CN108109010A - A kind of intelligence AR advertisement machines - Google Patents
A kind of intelligence AR advertisement machines Download PDFInfo
- Publication number
- CN108109010A CN108109010A CN201711428451.5A CN201711428451A CN108109010A CN 108109010 A CN108109010 A CN 108109010A CN 201711428451 A CN201711428451 A CN 201711428451A CN 108109010 A CN108109010 A CN 108109010A
- Authority
- CN
- China
- Prior art keywords
- face
- advertisement
- camera
- intelligence
- advertisement machine
- 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
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T19/00—Manipulating 3D models or images for computer graphics
- G06T19/006—Mixed reality
-
- 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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Business, Economics & Management (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Human Computer Interaction (AREA)
- General Health & Medical Sciences (AREA)
- Accounting & Taxation (AREA)
- Multimedia (AREA)
- Strategic Management (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Economics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Computer Graphics (AREA)
- Computer Hardware Design (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention discloses a kind of intelligence AR advertisement machines, including screen, processor, memory, network connection, camera etc., camera is connected to processor by interface, user is by camera collection image during advertisement machine camera and is transmitted to processor, the advertisement machine receives the image of camera acquisition, carry out Face datection, carry out face critical point detection and tracking, identify human face expression variation, and pass through network connection and be uploaded to high in the clouds automatically, the link of wechat public platform Quick Response Code is returned after server calls wechat API, advertisement machine generates Quick Response Code according to the link of return and is shown on advertisement machine screen, when scanning input Quick Response Code and concern or entrance public platform, wechat public platform sends photo to user automatically.The present invention realizes the real-time AR human-computer interactions of electrical screen under line, and interactive effect is made cost-effective while to have obtained effective promotion.
Description
Technical field
The invention belongs to retail customer's management domains under line, relate in particular to a kind of intelligence AR advertisement machines.
Background technology
Advertisement machine has obtained substantial amounts of application at present, but existing advertisement machine is in many aspects that there are shortcomings.It is such as first
First in terms of human-computer interaction, current technology is unable to reach the human-computer interaction realized in real time by face, and existing electrical screen is all
That unidirectional information is transferred, i.e., the interaction for not interacting or being realized by touch screen mode.Secondly in content appearance form side
Face, existing technology can be realized through camera to take pictures and be presented photo on the screen, it might even be possible to add after taking pictures
It is added on photo and adds virtual image, but can not technically realize and just add virtual shape when opening camera in real time
As so as to be unable to reach interactive effect.In addition unidirectional information is transferred so that electrical screen is difficult to directly precisely collect user
Information, so as to be difficult to precisely launch related content or advertisement for specific user.
Current electronic curtain has been equipped with Android system, and is equipped with processor and memory, this causes electrical screen
Curtain can install application program as smart mobile phone.Some electrical screens have the function of touch screen, can allow user's touch screen
Mode and screen are interactive.On the whole, the content of electronic curtain displaying at present still concentrates on rolling advertisement content displaying or regards
Frequency plays.
On the other hand, occur the application of face AR, such as the Snapchat applications in the U.S. above existing smart mobile phone,
Can virtual paster model, even virtual advertisement image be rendered on user's face by mobile phone camera in real time.This
The process that a application is realized is as follows:Detect face;Find out the key point above face;These key points of real-time tracking;
One virtual object of real-time rendering on the basis of these key points.But this formula is only realized at present in mobile phone terminal.
Electrical screen under current line, due to the limitation of algorithm, existing electronic curtain application program is difficult to realize real-time AR
Interactive function, mostly concentrates on rolling advertisement content displaying or video playing, it is few, interior so to cause electrical screen attention rate
The problems such as holding dullness etc..Large screen can be realized under only a few line renders a virtual portrait or object in real picture,
But it can not realize the interaction between this virtual image and true man.Electrical screen is all by touching under the line with human-computer interaction at present
Shield what is realized, but so add the cost of electrical screen.
Additionally, due on the electronic curtain under online it is the distance between man-machine much larger than people with mobile phone self-timer when distance,
Therefore the interactive effects of AR for such as Snapchat classes realized by smart mobile phone play electronic curtain to be difficult to realize online at present.
Current existing algorithm does not go to optimize in this respect, and face can not be captured in medium and long distance or frame losing etc. occurs now
As.
The content of the invention
For the above-mentioned problems in the prior art, the present invention proposes a kind of new intelligent AR advertisement machines, with compared with
Low cost realizes the human-computer interaction of advertisement machine, and can effectively collect user information, realizes that accurately advertisement is launched.
