image carry out human testing, to obtain pedestrian image frame；Image generation module 807 is used for from described
The corresponding region of pedestrian is extracted in pedestrian image frame, to form the pedestrian image.
Further, substep pond module 803 includes zoning unit 8031 and pond unit 8032.
Specifically, zoning unit 8031, it is multiple to be formed for the global characteristics figure to be averaged piecemeal along y direction
The region；Pond unit 8032, for carrying out average pond to each region respectively, to obtain multiple areas Chi Hua
Domain.
In this example embodiment, human bioequivalence device 800 further includes pond characteristic extracting module 808, for passing through
The convolution kernel of default size extracts the feature in each pond region, to obtain the pond provincial characteristics.
In this example embodiment, the pedestrian image includes pedestrian image and target pedestrian image to be checked, and human body is known
Other device 800 further includes similarity calculation module 809 and judgment module 810.
Specifically, similarity calculation module 809, for special according to the corresponding pedestrian image to be checked of the pedestrian image to be checked
The corresponding target pedestrian characteristics of image of the target pedestrian image of seeking peace, calculate the target pedestrian characteristics of image with it is described to be checked
The similarity of pedestrian image feature；Judgment module 810, for judging the row in the pedestrian image to be checked according to the similarity
Whether people is pedestrian in the target pedestrian image.
Further, similarity calculation module 809 includes metrics calculation unit 8091 and similarity calculated 8092.
Specifically, metrics calculation unit 8091 are schemed for calculating the target pedestrian characteristics of image and the pedestrian to be checked
As the distance of feature；Similarity calculated 8092, for being schemed according to the target pedestrian characteristics of image and the pedestrian to be checked
As the distance of feature, the similarity of the target pedestrian characteristics of image Yu the pedestrian image feature to be checked is determined.
In this example embodiment, human bioequivalence device 800 further includes arrangement module 811, for according to the target
Size of the pedestrian image feature at a distance from the pedestrian image feature to be checked, arranges the pedestrian image to be checked from the near to the remote.
In this example embodiment, human bioequivalence device 800 further includes that third obtains module 812 and pedestrian image to be checked
Generation module 813.
Specifically, third obtains module 812, for obtaining the original graph comprising pedestrian acquired by image capture device
Picture；Pedestrian image generation module 813 to be checked, for carrying out human testing to the original image and extracting human region, to obtain
Obtain the pedestrian image to be checked.
In this example embodiment, human bioequivalence device 800 further includes target pedestrian image generation module 814, is used for
The human body image of user's selection is received, and using the human body image as the target pedestrian image.
In this example embodiment, human bioequivalence device 800 further includes the 4th acquisition module 815 and model training module
816。
Specifically, the 4th module 815 is obtained, for obtaining multiple human body mark images, each human body marks image pair
Answer different scenes；Model training module 816, for human body mark image to be input to pedestrian's identification model, with right
Pedestrian's identification model is trained.
Each functional module and above-mentioned human body recognition method due to the human bioequivalence device of example embodiments of the present invention
Example embodiment the step of it is corresponding, therefore details are not described herein.
It should be noted that although being referred to several modules or unit of human bioequivalence device in the above detailed description,
It is that this division is not enforceable.In fact, embodiment according to the present invention, two or more above-described modules or
The feature and function of person's unit can embody in a module or unit.Conversely, an above-described module or
The feature and function of unit can be to be embodied by multiple modules or unit with further division.
Those skilled in the art after considering the specification and implementing the invention disclosed here, will readily occur to of the invention its
Its embodiment.This application is intended to cover any variations, uses, or adaptations of the invention, these modifications, purposes or
Person's adaptive change follows general principle of the invention and including the undocumented common knowledge in the art of the present invention
Or conventional techniques.The description and examples are only to be considered as illustrative, and true scope and spirit of the invention are by appended
Claim is pointed out.
It should be understood that the present invention is not limited to the precise structure already described above and shown in the accompanying drawings, and
And various modifications and changes may be made without departing from the scope thereof.The scope of the present invention is only limited by the attached claims.
