CN108830151A - Mask detection method based on gauss hybrid models - Google Patents
Mask detection method based on gauss hybrid models Download PDFInfo
- Publication number
- CN108830151A CN108830151A CN201810426435.0A CN201810426435A CN108830151A CN 108830151 A CN108830151 A CN 108830151A CN 201810426435 A CN201810426435 A CN 201810426435A CN 108830151 A CN108830151 A CN 108830151A
- Authority
- CN
- China
- Prior art keywords
- frame
- face
- hybrid models
- gauss hybrid
- key frame
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/40—Spoof detection, e.g. liveness detection
- G06V40/45—Detection of the body part being alive
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06V40/172—Classification, e.g. identification
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to technical field of computer vision more particularly to a kind of mask detection methods based on gauss hybrid models, include the following steps：Gauss hybrid models are established according to facial image sample；The key frame containing face is screened from video flowing, and face characteristic is extracted from key frame；The face characteristic extracted in key frame feeding gauss hybrid models are matched, judge whether the face in key frame wears mask according to matching result.By using the present invention, following effect may be implemented：Face picture library is classified using gauss hybrid models, can effectively differentiate real human face and mask；It is screened using key frame three times, removes redundant frame, reduce operation, improve detection efficiency.
Description
Technical field
The present invention relates to technical field of computer vision more particularly to a kind of mask detection sides based on gauss hybrid models
Method.
Background technique
With current e-commerce and the fast development of mobile payment, the facial image for obtaining user prevents fraud
A kind of effective means.If user has worn mask during taking pictures, it is likely to bring fraud.However, existing
Human face detection tech cannot differentiate real human face and mask well, therefore, lack a kind of effective mask detection skill at present
Art.
Summary of the invention
To solve the above problems, the present invention proposes a kind of mask detection method based on gauss hybrid models, for detecting
Whether the face in video flowing wears mask.
A kind of mask detection method based on gauss hybrid models, includes the following steps：It is established according to facial image sample
Gauss hybrid models；The key frame containing face is screened from video flowing, and face characteristic is extracted from key frame；By key frame
The face characteristic of middle extraction is sent into gauss hybrid models and is matched, and judges whether the face in key frame wears according to matching result
It wears a mask.
Preferably, the method that face characteristic is extracted from key frame is：It is by the Color Image Processing in key frame
Gray level image；Face datection is carried out to gray level image using human-face detector, human face region is found out in gray level image and is utilized
Rectangle frame is marked；According to the rectangle frame of label, face characteristic is extracted by facial feature points detection algorithm in rectangle frame.
Preferably, the gauss hybrid models are：
Wherein, K represents the quantity of model；πkWeight is represented, is metN(x；μk,∑k) it is mixing
K-th of component of model.
Preferably, the affiliated method that the key frame containing face is screened from video flowing is：Video is extracted from video flowing
Frame；The redundant frame that face is not contained in video frame is filtered out using human-face detector；Repetition is filtered out from the video frame containing face
Redundant frame, obtain key frame.
Preferably, described that duplicate redundant frame is filtered out from the video frame containing face, the method for obtaining key frame is：It obtains
The characteristic value of each frame in key frame is taken, and characteristic value substitution following formula is acquired to the phase of face characteristic in two adjacent frames
Like degree
Wherein, xijRepresent the jth dimensional feature value of the i-th frame, x(i+1)jRepresent the jth dimensional feature value of i+1 frame；Threshold value is set
Tf, whenWhen, then i+1 frame is redundant frame, deletes i+1 frame；Otherwise, retain i+1 frame.
Preferably, the method that whether face judged in key frame wears mask is：The people that will be extracted in key frame
Face feature is sent into after gauss hybrid models are matched and obtains several probability density, and the maximum value in several probability density is acquired
Pmax；Empirical value T is set, if Pmax> T, then be judged as and do not wear mask；Otherwise, then it is judged as and has worn mask.
By using the present invention, following effect may be implemented：
1. the present invention is classified face picture library using gauss hybrid models, real human face and face can be effectively differentiated
Tool；
2. screening using key frame three times, redundant frame is removed, reduces operation, improves detection efficiency.
Detailed description of the invention
The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.
Fig. 1 is flow diagram of the invention.
Specific embodiment
Below in conjunction with attached drawing, technical scheme of the present invention will be further described, but the present invention is not limited to these realities
Apply example.
Basic thought the present invention is based on the mask detection method of gauss hybrid models is established according to facial image sample
Gauss hybrid models；Repeatedly screening obtains the key frame containing face from video flowing, and face characteristic is extracted from key frame；
The face characteristic extracted in key frame feeding gauss hybrid models are matched, the people in key frame is judged according to matching result
Whether face wears mask, to realize quick, effective discrimination of real human face and mask.
