CN106295522A - A kind of two-stage anti-fraud detection method based on multi-orientation Face and environmental information - Google Patents
A kind of two-stage anti-fraud detection method based on multi-orientation Face and environmental information Download PDFInfo
- Publication number
- CN106295522A CN106295522A CN201610614086.6A CN201610614086A CN106295522A CN 106295522 A CN106295522 A CN 106295522A CN 201610614086 A CN201610614086 A CN 201610614086A CN 106295522 A CN106295522 A CN 106295522A
- Authority
- CN
- China
- Prior art keywords
- face
- background
- orientation
- detected
- fraud detection
- 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.)
- Granted
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 48
- 230000007613 environmental effect Effects 0.000 title claims abstract description 21
- 230000001815 facial effect Effects 0.000 claims abstract description 29
- 238000002372 labelling Methods 0.000 claims 1
- 238000000034 method Methods 0.000 description 23
- 238000012549 training Methods 0.000 description 11
- 238000001727 in vivo Methods 0.000 description 8
- 210000000887 face Anatomy 0.000 description 6
- 210000003128 head Anatomy 0.000 description 6
- 238000012360 testing method Methods 0.000 description 5
- 238000001574 biopsy Methods 0.000 description 3
- 230000004397 blinking Effects 0.000 description 2
- 230000009977 dual effect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000002452 interceptive effect Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 206010000210 abortion Diseases 0.000 description 1
- 231100000176 abortion Toxicity 0.000 description 1
- 238000013475 authorization Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 230000004886 head movement Effects 0.000 description 1
- 231100000518 lethal Toxicity 0.000 description 1
- 230000001665 lethal effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000013011 mating Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000007 visual effect Effects 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
- G06V40/10—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
- G06V40/172—Classification, e.g. identification
-
- G—PHYSICS
- G07—CHECKING-DEVICES
- G07C—TIME OR ATTENDANCE REGISTERS; REGISTERING OR INDICATING THE WORKING OF MACHINES; GENERATING RANDOM NUMBERS; VOTING OR LOTTERY APPARATUS; ARRANGEMENTS, SYSTEMS OR APPARATUS FOR CHECKING NOT PROVIDED FOR ELSEWHERE
- G07C1/00—Registering, indicating or recording the time of events or elapsed time, e.g. time-recorders for work people
- G07C1/10—Registering, indicating or recording the time of events or elapsed time, e.g. time-recorders for work people together with the recording, indicating or registering of other data, e.g. of signs of identity
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of two-stage anti-fraud detection method based on multi-orientation Face and environmental information, comprise the following steps: the face in detection region is tracked, face is carried out multi-orientation detection, the facial image angularly key words sorting detected is stored;Background contrast region is determined according to the face location detected;The background under the background of current face and unmanned environment is contrasted in background contrast region;Judge whether the face that detects meets multi-orientation Face condition, it is judged that in background contrast region, whether the matching degree of the background under the background of current face and unmanned environment meets condition;Two conditions are then judged to when being satisfied by that currently detected face is real human face, otherwise judge that currently detected face is as human face photo or video;Store the positive face image in the multi-orientation Face detected.On Cao Zuo more friendly to user, can rapidly and efficiently carry out face anti-fraud detection, the safety and reliability for face identification system provides safeguard.
Description
Technical field
The present invention relates to technical field of face recognition, be specifically related to a kind of based on multi-orientation Face with the two-stage of environmental information
Anti-fraud detection method.
Background technology
In all of biometrics identification technology, recognition of face is a kind of biometric identity the most natural, visual
Recognition method, meets the physiological vision custom of mankind itself, and mankind itself is through face to differentiate people's.In recent years,
Face recognition technology develops rapidly and has obtained increasingly being widely applied, and the safety of recognition of face is also by more and more therewith
Many concerns, spoofing attack is then the lethal challenge of face identification system.Through the development of decades, in field of face identification
In, scholars have been proposed for the technology of a large amount of In vivo detection and method is applied to enterprise's work attendance recognition of face that user capacity is little
In system, and the colleges and universities that place big for user capacity, that flow of the people is many, such as course are numerous, student's capacity is big, also do not propose
A kind of preferably anti-fraud biopsy method.