Claims (15)
1. a kind of human body recognition method characterized by comprising
Obtain pedestrian image；
The feature of the pedestrian image is extracted, to obtain global characteristics figure；
The global characteristics figure is divided into multiple regions along preset direction, and pond is carried out to obtain and each institute to each region
State the corresponding pond region in region；
The corresponding pond provincial characteristics in each pond region is sequentially connected along the preset direction, to obtain pedestrian's figure
As corresponding pedestrian image feature.
2. human body recognition method according to claim 1, which is characterized in that before obtaining pedestrian image, the human body
Recognition methods further include:
Obtain the original image comprising pedestrian；
By carrying out human testing to the original image, to obtain pedestrian image frame；
The corresponding region of pedestrian is extracted, from the pedestrian image frame to form the pedestrian image.
3. human body recognition method according to claim 1, which is characterized in that by the global characteristics figure along preset direction point
For multiple regions, and pond is carried out to each region to obtain pond region corresponding with each region, comprising:
The global characteristics figure is averaged piecemeal along y direction, to form multiple regions；
Average pond is carried out to each region respectively, to obtain multiple pond regions.
4. human body recognition method according to claim 1, which is characterized in that the human body recognition method further include:
The feature of the pedestrian image is extracted, by feature extraction network to obtain the global characteristics figure；
The global characteristics figure is divided into multiple regions along the preset direction by substep pond network, and to each region
Pond is carried out to obtain multiple pond regions；
The corresponding pond provincial characteristics in each pond region is sequentially connected along the preset direction by fully-connected network, with
Obtain the pedestrian image feature；
Wherein, the feature extraction network, substep pond network and the fully-connected network constitute pedestrian's identification model.
5. human body recognition method according to claim 4, which is characterized in that pedestrian's identification model is based on dense volume
The model of product neural network.
6. human body recognition method according to claim 1, which is characterized in that the human body recognition method further include:
Convolution kernel by presetting size extracts the feature in each pond region, to obtain the pond provincial characteristics.
7. human body recognition method according to claim 1, which is characterized in that the pedestrian image includes pedestrian image to be checked
With target pedestrian image；The human body recognition method further include:
According to the corresponding pedestrian image feature to be checked of pedestrian image to be checked and the corresponding target line of the target pedestrian image
People's characteristics of image calculates the similarity of the target pedestrian characteristics of image Yu the pedestrian image feature to be checked；
Judge whether the pedestrian in the pedestrian image to be checked is pedestrian in the target pedestrian image according to the similarity.
8. human body recognition method according to claim 7, which is characterized in that calculate the target pedestrian characteristics of image and institute
State the similarity of pedestrian image feature to be checked, comprising:
The target pedestrian characteristics of image is calculated at a distance from the pedestrian image feature to be checked；
According to the target pedestrian characteristics of image at a distance from the pedestrian image feature to be checked, the target pedestrian image is determined
The similarity of feature and the pedestrian image feature to be checked.
9. human body recognition method according to claim 8, which is characterized in that the human body recognition method further include:
According to size of the target pedestrian characteristics of image at a distance from the pedestrian image feature to be checked, institute is arranged from the near to the remote
State pedestrian image to be checked.
10. human body recognition method according to claim 7, which is characterized in that the human body recognition method further include:
Obtain the original image comprising pedestrian acquired by image capture device；
Human testing is carried out to the original image and extracts human region, to obtain the pedestrian image to be checked.
11. human body recognition method according to claim 7, which is characterized in that the human body recognition method further include:
The human body image of user's selection is received, and using the human body image as the target pedestrian image.
12. human body recognition method according to claim 4, which is characterized in that the human body recognition method further include:
Multiple human body mark images are obtained, each human body mark image corresponds to different scenes；
Human body mark image is input to pedestrian's identification model, to be trained to pedestrian's identification model.