Specifically, present invention employs following technical schemes:
A kind of intelligence AR advertisement machines, the advertisement machine include display screen, processor, memory, network connection, camera, electricity
Source, which is characterized in that the camera is located at 1.3 meters to 1.8 meters of ground, and camera is connected to processing by interface
Device, user is by camera collection image during advertisement machine camera and is transmitted to processor, and the advertisement machine performs following steps
Operation:
1)The image of camera acquisition is received, face inspection is carried out using the Face datection algorithm such as MTCNN based on deep learning
It surveys, captures facial image and return to face rectangle frame data;
2)Face critical point detection and tracking are carried out to the face captured, capture 68 key points on face;
3)Pre-production face 3D models, and adjust according to the key point information being detected above the attitude angle of face 3D models
Degree wherein representing the angle of face pitching, deflection and overturning with Eulerian angles, is recorded when face rotates on curtain every 5 degree and tracked
Three angle values that device returns are carried out linear regression to three angle values obtained using which, obtain pitching, deflect and turn over
Turn the regression model of three angle values, the angle value that tracker is returned inputs regression model, estimates human face three-dimensional model
Conjecture face model is gone to corresponding angle by rotational angle further according to the angle of the estimation;
4)Make a set of comprising the set several groups of human face expressions with merging amplitude of deformation, each group of human face expression is with merging deformation
Amplitude includes several key points, and each group of expression is calculated with merging deformation width according to the key point information obtained in above step
Human face expression distance in degree;
5)According to human face expression distance identification human face expression variation, and taken pictures according to the Given Face expression triggering recognized;
6)Photo and label information are uploaded to high in the clouds automatically by network connection, and it is public to return to wechat after server calls wechat API
Crowd's Quick Response Code links, and advertisement machine generates Quick Response Code according to the link of return and shown on advertisement machine screen, when scanning input two
When tieing up code and paying close attention to or enter public platform, wechat public platform sends photo to user automatically.
Wherein in step 1)In, carrying out Face datection using MTCNN algorithms includes three phases:Stage 1:In structure image
It on the basis of pyramidal, is detected using full convolutional network, while is returned using Boundingbox and be modified with NMS;Rank
Section 2:It will be determined whether by the input of all windows in stage 1, while still be Boundingbox recurrence and NMS;Stage
3:Window input is made to judge in the case where further increasing the constraint of 5 face key points and does Boundingbox recurrence
And NMS.
In step 2)In, face critical point detection includes with tracking:When reading first frame picture frame, current image frame is captured
In human face region, using machine learning algorithm detect face in key point, the later picture frame of first frame use feature
The tracking mode matched somebody with somebody detects above-mentioned key point near former frame human face region, and captures the human face region of current image frame.
In step 4)In, it is obtained by calculating each group of human face expression to merging the coordinate difference of corresponding key point in amplitude of deformation
Go out the corresponding human face expression distance on this key point.
Further, after photo and label information are uploaded, the step of image is further analyzed is further included, it is described
Analysis includes basic data analysis, recognition of face, gender identification, age identification, mood identification, and to the face size on image
Standardization is adjusted, all analysis result information are finally combined storage.Further, after analytical procedure, also
Including according to the obtained result information of analysis come the step of targetedly launching corresponding advertisement.Wherein preferably, launched
Each ad content can be preceded by adaptation population's label in advance, the result information that server is obtained according to previous analysis carries out
Match somebody with somebody, ranking is carried out to most suitable ad content, and select most suitable advertisement return advertisement machine screen and play out.In addition it is excellent
Selection of land is collected simultaneously the viewing duration and expression shape change of beholder when advertisement plays.
Advantageous effect:The present invention realizes the real-time AR human-computer interactions of electrical screen under line, makes cost-effective while mutually
Dynamic effect has obtained effective promotion.The present invention realizes and multiple faces is quick and precisely captured when medium and long distance, and
And phenomena such as improving stability, being not in frame losing.Using AR effects in the advertisement machine of the present invention, known by application face
Do not understand user, and collect user feedback, filled up the market vacancy under this scene, realize for different use
The function of different AR ad contents is launched at family according to face recognition algorithms and big data analysis respectively.
Description of the drawings
Fig. 1 is the realization flow chart of the solution of the present invention;
Fig. 2 is the flow chart of Face datection identification used in the present invention program;
Fig. 3 is the flow chart that AR effects are realized in the present invention program;
Fig. 4 is the schematic diagram of image analysis process.