Under normal conditions, face identification system be faced with three kinds of frauds:
1. use the photo of validated user;
2. use the video of validated user;
3. use the threedimensional model of validated user.
Using validated user photograph to carry out spoofing attack is the most cheap and convenient means, because human face photo is very easy to
Obtain, such as surf the Net download, photographic head shooting etc..Video deception can present the live body movement effects identical with being really face,
Such as motion of blinking, head movement, expression motion etc..And relative first two fraud, just manufacture the threedimensional model of a face
Appear to much more complex, and the case that currently used this mode carries out spoofing attack to face identification system be the most fewer,
So we the most temporarily cheat, just for photo and video, research of discussing.
Application publication number is " the interactive video biopsy method controlled based on human face posture of CN 105260726 A
And system " proposing a kind of interactive video identification system, it is when carrying out Face datection, and user is required completion system institute
Point out actions such as blinking or open one's mouth, use photo to carry out the user played tricks to filter.This system work attendance is applied significantly to carry
The cost of high equipment, the In vivo detection long period to be spent to everyone, and strong depend-ence user coordinate, and are unfavorable for that user holds
The place that amount is big, flow of the people is many is suitable for;Authorization Notice No. is the " a kind of based on the motion of face physiological of CN 100514353 C
In-vivo detection method and system " propose a kind of In vivo detection system, this system is by the face physiological in detection region
Real human face and photo face are distinguished in movable detection, and if this system is carried out video spoofing attack, just can crack it
Biopsy method, allows system become unreliable;Application publication number is " a kind of face In vivo detection side of CN 103886301 A
Method " in face In vivo detection grader be required for particular person and be individually trained, the corresponding people of the most each accredited personnel
Face live body grader, this mode is only applicable to the less enterprise's work attendance of user capacity or the work attendance of special sector equally, when with
When family amount is excessive, it is clear that be the most uneconomic to everyone one live body grader of independent training.
Summary of the invention
The technical problem to be solved in the present invention is, in view of the foregoing defects the prior art has, it is provided that a kind of based on many
The two-stage anti-fraud detection method of angle face and environmental information, more friendly to user in operation, can rapidly and efficiently carry out
Face anti-fraud detects, and the safety and reliability for face identification system provides safeguard..
The present invention solves that above-mentioned technical problem be the technical scheme is that
A kind of two-stage anti-fraud detection method based on multi-orientation Face and environmental information, comprises the following steps:
1) face in detection region is tracked, face is carried out multi-orientation detection, the facial image that will detect
Angularly key words sorting stores;
2) background contrast region is determined according to the face location detected;
3) in background contrast region, contrast the background under the background of current face and unmanned environment;
4) judge whether the face detected meets multi-orientation Face condition, it is judged that current face in background contrast region
Whether the matching degree of the background under background and unmanned environment meets condition;
5) when step 4) in two conditions be then judged to that currently detected face is real human face when being satisfied by, enter
To next step, otherwise judge that currently detected face is as human face photo or video;
6) the positive face image in the multi-orientation Face detected is stored.
Connect technique scheme, in described step 1) in, it is to use Haar-like feature that face carries out multi-orientation detection
Face is carried out with each two field picture of the many classification Adaboost algorithm adjusted based on the Weak Classifier face video to detecting
Detection.
Connect technique scheme, in step 1) in, angularly facial image is divided three classes, including positive face type, half side
Face type and full side face type, positive face type includes the scope with face front for zero point rotation of the coordinate system [-20 °, 20 °], half side
Face type includes the scope with face front for zero point rotation of the coordinate system [-60 ° ,-20 °) and (20 °, 60 °], and full side face type includes
With face front for the scope of zero point rotation of the coordinate system [-90 ° ,-60 °) and (60 °, 90 °].