13. a kind of human bioequivalence device characterized by comprising
First obtains module, for obtaining pedestrian image；
Characteristic extracting module, for extracting the feature of the pedestrian image, to obtain global characteristics figure；
Substep pond module, for the global characteristics figure to be divided into multiple regions along preset direction, and to each region into
Row pond is to obtain pond region corresponding with each region；
Full link block, for the corresponding pond provincial characteristics in each pond region to be sequentially connected along the preset direction,
To obtain the corresponding pedestrian image feature of the pedestrian image.
14. a kind of computer-readable medium, is stored thereon with computer program, which is characterized in that described program is held by processor
Such as human body recognition method of any of claims 1-12 is realized when row.
15. a kind of electronic equipment characterized by comprising
One or more processors；
Storage device, for storing one or more programs, when one or more of programs are by one or more of processing
When device executes, so that one or more of processors realize such as human bioequivalence side of any of claims 1-12
Method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810681840.7A CN108960114A (en) | 2018-06-27 | 2018-06-27 | Human body recognition method and device, computer readable storage medium and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810681840.7A CN108960114A (en) | 2018-06-27 | 2018-06-27 | Human body recognition method and device, computer readable storage medium and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108960114A true CN108960114A (en) | 2018-12-07 |
Family
ID=64487364
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810681840.7A Pending CN108960114A (en) | 2018-06-27 | 2018-06-27 | Human body recognition method and device, computer readable storage medium and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108960114A (en) |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109977823A (en) * | 2019-03-15 | 2019-07-05 | 百度在线网络技术（北京）有限公司 | Pedestrian's recognition and tracking method, apparatus, computer equipment and storage medium |
| CN110175587A (en) * | 2019-05-30 | 2019-08-27 | 黄岩 | A kind of video frequency tracking method based on recognition of face and Algorithm for gait recognition |
| CN110334677A (en) * | 2019-07-11 | 2019-10-15 | 山东大学 | A kind of recognition methods again of the pedestrian based on skeleton critical point detection and unequal subregion |
| CN110543841A (en) * | 2019-08-21 | 2019-12-06 | 中科视语(北京)科技有限公司 | Pedestrian re-identification method, system, electronic device and medium |
| CN110555401A (en) * | 2019-08-26 | 2019-12-10 | 浙江大学 | self-adaptive emotion expression system and method based on expression recognition |
| CN110555420A (en) * | 2019-09-09 | 2019-12-10 | 电子科技大学 | fusion model network and method based on pedestrian regional feature extraction and re-identification |
| CN110598716A (en) * | 2019-09-09 | 2019-12-20 | 北京文安智能技术股份有限公司 | Personnel attribute identification method, device and system |
| CN110738101A (en) * | 2019-09-04 | 2020-01-31 | 平安科技（深圳）有限公司 | Behavior recognition method and device and computer readable storage medium |
| CN110929711A (en) * | 2019-11-15 | 2020-03-27 | 智慧视通(杭州)科技发展有限公司 | Method for automatically associating identity information and shape information applied to fixed scene |
| CN111078940A (en) * | 2019-12-16 | 2020-04-28 | 腾讯科技（深圳）有限公司 | Image processing method, image processing device, computer storage medium and electronic equipment |
| CN111581418A (en) * | 2020-04-29 | 2020-08-25 | 山东科技大学 | Target person searching method based on image associated person information |
| CN112054979A (en) * | 2020-09-14 | 2020-12-08 | 四川大学 | Radio automatic modulation identification method based on fuzzy dense convolution network |
| WO2021017316A1 (en) * | 2019-07-30 | 2021-02-04 | 平安科技（深圳）有限公司 | Residual network-based information recognition method, apparatus, and computer device |
| CN112801008A (en) * | 2021-02-05 | 2021-05-14 | 电子科技大学中山学院 | Pedestrian re-identification method and device, electronic equipment and readable storage medium |