Specific embodiment
In the prior art, easily can realize that AR is interactive on mobile phone, but online under electrical screen scene in,
Since passing crowd is stranger group, i.e., have very that maximum probability is not same person using the people of electrical screen every time, so having
Necessity goes to understand current user by recognition of face, this is that current phone AR cannot accomplish.It is and electric under existing line
Sub-screen presentation mode is single, and advertisement dispensing form is extensive, it is difficult to user information is directly precisely collected, so as to be difficult to pin
Related content or advertisement are precisely launched to specific user, so as to cannot effectively play the effect of electronic curtain under line.In view of this,
The invention discloses a kind of intelligence AR advertisement machines, realize the two-way interaction of electronic curtain under line, and can be according to user's essence
Standard launches advertisement and more effectively collects user information.
The intelligent AR electronic curtains of the present invention include display screen, processor, memory, network connection, camera, power supply
Deng camera can be connected by USB interface, and camera is located at 1.3 meters to 1.6 meters of ground.Referring to Fig. 1, wherein institute
It states advertisement machine and performs following operation:
1)Video image is gathered by camera in real time, the image of each frame is calculated using the Face datection based on deep learning
Method.One of them algorithm used is MTCNN algorithms, it is a Face datection algorithm increased income(A kind of frame of deep learning
Frame)Carry out Face datection.MTCNN algorithms have been done integrated optimization by us, are particularly done on the input picture to algorithm
Engineering optimization, face figure is cut out by we using the coordinate system of face frame, and face frame is done using histogram technology and is exposed
Light balance optimizing carries out quality testing to filter out improper knowledge to increase recognition accuracy, then for eye recognition to face figure
Other face figure, then will be detected and suitably be compressed with processed face figure, operation is identified in upload server.
Sdk development kits are reduced and rely on storehouse, and are transplanted in Android system.The detection method can accomplish 10 on Android machine
Frame per second captures multiple faces simultaneously, returns to the face rectangle frame data detected, and idiographic flow is as shown in Figure 2.
Compared with Face datection algorithm existing in the market, human face detection tech of the invention is not only realized in mobile terminal
Real-time detection, but also the Android system in advertisement machine is optimized:Because most people are rested in face of advertisement machine
When, the distance between advertisement machine screen more than 1 meter, in the range of 5 meters of screen all may be used by the MTCNN algorithms that we select
To quickly detect face, existing solution is then apart from mobile phone self-timer field of the face less than 1 meter above existing market
Scape carries out Face datection.We can understand that it is because the algorithm of deep learning is based on nerve net in itself to reach more remote detection
Understanding of the network for face features, the especially detection at eyes, nose and two corners of the mouths are based on substantial amounts of mark
The training that the face characteristic being poured in carries out, although becoming smaller in larger distance face, but characteristic value still has, so using this
Method still can detect face.In comparison, the method for conventional depth study is simply judged with specific empirical equation
The position of face for the bad adaptability of different distance, becomes larger, it will be difficult to examine once the distance of face becomes remote either angle
Measure face.
Remote Face datection is realized as core algorithm present invention uses MTCNN, MTCNN is the frame of deep learning
Frame can be carried out at the same time more people's human face detection and tracings.
MTCNN algorithms divide three phases:
Stage 1:On the basis of image pyramid is built, it is detected using full convolutional network, is utilized simultaneously
Boundingbox returns with NMS to be modified.(Pay attention to:Here full convolutional network and the net with deconvolution inside R-CNN
Network is different, is only referred to only convolutional layer here, can receive the input of arbitrary dimension, is come by network stride automatic complete
Into sliding window);
Stage 2:Will by all windows in stage 1 input determine whether, while to be also Boundingbox return with
NMS。
Stage 3:It is similar with the stage 2, only add stronger constraint:5 face key points.
MTCNN network structures are slightly deep compared with CascadeCNN but less per layer parameter, institute's same speed per hour of better performances in this way
Degree and CascadeCNN are also very nearly the same.In the training process, a variety of wide-angle face atlas are increased to training set, increases distant view
Face atlas allocates each colour of skin ethnic group face atlas ratio.
2)With reference to figure 3, after face is detected, the key point of face is captured, and it is real with the change in location of face
When track human faces:The face tracking technology of the present invention is optimized for the purposes of the scene of advertisement machine, existing 68 passes
On the basis of key point captures, medium and long distance is realized(1-5 meters of scopes)Tracking effect, and in real time return the number of people turn to
Yaw, roll and pitch angle value.