Fig. 1 is flow diagram of the invention, according to Fig. 1, it can be seen that the invention mainly comprises following steps.
Step 1 establishes gauss hybrid models according to facial image sample.
Specifically, in the present embodiment, present invention employs Chinese Academy of Sciences's face databases, 10000 people are therefrom had chosen
Face image is as training set, including different genders, age, head pose, expression etc..Firstly, being ash by Color Image Processing
Spend image.Secondly, carrying out Face datection to gray level image using human-face detector, human face region is found out in a gray level image
And it is marked using rectangle frame.Usually in an image, the ratio that face accounts for whole image is uncertain, utilizes Face datection
Device is marked, and is conducive to extract face characteristic.Finally, being mentioned in the picture by facial feature points detection algorithm according to label
Take face characteristic.
Gauss hybrid models are established according to the face characteristic of extraction.In the present embodiment, gauss hybrid models are：
Wherein, K represents the quantity of model；πkWeight is represented, is metN(x；μk,Σk) it is mixed
K-th of component of molding type.
Gauss hybrid models (GMM), which just refer to, estimates the probability density distribution of sample, and estimates the model used
It is the weighted sum of several Gauss models.Each Gauss model just represents a class.To the data in sample respectively several high
It is projected on this model, the probability in each class will be respectively obtained.Then we can choose the class of maximum probability and are
Court verdict.In the present embodiment, for established gauss hybrid models, after the completion of these face characteristic data clusters, meeting
It is divided into multiclass, for example man is divided into one kind, woman is divided into one kind, and child is divided into one kind, and older is divided into one kind, face circle
People is divided into one kind etc..By every a kind of image, the sub-belt energy feature of corresponding one group two layers of discrete wavelet transformation obtains multiple groups spy
Vector is levied, and the more of corresponding multiple groups feature vector are respectively trained using EM algorithm using the multiple groups feature vector as training sample
A mixed Gauss model.Face characteristic data are finely divided as multiclass using gauss hybrid models, so that final discrimination knot
Fruit is more accurate.
Step 2 screens the key frame containing face from video flowing, and face characteristic is extracted from key frame.
Specifically, in the present embodiment, to reduce the operation of model and improving detection efficiency, to the video frame in video flowing
It is screened three times, obtains the key frame for containing only face.
It screens for the first time：According to n frame/s decimation in frequency video frame from video flowing, video frame is obtained, n generally takes 1.?
Certain decimation in frequency video frame is pressed in video flowing, and guarantees that the face occurred in video flowing is not omitted.
Programmed screening：The redundant frame that face is not contained in video frame is filtered out using human-face detector.Due in video flowing
In the presence of much not no scenes of face, so can also there is the video frame of many not faces in the extraction process of video frame.
Face is not present in this partial video frame, so not needing to carry out mask detection to it.This part is not present in the present embodiment
The redundant frame of face is screened by human-face detector, to reduce the operation of gauss hybrid models, improves true people
The efficiency that face and mask distinguish.Since there are errors in identification process for human-face detector, so the video after primary screening
There is likely to be the video frames of not face in frame, in practical screening process, can repeatedly be screened by human-face detector.
Third time is screened：Duplicate redundant frame is filtered out from the video frame containing face, obtains key frame.Specifically, obtaining
The characteristic value of each frame in N frame key frame is taken, and characteristic value substitution following formula is acquired into face characteristic in two adjacent frames
Similarity degree
Wherein, xijRepresent the jth dimensional feature value of the i-th frame, x(i+1)jRepresent the jth dimensional feature value of i+1 frame；Threshold value is set
Tf, whenWhen, then i+1 frame is redundant frame, deletes i+1 frame；Otherwise, retain i+1 frame.After programmed screening
Video frame in there are the video frames of same similar scene, so do not need to duplicate video frame carry out mask detection.At this
In embodiment, the characteristic value that former and later two video frames carry out 128 dimensions is compared, if obtained result is less than the threshold value of setting,
Judge that the two video frames are excessively similar, determines that a later frame is redundant frame.Due to not needing to carry out operation to similar redundant frame,
Improve the efficiency of real human face and mask discrimination.In the present embodiment, characteristic value comparison is carried out by former and later two video frames
Determine whether redundant frame, but applicant can also use other figures not being defined to the decision procedure of redundant frame
As processing method determines it.
The secondary static picture of one frame i.e. one.In the present embodiment, extracted from key frame face characteristic with from facial image
The method that face characteristic is extracted in sample is identical, and explanation is not repeated herein.