| CN112801020A (en) * | 2021-02-09 | 2021-05-14 | 福州大学 | Pedestrian re-identification method and system based on background graying |
| WO2021196547A1 (en) * | 2020-03-31 | 2021-10-07 | 北京迈格威科技有限公司 | Person re-identification method and apparatus, electronic device and storage medium |
| CN113837244A (en) * | 2021-09-02 | 2021-12-24 | 哈尔滨工业大学 | Confrontation sample detection method and device based on multilayer significance characteristics |
| CN113870454A (en) * | 2021-09-29 | 2021-12-31 | 平安银行股份有限公司 | Attendance checking method and device based on face recognition, electronic equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107145889A (en) * | 2017-04-14 | 2017-09-08 | 中国人民解放军国防科学技术大学 | Target identification method based on double CNN networks with RoI ponds |
| US20170270387A1 (en) * | 2016-03-15 | 2017-09-21 | Tata Consultancy Services Limited | Method and system for unsupervised word image clustering |
-
2018
- 2018-06-27 CN CN201810681840.7A patent/CN108960114A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170270387A1 (en) * | 2016-03-15 | 2017-09-21 | Tata Consultancy Services Limited | Method and system for unsupervised word image clustering |
| CN107145889A (en) * | 2017-04-14 | 2017-09-08 | 中国人民解放军国防科学技术大学 | Target identification method based on double CNN networks with RoI ponds |
Non-Patent Citations (2)
| Title |
|---|
| SUN, YIFAN，ET AL: "《Beyond Part Models: Person Retrieval with Refined Part Pooling (and a Strong Convolutional Baseline)》", 《ARXIV E-PRINTS》 * |
| 王亦民: "《面向监控视频的行人重识别技术研究》", 《中国优秀博硕士学位论文全文数据库(博士)信息科技辑》 * |
Cited By (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109977823A (en) * | 2019-03-15 | 2019-07-05 | 百度在线网络技术（北京）有限公司 | Pedestrian's recognition and tracking method, apparatus, computer equipment and storage medium |
| CN109977823B (en) * | 2019-03-15 | 2021-05-14 | 百度在线网络技术（北京）有限公司 | Pedestrian recognition and tracking method and device, computer equipment and storage medium |
| CN110175587B (en) * | 2019-05-30 | 2020-03-24 | 黄岩 | Video tracking method based on face recognition and gait recognition algorithm |
| CN110175587A (en) * | 2019-05-30 | 2019-08-27 | 黄岩 | A kind of video frequency tracking method based on recognition of face and Algorithm for gait recognition |
| CN110334677A (en) * | 2019-07-11 | 2019-10-15 | 山东大学 | A kind of recognition methods again of the pedestrian based on skeleton critical point detection and unequal subregion |
| WO2021017316A1 (en) * | 2019-07-30 | 2021-02-04 | 平安科技（深圳）有限公司 | Residual network-based information recognition method, apparatus, and computer device |
| CN110543841A (en) * | 2019-08-21 | 2019-12-06 | 中科视语(北京)科技有限公司 | Pedestrian re-identification method, system, electronic device and medium |
| CN110555401A (en) * | 2019-08-26 | 2019-12-10 | 浙江大学 | self-adaptive emotion expression system and method based on expression recognition |
| CN110555401B (en) * | 2019-08-26 | 2022-05-03 | 浙江大学 | Self-adaptive emotion expression system and method based on expression recognition |
| CN110738101A (en) * | 2019-09-04 | 2020-01-31 | 平安科技（深圳）有限公司 | Behavior recognition method and device and computer readable storage medium |
| CN110738101B (en) * | 2019-09-04 | 2023-07-25 | 平安科技（深圳）有限公司 | Behavior recognition method, behavior recognition device and computer-readable storage medium |
| WO2021042547A1 (en) * | 2019-09-04 | 2021-03-11 | 平安科技（深圳）有限公司 | Behavior identification method, device and computer-readable storage medium |
| CN110598716A (en) * | 2019-09-09 | 2019-12-20 | 北京文安智能技术股份有限公司 | Personnel attribute identification method, device and system |
| CN110555420A (en) * | 2019-09-09 | 2019-12-10 | 电子科技大学 | fusion model network and method based on pedestrian regional feature extraction and re-identification |
| CN110555420B (en) * | 2019-09-09 | 2022-04-12 | 电子科技大学 | Fusion model network and method based on pedestrian regional feature extraction and re-identification |
| CN110929711A (en) * | 2019-11-15 | 2020-03-27 | 智慧视通(杭州)科技发展有限公司 | Method for automatically associating identity information and shape information applied to fixed scene |