The present invention realizes medium and long distance Face datection on the basis of MTCNN human-face detectors.It is this to be based on depth
The method of study will not be influenced by bright and dark light, so as to ensure the stability of Face datection.After detecting face, pass through identification
The tracer technique of 68 key points of face stablizes the effect of tracking to realize, and according to face 3D model corner model data meters
Calculate yaw, roll, pitch three classes angle values.
3)By camera when face is detected automatically in backstage sectional drawing, and the photo upload containing face extremely
Cloud Server does face discriminance analysis.This sectional drawing and the mechanism uploaded, are combined with detection and identification model, are only detecting
First frame occurs, and avoids Multiple detection and more occupancy network flow bandwidths.The present invention realizes in engineering to be taken the photograph by single
As head is integrated with the effect of recognition of face and face augmented reality, this is global the first.
When user passes through advertisement machine, detector quickly detects face, then detects human face data by tracker, captures people
68 key point location informations of face draw 3DAR models according to these face informations.When user's expression changes(Such as choose
Eyebrow is opened one's mouth), algorithm can identify that the particular emotion triggering of user is taken pictures, and onscreen cue is taken pictures countdown, and countdown terminates just
Automatically take pictures for user, after taking pictures, photo and label information are uploaded to high in the clouds automatically by wireless network, and server calls are micro-
The link of wechat public platform Quick Response Code is returned to after believing API, advertisement machine generates Quick Response Code on advertisement machine screen according to the link of return
It has been shown that, display time can set, and user can scan the two-dimensional code within the display time, pay close attention to or into public platform, the wechat public
Photo number will be sent automatically to user, and photo shows after the time that advertisement machine recovers camera and detects state again.
4)Further as shown in figure 4, building the background analysis frame of Cloud Server for intelligent AR advertisement machines.Intelligent AR advertisements
Machine is sent to backstage after the image information of acquisition user, by wireless network, and user images are analyzed and processed from the background.Point
Not Tiao Yong basic data engine, face recognition engine, gender identification engine, age identification engine, mood identification engine etc., to figure
As carrying out information analysis.In analysis, picture background denoising sound function can be provided, the automation of face size is provided and adjusts verification
Deng ensureing the accuracy of information.The multi engine technology of the present invention can handle a pictures in hundreds of milliseconds and return to intelligence
AR advertisement machines, entire picture processing time, including network traffic hours, no more than 200 milliseconds.Cloud service backstage backstage relies on deep
Degree study state-of-the-art technology carries out the accurate identification prediction of higher, with efficient parallel computing capability quick obtaining analysis result, meets
The concurrent Real time request processing of more intelligence AR advertisement machines.
In the result of return, scanning information of unique ID of face, the ID of advertisement machine, Quick Response Code etc., this hair are contained
Bright intelligent platform precisely can launch advertisement according to this result.
5)One is fed back accurately after advertisement special efficacy ID in Cloud Server, and advertisement machine front end is thrown in real time according to this ID
The AR interactive type advertisements of a 3D are put, achieve the effect that precisely to launch.Each ad content can be preceded by adaptation population's mark in advance
Label, the rear end of intelligent advisement player can generate the content of corresponding label after the face in face of analysis, and field includes judging current
Age of audient, gender, mood, height and whether be member.It rear end can be according to algorithmic match come in most suitable advertisement
Hold and carry out ranking, and select most suitable advertisement return advertisement machine screen and play out.When advertisement plays, receive simultaneously
Collect the data messages such as viewing duration and the expression shape change of audient.It is the situation of member in spectators, it is automatic to play member welcome letter
Breath, and corresponding content is played according to the buying behavior of member record.
After Cloud Server handles image information, it can be by information such as the gender of user, age, moods with specific coding
Form return to advertisement machine, advertisement machine is calculated advertisement ID the most matched therewith, is passed through by the parsing to specific coding
Advertising model in ID reading databases will launch out corresponding AR interactive advertisements on advertisement machine screen.User passes through screen
On signal language can carry out the diversified interaction such as gesture, action, expression, overturn the pain spot that conventional ads " can only be seen ".It is logical
The accurate identification to user is crossed, can advertisement more effectively be propagated, advertiser can spend less money to reach maximum
Effect.User can take pictures in AR advertisement machines, pay close attention to businessman's wechat public platform, download photo, and then photo is turned
It sends out to achieve the effect that secondary propagation.In addition the wechat public platform rear end of businessman can also according to user pay close attention to public platform when
Between, place, personal information etc., further analyze user behavior custom and distribution situation etc..