Step 3 matches the face characteristic extracted in key frame feeding gauss hybrid models, according to matching result
Judge whether the face in key frame wears mask.
Specifically, the face characteristic extracted in key frame is sent into gauss hybrid models, in the cluster of each model
The heart carries out distance metric and obtains several probability density, acquires the maximum value P in several probability densitymax；Empirical value T is set,
In the present embodiment, T=96.If Pmax> T, then be judged as and do not wear mask；Otherwise, then it is judged as and has worn mask.
Those skilled in the art can make various modifications to described specific embodiment
Or supplement or be substituted in a similar manner, however, it does not deviate from the spirit of the invention or surmounts the appended claims determines
The range of justice.
Claims (6)
1. a kind of mask detection method based on gauss hybrid models, which is characterized in that include the following steps：
Gauss hybrid models are established according to facial image sample；
The key frame containing face is screened from video flowing, and face characteristic is extracted from key frame；
The face characteristic extracted in key frame feeding gauss hybrid models are matched, are judged in key frame according to matching result
Face whether wear mask.
2. the mask detection method according to claim 1 based on gauss hybrid models, which is characterized in that described from key
The method of extraction face characteristic is in frame：
It is gray level image by the Color Image Processing in key frame；
Face datection is carried out to gray level image using human-face detector, human face region is found out in gray level image and utilizes rectangle frame
It is marked；
According to the rectangle frame of label, face characteristic is extracted by facial feature points detection algorithm in rectangle frame.
3. the mask detection method according to claim 1 based on gauss hybrid models, which is characterized in that the Gauss is mixed
Molding type is：
Wherein, K represents the quantity of model；πkWeight is represented, is metN(x；μk,∑k) it is mixed model
K-th of component.
4. the mask detection method according to claim 1 based on gauss hybrid models, which is characterized in that affiliated from video
The method of key frame of the screening containing face is in stream：
Video frame is extracted from video flowing；
The redundant frame that face is not contained in video frame is filtered out using human-face detector；
Duplicate redundant frame is filtered out from the video frame containing face, obtains key frame.
5. the mask detection method according to claim 4 based on gauss hybrid models, which is characterized in that described from containing
Duplicate redundant frame is filtered out in the video frame of face, the method for obtaining key frame is：
The characteristic value of each frame in key frame is obtained, and characteristic value substitution following formula is acquired into face spy in two adjacent frames
The similarity degree of sign
Wherein, xijRepresent the jth dimensional feature value of the i-th frame, x(i+1)jRepresent the jth dimensional feature value of i+1 frame；
Threshold value T is setf, whenWhen, then i+1 frame is redundant frame, deletes i+1 frame；Otherwise, retain i+1 frame.
6. the mask detection method according to claim 1 based on gauss hybrid models, which is characterized in that the judgement is closed
The method whether face in key frame wears mask is：The face characteristic extracted in key frame is sent into gauss hybrid models to carry out
Several probability density are obtained after matching, acquire the maximum value P in several probability densitymax；Empirical value T is set, if Pmax>
T is then judged as and does not wear mask；Otherwise, then it is judged as and has worn mask.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810426435.0A CN108830151A (en) | 2018-05-07 | 2018-05-07 | Mask detection method based on gauss hybrid models |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810426435.0A CN108830151A (en) | 2018-05-07 | 2018-05-07 | Mask detection method based on gauss hybrid models |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108830151A true CN108830151A (en) | 2018-11-16 |
Family
ID=64147636
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810426435.0A Pending CN108830151A (en) | 2018-05-07 | 2018-05-07 | Mask detection method based on gauss hybrid models |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108830151A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110879972A (en) * | 2019-10-24 | 2020-03-13 | 深圳云天励飞技术有限公司 | Face detection method and device |
| CN112650461A (en) * | 2020-12-15 | 2021-04-13 | 广州舒勇五金制品有限公司 | Relative position-based display system |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101464950A (en) * | 2009-01-16 | 2009-06-24 | 北京航空航天大学 | Video human face identification and retrieval method based on on-line learning and Bayesian inference |