Connecing technique scheme, the facial image detected is stored in a reservoir by Key-Value, the people in the range of positive face
The value of face image is labeled as 1, and the value of the facial image in the range of half side face is labeled as 0, the face in the range of the face of full side
The value of image is labeled as-1.
Connect technique scheme, described step 4) in, when detecting same face, in storage container, value is 1,0 ,-1 successively
Value time, then judge detect that face meets multi-orientation Face condition.
Connect technique scheme, in described step 6) in, it is facial image when 1 that positive face image is value, works as judgement
When the face detected is real human face, by positive face image storage and send into identification system.
Connect technique scheme, described step 3) in, background contrast region A is positioned at for the face left and right sides, background contrast
Region does not comprise the region of ear and hair, and the height in background contrast region is identical with the height of face, and face width is d,
The width of the background contrast region A of the face left and right sides is 1/8d~1/3d, background contrast region A and homonymy face edge away from
From for 1/8d~1/3d.
Connect technique scheme, described step 4) in, in background contrast region under the background of current face and unmanned environment
When background the match is successful rate is more than 70%, it is judged that for the match is successful, meeting background environment matching condition, otherwise it fails to match,
It is unsatisfactory for background environment matching condition.
The method have the advantages that
The method that the present invention provides, is to prevent modal photo spoofing attack and the dual fail-safe of video spoofing attack, uses
Family is walked close to turn round, after photographic head prepares detection and detected, the action left, and multi-orientation Face just can be detected, so this
The method of kind is not required to user and coordinates, the most friendly to user, and background contrast also makes face anti-fraud detection work accurately and efficiently
Can be quickly completed, be applicable not only to enterprise's work attendance that user capacity is little, be more suitable for user capacity big time work attendance, by multi-angle
Face datection and background contrast combine, and just can obtain a kind of quick, efficient, to the detection of user-friendly face anti-fraud
Method.
Accompanying drawing explanation
Fig. 1 is the flow process of two-stage anti-fraud detection method based on multi-orientation Face and environmental information in the embodiment of the present invention
Figure;
Fig. 2 is multi-orientation Face sample graph in the embodiment of the present invention;
Fig. 3 is background contrast region in the embodiment of the present invention;
Detailed description of the invention
The present invention is described in detail with embodiment below in conjunction with the accompanying drawings.
With reference to shown in Fig. 1, in the embodiment that the present invention provides based on multi-orientation Face and the two-stage of environmental information
Anti-fraud detection method, comprises the following steps:
1) face in detection region is tracked, face is carried out multi-orientation detection, the facial image that will detect
Angularly key words sorting stores;
2) background contrast region is determined according to the face location detected;
3) in background contrast region, contrast the background under the background of current face and unmanned environment;
4) judge whether the face detected meets multi-orientation Face condition, it is judged that current face in background contrast region
Whether the matching rate of the background under background and unmanned environment meets condition;
5) when step 4) in two conditions be then judged to that currently detected face is real human face when being satisfied by, enter
To next step, otherwise judge that currently detected face is as human face photo or video;
6) the positive face image in the multi-orientation Face detected is stored.
Further, in described step 1) in, face is carried out multi-orientation detection be use Haar-like feature and based on
Each two field picture of the many classification Adaboost algorithm that Weak Classifier the adjusts face video to detecting carries out Face datection.
Further, described facial image is the face video sub-frame processing detected to be formed.
Further, in step 1) in, angularly facial image is divided three classes, including positive face type, half side face type
With full side face type, positive face type includes the scope with face front for zero point rotation of the coordinate system [-20 °, 20 °], half side face type
Including for face front for the scope of zero point rotation of the coordinate system [-60 ° ,-20 °) and (20 °, 60 °], full side face type includes with people
Face front is the scope of zero point rotation of the coordinate system [-90 ° ,-60 °) and (60 °, 90 °], as in figure 2 it is shown, the first positive face of behavior in Fig. 2
Type sample, the second behavior half side face type sample, the third line is full side face type sample.