The implementation of invention is based on the application on the operating system of mobile terminal, suitable for Android system and iOS system.This hair
Bright intelligent AR electrical screens can replace its hardware, including processor, memory, hard disk, camera etc., as long as keeping hardware device
Most basic requirement can be reached, i.e.,:Memory size not less than 4GB, processor for 2014 and produce afterwards species,
Camera pixel is not less than 4.0 not less than 2,000,000 pixels, Android system version and possesses the function of network connection.
The implementation of the present invention can be by being transformed common advertisement machine to carry out, and at low cost, effect is good.Common
Advertisement machine on pass through the USB interface built in machine or special utilizing camera interface one high-definition camera, camera picture be installed
Element not less than 2,000,000, by camera be fixed on it is substantially concordant with human eye height locate, it is 1.3-1.8 meters general.Pass through USB interface again
Mouse is connected, installation procedure connects wireless network.Network connection to backstage platform, the platform include face identification system and
Database, exploitation wechat public platform interface are connected with database.
The key point of the present invention is to the recognition of face under advertisement machine scene(People is in the situation apart from advertisement machine certain distance
Under can accurately detect face), the not frame losing tracking effect under medium and long distance avoids Multiple detection and occupies network traffics more
The sectional drawing and upload mechanism of bandwidth targetedly launch respective advertisement according to the result of identification.
The present invention is that can be achieved at the same time using face as the AR driven using same camera by upgrading electrical screen under line
Interactive type advertisement function, recognition of face and big data analysis function.Solve the following pain spot of electric advertisement screen under current line:1st, it is interior
Appearance form is dull, i.e., the one direction information based on picture and video is transferred;2nd, interactive touch screen advertisement machine is of high cost;3rd, it is wide
It is popular to accuse content, does not launch precisely;4th, it is difficult to collect audient's feedback.
The intelligent AR electrical screens of the present invention have following advantage:1st, AR interactive advertisement contents have reached man-machine bidirectional information
Interaction;2nd, compare the interaction of touch screen type, and the interactive electronic screen using face as driving has been terminated an agreement touch screen cost about 50%;3rd, base
In the big data analysis of recognition of face, age, gender of audient etc. can be understood in real time, and according to the accurate formula of these factors
It launches;4th, collect that audient watches duration, audient's mood, audient whether there is with sweeping Quick Response Code data collects feedback in real time.
Embodiments of the present invention are described in detail above in conjunction with the drawings and specific embodiments, but the present invention is not
It is limited to the above embodiment, in the knowledge possessed in technical field those of ordinary skill, can also departing from
It is made a variety of changes on the premise of present inventive concept.
Claims (9)
1. a kind of intelligence AR advertisement machines, the advertisement machine include display screen, processor, memory, network connection, camera,
Power supply, camera are connected to processor by interface, and user is by camera collection image during advertisement machine camera and is transmitted to
Processor, the advertisement machine perform following operation:
1)The image of camera acquisition is received, Face datection is carried out using the Face datection algorithm based on deep learning, captures
Facial image simultaneously returns to face rectangle frame data;
2)Face critical point detection and tracking are carried out to the face captured, capture 68 key points on face;
3)Pre-production face 3D models, and adjust according to the key point information being detected above the attitude angle of face 3D models
Degree wherein representing the angle of face pitching, deflection and overturning with Eulerian angles, is recorded when face rotates on curtain every 5 degree and tracked
Three angle values that device returns are carried out linear regression to three angle values obtained using which, obtain pitching, deflect and turn over
Turn the regression model of three angle values, the angle value that tracker is returned inputs regression model, estimates human face three-dimensional model
Conjecture face model is gone to corresponding angle by rotational angle further according to the angle of the estimation;
4)Make a set of comprising the set several groups of human face expressions with merging amplitude of deformation, each group of human face expression is with merging deformation
Amplitude includes several key points, and each group of expression is calculated with merging deformation width according to the key point information obtained in above step
Human face expression distance in degree;
5)According to human face expression distance identification human face expression variation, and taken pictures according to the Given Face expression triggering recognized;
6)Photo and label information are uploaded to high in the clouds automatically by network connection, and it is public to return to wechat after server calls wechat API
Crowd's Quick Response Code links, and advertisement machine generates Quick Response Code according to the link of return and shown on advertisement machine screen, when scanning input two
When tieing up code and paying close attention to or enter public platform, wechat public platform sends photo to user automatically.