| CN111078940A (en) * | 2019-12-16 | 2020-04-28 | 腾讯科技（深圳）有限公司 | Image processing method, image processing device, computer storage medium and electronic equipment |
| CN111078940B (en) * | 2019-12-16 | 2023-05-23 | 腾讯科技（深圳）有限公司 | Image processing method, device, computer storage medium and electronic equipment |
| WO2021196547A1 (en) * | 2020-03-31 | 2021-10-07 | 北京迈格威科技有限公司 | Person re-identification method and apparatus, electronic device and storage medium |
| CN111581418A (en) * | 2020-04-29 | 2020-08-25 | 山东科技大学 | Target person searching method based on image associated person information |
| CN112054979A (en) * | 2020-09-14 | 2020-12-08 | 四川大学 | Radio automatic modulation identification method based on fuzzy dense convolution network |
| CN112801008A (en) * | 2021-02-05 | 2021-05-14 | 电子科技大学中山学院 | Pedestrian re-identification method and device, electronic equipment and readable storage medium |
| CN112801008B (en) * | 2021-02-05 | 2024-05-31 | 电子科技大学中山学院 | Pedestrian re-recognition method and device, electronic equipment and readable storage medium |
| CN112801020A (en) * | 2021-02-09 | 2021-05-14 | 福州大学 | Pedestrian re-identification method and system based on background graying |
| CN113837244A (en) * | 2021-09-02 | 2021-12-24 | 哈尔滨工业大学 | Confrontation sample detection method and device based on multilayer significance characteristics |
| CN113870454A (en) * | 2021-09-29 | 2021-12-31 | 平安银行股份有限公司 | Attendance checking method and device based on face recognition, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108960114A (en) | Human body recognition method and device, computer readable storage medium and electronic equipment | |
| CN108960090B (en) | Video image processing method and device, computer readable medium and electronic equipment | |
| CN108269254B (en) | Image quality evaluation method and device | |
| CN108256479B (en) | Face tracking method and device | |
| CN108898086A (en) | Method of video image processing and device, computer-readable medium and electronic equipment | |
| CN109145766A (en) | Model training method, device, recognition methods, electronic equipment and storage medium | |
| CN110765954A (en) | Vehicle weight recognition method, equipment and storage device | |
| CN107918767B (en) | Object detection method, device, electronic equipment and computer-readable medium | |
| CN113449700B (en) | Training of video classification model, video classification method, device, equipment and medium | |
| US20230060211A1 (en) | System and Method for Tracking Moving Objects by Video Data | |
| CN109902681B (en) | User group relation determining method, device, equipment and storage medium | |
| Zhang et al. | Indoor space recognition using deep convolutional neural network: a case study at MIT campus | |
| CN113139540B (en) | Backboard detection method and equipment | |
| CN111709382A (en) | Human body trajectory processing method and device, computer storage medium and electronic equipment | |
| CN112232311A (en) | Face tracking method and device and electronic equipment | |
| CN115439927A (en) | Gait monitoring method, device, equipment and storage medium based on robot | |
| Singh et al. | Performance enhancement of salient object detection using superpixel based Gaussian mixture model | |
| US20220300774A1 (en) | Methods, apparatuses, devices and storage media for detecting correlated objects involved in image | |
| Gao et al. | Occluded person re-identification based on feature fusion and sparse reconstruction | |
| CN116824641B (en) | Gesture classification method, device, equipment and computer storage medium | |
| CN111310595B (en) | Method and device for generating information | |
| CN110309790B (en) | Scene modeling method and device for road target detection | |
| CN116958873A (en) | Pedestrian tracking method, device, electronic equipment and readable storage medium | |
| CN114694257B (en) | Multi-user real-time three-dimensional action recognition evaluation method, device, equipment and medium | |
| CN110738149A (en) | Target tracking method, terminal and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181207 |