Further, the facial image detected is stored in a reservoir by Key-Value, the facial image in the range of positive face
Value be labeled as 1, the value of the facial image in the range of half side face is labeled as 0, the facial image in the range of the face of full side
Value is labeled as-1;During whole detection, the face of user will be tracked, to prevent the midway user people in certification
Face is switched.
Further, described step 4) in, when detecting same face, in storage container, occurrence is 1,0 ,-1 successively
During value, then judge to detect that face meets multi-orientation Face condition, storage container occurs successively show that face is continuous print
Rather than interrupt, also it is as the foundation determining whether real human face.
Further, in described step 6) in, it is facial image when 1 that positive face image is value, when judging to detect
Face when being real human face, by positive face image storage and send into identification system.
Further, described step 3) in, as it is shown on figure 3, background contrast region A is positioned at for the face left and right sides, background
Contrast district does not comprise the region of ear and hair, and the height in background contrast region is identical with the height of face, background contrast district
The left and right sides width of territory A is the 1/6 of face width d, and background contrast region A is face width d with the distance at face edge
1/6.
Further, described step 4) in, in background contrast region under current face's background and unmanned environment background
When being made into power more than 70%, it is judged that for the match is successful, meeting background environment matching condition, otherwise it fails to match, discontented instep
Scape environments match condition.
Further, described step 4) in, the background of current face and the background under unmanned environment in background contrast region
Rate that the match is successful, refers in multiple facial images currently detected under the face background of individual facial image and unmanned environment
Background is in background contrast region, and the quantity that the match is successful accounting in the sum of current face's image is that the match is successful
Rate.
Further, in individual facial image the matching degree of the background under the background of face and unmanned environment more than 70%
Time, i.e. think that the Background matching of individual facial image is successful.
Further, extract the background area under current face background area and unmanned environment in background contrast region, will
Rectangular histogram in two regions extracted is mated, when mating numerical value more than 0.7, it is judged that for the match is successful, otherwise mate
Failure.
In one embodiment of the present of invention, the operation principle of the present invention:
Face in detection region is tracked, and detects the multi-orientation Face of user:
For each two field picture of input video, use Haar similar characteristics and the many classification adjusted based on Weak Classifier
Adaboost algorithm detects face and to the face classification detected.
Adaboost algorithm is that Paul Viola and Michael Jones proposed in calendar year 2001.It is that a kind of iteration is calculated
Method, its basic thought is to train same grader (Weak Classifier), then these different training for different training sets
Obtain grader on collection to gather, constitute a final strong classifier.
Adaboost algorithm requires that Weak Classifier accuracy is more than 1/2.But in many classification problems, due to random guess
Accuracy only has 1/k, and compared with two class AdaBoost algorithms, many classification AdaBoost algorithms are difficult to find that accuracy is more than 1/2
Weak Classifier, the Weak Classifier number that obtains in actual applications may be caused not enough, it is impossible to ensure to obtain the best strong
Grader.
The many classification Adaboost algorithm adjusted based on Weak Classifier proposes: if both can be to Weak Classifier accuracy
Requirement be reduced to more than 1/k, low-quality Weak Classifier can be generated by limiting again, ensure that and obtain sufficiently strong strong classification
Device.
When training Weak Classifier, it is judged that the result of this Weak Classifier, in the classification of all samples belonging to a class together, just
Really the sample of classification weights and whether score to other any type weights with greatly.If meeting this condition, proceed
Weighed value adjusting and next iteration.If be unsatisfactory for, then the Weak Classifier being likely due to train is the best, can be at weights
Re-training Weak Classifier in the case of constant, judges whether new Weak Classifier meets the condition described in top the most again,
If meet entrance to adjust next time, it is unsatisfactory for then re-training Weak Classifier.Through the restriction of this condition, when T levels off to nothing
Time poor, the certain score of correct label of classifying is many to the quantity of other kinds label, and the error rate of final grader levels off to
0。
Assume m sample sequence { (x1,y1),(x2,y2),…,(xm,ym), wherein yiFor classification, and yi∈ Y={1,
2 ..., K}, Weak Classifier ht(x).Foryi=P, P ∈ 1,2 ..., K}, Weak Classifier htX (), at P class sample
In this classification, assigning to all kinds of probability is { p1,p2,…,pK, if there being a=P, then require pa>pb,b∈{1,2,…,a-1,a+
1,…,K}.If being unsatisfactory for this condition, then re-training htX (), if until training classification htX () meets above-mentioned condition.