2. intelligence AR advertisement machines as described in claim 1, which is characterized in that step 1)Described in the face based on deep learning
Detection algorithm is MTCNN.
3. intelligence AR advertisement machines as claimed in claim 2, which is characterized in that step 1)It is middle to carry out face using MTCNN algorithms
Detection includes three phases:Stage 1:On the basis of image pyramid is built, it is detected using full convolutional network, while profit
It is returned with Boundingbox and NMS is modified;Stage 2:It will be determined whether by the input of all windows in stage 1, together
When be still Boundingbox return and NMS;Stage 3:It is right in the case where further increasing the constraint of 5 face key points
Window input is made to judge and be Boundingbox recurrence and NMS.
4. intelligence AR advertisement machines as described in claim 1, which is characterized in that step 2)Middle face critical point detection and tracking bag
It includes:When reading first frame picture frame, the human face region in current image frame is captured, is detected using machine learning algorithm in face
Key point, the later picture frame of first frame use the tracking mode of characteristic matching, are detected near former frame human face region above-mentioned
Key point, and capture the human face region of current image frame.
5. intelligence AR advertisement machines as described in claim 1, which is characterized in that in step 4)In, by calculating each group of face
Expression draws the corresponding human face expression distance on this key point to the coordinate difference of corresponding key point in fusion amplitude of deformation.
6. intelligence AR advertisement machines as described in claim 1, which is characterized in that after photo and label information are uploaded, also wrap
The step of image is further analyzed is included, the analysis includes basic data analysis, recognition of face, gender identification, age
Identification, mood identification, and standardization is adjusted to the face size on image, finally all analysis result information are tied
Close storage.
7. intelligence AR advertisement machines as claimed in claim 6, which is characterized in that after analytical procedure, further include according to analysis
Obtained result information targetedly launches the step of corresponding advertisement.
8. intelligence AR advertisement machines as claimed in claim 7, which is characterized in that each ad content launched can add in advance
Upper adaptation population's label, the result information that server is obtained according to previous analysis are matched, to most suitable ad content into
Row ranking, and select most suitable advertisement return advertisement machine screen and play out.
9. intelligence AR advertisement machines as claimed in claim 7, which is characterized in that be collected simultaneously beholder when advertisement plays
Viewing duration and expression shape change.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711428451.5A CN108109010A (en) | 2017-12-26 | 2017-12-26 | A kind of intelligence AR advertisement machines |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711428451.5A CN108109010A (en) | 2017-12-26 | 2017-12-26 | A kind of intelligence AR advertisement machines |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108109010A true CN108109010A (en) | 2018-06-01 |
Family
ID=62213284
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711428451.5A Pending CN108109010A (en) | 2017-12-26 | 2017-12-26 | A kind of intelligence AR advertisement machines |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108109010A (en) |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108830946A (en) * | 2018-08-31 | 2018-11-16 | 佛山科学技术学院 | A kind of 3 D intelligent AR examination forwarding method and examination transmitting apparatus |
CN108960186A (en) * | 2018-07-20 | 2018-12-07 | 南京开为网络科技有限公司 | A kind of advertisement machine user identification method based on face |
CN109191184A (en) * | 2018-08-14 | 2019-01-11 | 微梦创科网络科技(中国)有限公司 | Advertisement placement method and system based on image recognition |
CN109615433A (en) * | 2018-12-17 | 2019-04-12 | 海南大学 | Merge the user satisfaction modeling and spacial flex method of adjustment of equitable degree, experience sense and price |
CN109993589A (en) * | 2019-04-16 | 2019-07-09 | 南京海帆数据科技有限公司 | A kind of advertisement machine passenger flow data analysis method based on machine vision |