| CN104361326A (en) * | 2014-11-18 | 2015-02-18 | 新开普电子股份有限公司 | Method for distinguishing living human face |
| CN105426515A (en) * | 2015-12-01 | 2016-03-23 | 小米科技有限责任公司 | Video classification method and apparatus |
| CN106446772A (en) * | 2016-08-11 | 2017-02-22 | 天津大学 | Cheating-prevention method in face recognition system |
| CN107194985A (en) * | 2017-04-11 | 2017-09-22 | 中国农业大学 | A kind of three-dimensional visualization method and device towards large scene |
| CN107194376A (en) * | 2017-06-21 | 2017-09-22 | 北京市威富安防科技有限公司 | Mask fraud convolutional neural networks training method and human face in-vivo detection method |
| CN107844779A (en) * | 2017-11-21 | 2018-03-27 | 重庆邮电大学 | A kind of video key frame extracting method |
-
2018
- 2018-05-07 CN CN201810426435.0A patent/CN108830151A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101464950A (en) * | 2009-01-16 | 2009-06-24 | 北京航空航天大学 | Video human face identification and retrieval method based on on-line learning and Bayesian inference |
| CN104361326A (en) * | 2014-11-18 | 2015-02-18 | 新开普电子股份有限公司 | Method for distinguishing living human face |
| CN105426515A (en) * | 2015-12-01 | 2016-03-23 | 小米科技有限责任公司 | Video classification method and apparatus |
| CN106446772A (en) * | 2016-08-11 | 2017-02-22 | 天津大学 | Cheating-prevention method in face recognition system |
| CN107194985A (en) * | 2017-04-11 | 2017-09-22 | 中国农业大学 | A kind of three-dimensional visualization method and device towards large scene |
| CN107194376A (en) * | 2017-06-21 | 2017-09-22 | 北京市威富安防科技有限公司 | Mask fraud convolutional neural networks training method and human face in-vivo detection method |
| CN107844779A (en) * | 2017-11-21 | 2018-03-27 | 重庆邮电大学 | A kind of video key frame extracting method |
Non-Patent Citations (4)
| Title |
|---|
| GIRIJA CHETTY ET AL: "Liveness Verification in Audio-Video Speaker Authentication", 《10TH AUSTRALIAN INTERNATIONAL CONFERENCE ON SPEECH SCIENCE & TECHNOLOGY》 * |
| 刘伟锋等: "基于Gabor特征和混合高斯模型的人脸表情分析", 《计算机工程与应用》 * |
| 唐旭: "基于高斯混合模型分类的SAR图像检索", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 屈有佳: "基于SIFT特征的关键帧提取算法研究", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110879972A (en) * | 2019-10-24 | 2020-03-13 | 深圳云天励飞技术有限公司 | Face detection method and device |
| CN112650461A (en) * | 2020-12-15 | 2021-04-13 | 广州舒勇五金制品有限公司 | Relative position-based display system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yang et al. | Exposing GAN-synthesized faces using landmark locations | |
| Huang et al. | Tracknet: A deep learning network for tracking high-speed and tiny objects in sports applications | |
| Li et al. | Multiple-human parsing in the wild | |
| Agarwal et al. | Swapped! digital face presentation attack detection via weighted local magnitude pattern | |
| Luo et al. | Group sparsity and geometry constrained dictionary learning for action recognition from depth maps | |
| CN109815826B (en) | Method and device for generating face attribute model | |
| Wang et al. | Human action recognition by semilatent topic models | |
| Vrigkas et al. | Matching mixtures of curves for human action recognition | |
| CN109815874A (en) | A kind of personnel identity recognition methods, device, equipment and readable storage medium storing program for executing | |
| CN111401144A (en) | Escalator passenger behavior identification method based on video monitoring | |
| CN102034107B (en) | Unhealthy image differentiating method based on robust visual attention feature and sparse representation | |
| Winarno et al. | Multi-view faces detection using Viola-Jones method | |
| Zhu et al. | Action recognition in broadcast tennis video using optical flow and support vector machine | |
| CN106845456A (en) | A kind of method of falling over of human body monitoring in video monitoring system | |
| CN108830151A (en) | Mask detection method based on gauss hybrid models | |
| Mallet et al. | Using deep learning to detecting deepfakes | |
| CN101950448A (en) | Detection method and system for masquerade and peep behaviors before ATM (Automatic Teller Machine) | |
| CN108681928A (en) | A kind of intelligent advertisement put-on method | |
| CN104680189A (en) | Pornographic image detection method based on improved bag-of-words model | |
| Gunay et al. | Facial age estimation based on decision level fusion of amm, lbp and gabor features | |
| CN110008876A (en) | A kind of face verification method based on data enhancing and Fusion Features | |
| CN103971100A (en) | Video-based camouflage and peeping behavior detection method for automated teller machine | |
| Siddiquie et al. | Recognizing plays in american football videos | |
| Hu et al. | CoverTheFace: face covering monitoring and demonstrating using deep learning and statistical shape analysis | |
| Panicker et al. | Cardio-pulmonary resuscitation (CPR) scene retrieval from medical simulation videos using local binary patterns over three orthogonal planes |
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: 20181116 |