The face type detected is positive face, half side face or full side face by the present invention.Wherein positive face type includes with face
Front is the scope of zero point rotation of the coordinate system [-20 °, 20 °], half side face type include with face front for zero point rotation of the coordinate system [-
60 ° ,-20 °) and (20 °, 60 °] scope, full side face type include with face front for zero point rotation of the coordinate system [-90 ° ,-60 °)
(60 °, 90 °] scope.
Training grader specific requirement is as follows:
1) the true man's facial image in positive example sample (including positive face collection, half side face collection, full side face collection) should be as much as possible
Contain and block with or without eyes and the changing factor at sex, age etc.;
2) firstly the need of setting up a positive face storehouse, a half side face storehouse and face storehouse, a full side.Use camera collection respectively
30 boy students and 30 schoolgirl's face head portraits, everyone corresponding 30 positive faces, 20 half side faces in a left side, 20 half side faces in the right side, 20
Left full side face, full side, 20 right sides face.Picking rate is 20fps, and image resolution ratio is 640 pixel × 480 pixels;
3) respectively from network and with 5000 non-face negative datas under camera collection varying environment;
4) according to eye location, all images are converted into the Gray Face image of 24 × 24 pixels;
5) facial image (600 positive faces, 800 half side faces, 800 full side faces) of 20 people and 3500 are randomly selected
Non-face image is as training set, and remaining facial image and non-face image are all as test set;
6) according to above-mentioned many classification Adaboost algorithm one grader of training adjusted based on Weak Classifier, it is used for just distinguishing
Face, half side face, full side face and non-face;
7) in order to detect the effect of said method, testing grader with test set, test result is as shown in table 1.
Table 1 multi-orientation Face testing result
Classification | Positive face | Half side face | Full side face | Non-face |
Type judgment accuracy | 95.62% | 86.34% | 90.45% | 100.0% |
After face being detected and classifying, the facial image detected is stored in a reservoir by Key-Value, positive face
Value be 1, the value of half side face is 0, and the value of full side face is-1.
It is necessary to carry out current face's background and the contrast of unmanned environmental background after face being detected:
Traditional method be by Face datection before environmental background preserve, then with detection time face background carry out right
Ratio.
But the place that, environmental background a lot of at flow of the people is easily changed, university researchers as big in student's capacity, face is examined
Survey environment when carrying out and be likely occurred large change compared to the environment just having opened photographic head before Face datection, as detected
Time user have people's process, empty environment to become the environment etc. occupying artificial abortion behind.At this moment a real human face is arisen that
Judge into the situation of video human face by accident, because the real background of real human face differs with environmental background when just opening photographic head
Sample, then background contrast will be failed.
Contrast district is the region that the face left and right sides does not comprise hair, and as shown in the a-quadrant of Fig. 3, width is 1/6d, model
Enclosing for [rect.x-1/6d*2, rect.x-1/6d] and [rect.x+d+1/6d, rect.x+d+1/6d*2], height be face height
Degree rect.Height, scope is [rect.y, rect.y+h], and wherein rect is rectangular area face being detected, and d is detection
The width rect.width of the face rectangular area rect arrived.The height of the h face rectangular area rect for detecting is
rect.height。
Then by the contrast district, left and right of face background and and the rectangular histogram of contrast district, left and right of environmental background carry out right
Ratio, more than 0.7 after the Histogram Matching coefficient in [left region-left region] and [right region-right region] weights, then background contrast
Success, otherwise background contrast's failure.