CN110298706A (en) * | 2019-07-04 | 2019-10-01 | 山东浪潮人工智能研究院有限公司 | A kind of exhibition room visits interaction and takes pictures evaluation method |
CN110414419A (en) * | 2019-07-25 | 2019-11-05 | 四川长虹电器股份有限公司 | A kind of posture detecting system and method based on mobile terminal viewer |
CN110490656A (en) * | 2019-08-21 | 2019-11-22 | 深圳市易云数字科技有限责任公司 | The effect quantization method and system of advertisement under a kind of line |
CN111325051A (en) * | 2018-12-13 | 2020-06-23 | 普天信息技术有限公司 | Face recognition method and device based on face image ROI selection |
CN111711747A (en) * | 2020-05-21 | 2020-09-25 | 广州恒德智汇有限公司 | Interactive all-in-one exhibition hall group photo system |
CN112015292A (en) * | 2019-05-31 | 2020-12-01 | 上海分泽时代软件技术有限公司 | Touch screen operated advertising machine and operation method thereof |
CN113362353A (en) * | 2020-03-04 | 2021-09-07 | 上海分众软件技术有限公司 | Method for identifying advertising player frame by utilizing synthesis training picture |
CN113762095A (en) * | 2021-08-18 | 2021-12-07 | 东软集团股份有限公司 | Mask recognition method and device, storage medium and electronic equipment |
CN113807313A (en) * | 2021-10-08 | 2021-12-17 | 合肥安达创展科技股份有限公司 | AI platform analysis system based on Dlib face recognition |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101369307A (en) * | 2007-08-14 | 2009-02-18 | 索尼株式会社 | Iamge forming device, method and computer program |
CN103079034A (en) * | 2013-01-06 | 2013-05-01 | 北京百度网讯科技有限公司 | Perception shooting method and system |
CN104573619A (en) * | 2014-07-25 | 2015-04-29 | 北京智膜科技有限公司 | Method and system for analyzing big data of intelligent advertisements based on face identification |
CN204559698U (en) * | 2014-12-31 | 2015-08-12 | 上海点图网络科技有限公司 | The process of a kind of image instantaneous acquiring exports integration apparatus |
CN106055193A (en) * | 2016-05-13 | 2016-10-26 | 吴贤波 | Method and system for photo downloading |
CN106355147A (en) * | 2016-08-26 | 2017-01-25 | 张艳 | Acquiring method and detecting method of live face head pose detection regression apparatus |
CN106920274A (en) * | 2017-01-20 | 2017-07-04 | 南京开为网络科技有限公司 | Mobile terminal 2D key points rapid translating is the human face model building of 3D fusion deformations |
CN107273952A (en) * | 2017-06-15 | 2017-10-20 | 邂逅时尚(大连)科技有限公司 | One kind is accompanied sb. to do shopping machine |
CN107506707A (en) * | 2016-11-30 | 2017-12-22 | 奥瞳系统科技有限公司 | Using the Face datection of the small-scale convolutional neural networks module in embedded system |
-
2017
- 2017-12-26 CN CN201711428451.5A patent/CN108109010A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101369307A (en) * | 2007-08-14 | 2009-02-18 | 索尼株式会社 | Iamge forming device, method and computer program |
CN103079034A (en) * | 2013-01-06 | 2013-05-01 | 北京百度网讯科技有限公司 | Perception shooting method and system |
CN104573619A (en) * | 2014-07-25 | 2015-04-29 | 北京智膜科技有限公司 | Method and system for analyzing big data of intelligent advertisements based on face identification |
CN204559698U (en) * | 2014-12-31 | 2015-08-12 | 上海点图网络科技有限公司 | The process of a kind of image instantaneous acquiring exports integration apparatus |
CN106055193A (en) * | 2016-05-13 | 2016-10-26 | 吴贤波 | Method and system for photo downloading |
CN106355147A (en) * | 2016-08-26 | 2017-01-25 | 张艳 | Acquiring method and detecting method of live face head pose detection regression apparatus |
CN107506707A (en) * | 2016-11-30 | 2017-12-22 | 奥瞳系统科技有限公司 | Using the Face datection of the small-scale convolutional neural networks module in embedded system |
CN106920274A (en) * | 2017-01-20 | 2017-07-04 | 南京开为网络科技有限公司 | Mobile terminal 2D key points rapid translating is the human face model building of 3D fusion deformations |
CN107273952A (en) * | 2017-06-15 | 2017-10-20 | 邂逅时尚(大连)科技有限公司 | One kind is accompanied sb. to do shopping machine |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108960186A (en) * | 2018-07-20 | 2018-12-07 | 南京开为网络科技有限公司 | A kind of advertisement machine user identification method based on face |
CN108960186B (en) * | 2018-07-20 | 2022-07-05 | 南京开为网络科技有限公司 | Advertising machine user identification method based on human face |
CN109191184A (en) * | 2018-08-14 | 2019-01-11 | 微梦创科网络科技(中国)有限公司 | Advertisement placement method and system based on image recognition |
CN108830946A (en) * | 2018-08-31 | 2018-11-16 | 佛山科学技术学院 | A kind of 3 D intelligent AR examination forwarding method and examination transmitting apparatus |
CN111325051A (en) * | 2018-12-13 | 2020-06-23 | 普天信息技术有限公司 | Face recognition method and device based on face image ROI selection |
CN111325051B (en) * | 2018-12-13 | 2023-06-23 | 普天信息技术有限公司 | Face recognition method and device based on face image ROI selection |
CN109615433A (en) * | 2018-12-17 | 2019-04-12 | 海南大学 | Merge the user satisfaction modeling and spacial flex method of adjustment of equitable degree, experience sense and price |
CN109615433B (en) * | 2018-12-17 | 2020-08-04 | 海南大学 | User satisfaction modeling and display space adjusting method fusing fairness, experience feeling and price |
CN109993589A (en) * | 2019-04-16 | 2019-07-09 | 南京海帆数据科技有限公司 | A kind of advertisement machine passenger flow data analysis method based on machine vision |
CN112015292A (en) * | 2019-05-31 | 2020-12-01 | 上海分泽时代软件技术有限公司 | Touch screen operated advertising machine and operation method thereof |
CN110298706A (en) * | 2019-07-04 | 2019-10-01 | 山东浪潮人工智能研究院有限公司 | A kind of exhibition room visits interaction and takes pictures evaluation method |
CN110414419A (en) * | 2019-07-25 | 2019-11-05 | 四川长虹电器股份有限公司 | A kind of posture detecting system and method based on mobile terminal viewer |
CN110490656A (en) * | 2019-08-21 | 2019-11-22 | 深圳市易云数字科技有限责任公司 | The effect quantization method and system of advertisement under a kind of line |
CN113362353A (en) * | 2020-03-04 | 2021-09-07 | 上海分众软件技术有限公司 | Method for identifying advertising player frame by utilizing synthesis training picture |
CN111711747A (en) * | 2020-05-21 | 2020-09-25 | 广州恒德智汇有限公司 | Interactive all-in-one exhibition hall group photo system |
CN113762095A (en) * | 2021-08-18 | 2021-12-07 | 东软集团股份有限公司 | Mask recognition method and device, storage medium and electronic equipment |
CN113807313A (en) * | 2021-10-08 | 2021-12-17 | 合肥安达创展科技股份有限公司 | AI platform analysis system based on Dlib face recognition |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108109010A (en) | A kind of intelligence AR advertisement machines | |
CN111698523B (en) | Method, device, equipment and storage medium for presenting text virtual gift | |
CN107343211B (en) | Method of video image processing, device and terminal device | |
US11217006B2 (en) | Methods and systems for performing 3D simulation based on a 2D video image | |
US11830118B2 (en) | Virtual clothing try-on | |
US10771704B2 (en) | Method of displaying wide-angle image, image display system, and information processing apparatus | |
US20180364803A1 (en) | Methods and Systems for Integrating Virtual Content into an Immersive Virtual Reality World Based on Real-World Scenery | |
US8660319B2 (en) | Method, system and computer program product for automatic and semi-automatic modification of digital images of faces | |
CN108965982A (en) | Video recording method, device, electronic equipment and readable storage medium storing program for executing | |
US8265351B2 (en) | Method, system and computer program product for automatic and semi-automatic modification of digital images of faces | |
WO2018033137A1 (en) | Method, apparatus, and electronic device for displaying service object in video image | |
CN107257338B (en) | media data processing method, device and storage medium | |
WO2009096428A1 (en) | Display system, system for measuring display effect, display method, method for measuring display effect, and recording medium | |
CN107340852A (en) | Gestural control method, device and terminal device | |
US20120299945A1 (en) | Method, system and computer program product for automatic and semi-automatic modificatoin of digital images of faces | |
JP5667061B2 (en) | System and method for scene analysis of multiple viewpoints | |
EP1566788A2 (en) | Display | |
WO2017160370A1 (en) | Visualization of image themes based on image content | |
CN110633664A (en) | Method and device for tracking attention of user based on face recognition technology | |
CN106575450A (en) | Augmented reality content rendering via albedo models, systems and methods | |
CN107343225B (en) | The method, apparatus and terminal device of business object are shown in video image | |
CN108875539B (en) | Expression matching method, device and system and storage medium | |
JP2012504828A (en) | System and method for optimizing a scene | |
CN105763829A (en) | Image processing method and electronic device | |
CN105183477A (en) | System and method for acquiring virtual item information of application program |
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: 20180601 |