When the multi-orientation Face of a people being detected, i.e. store and the container of image occurring, value is 1,0 ,-1 successively
During image key, system judges that this is the dynamic human face (living body faces or video human face) of multi-orientation Face to be detected rather than quiet
State photo face.
Meanwhile, when multi-orientation Face being detected, it is calculated background contrast's success rate of these multi-orientation Faces, when this
When background contrast's success rate of a little frame faces is more than 70%, represent environments match success, detect be not video human face and
It it is living body faces.
In sum, face In vivo detection is the indispensable part of face identification system, the property of face In vivo detection
Can the quality of energy, decide face identification system and move towards actual application from theoretical research, and the method that the present invention provides, is to prevent
Modal photo spoofing attack and the dual fail-safe of video spoofing attack, user walks close to photographic head and prepares detection and detected
After turn round the action left, multi-orientation Face just can be detected, be in this way not required to user coordinate, the most friendly to user
Good.And background contrast also makes face anti-fraud detection work to be quickly completed accurately and efficiently, it is applicable not only to user capacity little
Enterprise's work attendance, be more suitable for user capacity big time work attendance.
The above presently preferred embodiments of the present invention that is only, can not limit the interest field of the present invention with this certainly,
Therefore the equivalence change made according to scope of the present invention patent, still belongs to protection scope of the present invention.
Claims (8)
1. a two-stage anti-fraud detection method based on multi-orientation Face and environmental information, it is characterised in that include following step
Rapid:
1) face in detection region is tracked, face is carried out multi-orientation detection, by the facial image that detects by angle
Degree key words sorting stores;
2) background contrast region is determined according to the face location detected;
3) in background contrast region, contrast the background under the background of current face and unmanned environment;
4) judge whether the face detected meets multi-orientation Face condition, it is judged that the background of current face in background contrast region
The match is successful whether rate meets background environment matching condition with the background under unmanned environment;
5) when step 4) in two conditions be then judged to that currently detected face is real human face when being satisfied by, enter into down
One step, otherwise judges that currently detected face is as human face photo or video;
6) the positive face image in the multi-orientation Face detected is stored.
Two-stage anti-fraud detection method based on multi-orientation Face and environmental information the most according to claim 1, its feature
It is, in described step 1) in, it is to use Haar-like feature and adjust based on Weak Classifier that face carries out multi-orientation detection
Each two field picture of many classification Adaboost algorithm face video to detecting carry out Face datection.
Two-stage anti-fraud detection method based on multi-orientation Face and environmental information the most according to claim 1, its feature
It is, in step 1) in, angularly facial image is divided three classes, including positive face type, half side face type and full side face type,
Positive face type includes the scope with face front for zero point rotation of the coordinate system [-20 °, 20 °], and half side face type includes with face front
For the scope of zero point rotation of the coordinate system [-60 ° ,-20 °) and (20 °, 60 °], full side face type includes with face front for zero point coordinate
Rotate the scope of [-90 ° ,-60 °) and (60 °, 90 °].
Two-stage anti-fraud detection method based on multi-orientation Face and environmental information the most according to claim 3, its feature
Being, the facial image detected is stored in a reservoir by Key-Value, the value labelling of the facial image in the range of positive face
Being 1, the value of the facial image in the range of half side face is labeled as 0, and the value of the facial image in the range of the face of full side is labeled as-
1。
Two-stage anti-fraud detection method based on multi-orientation Face and environmental information the most according to claim 4, its feature
It is, described step 4) in, when detecting same face, when storing in container the value that occurrence successively is 1,0 ,-1, then sentence
Break and detect that face meets multi-orientation Face condition.
Two-stage anti-fraud detection method based on multi-orientation Face and environmental information the most according to claim 4, its feature
It is, in described step 6) in, it is facial image when 1 that positive face image is value, when judging that the face detected is as true
During face, by positive face image storage and send into identification system.
Two-stage anti-fraud detection method based on multi-orientation Face and environmental information the most according to claim 1, its feature
It is, described step 3) in, background contrast region A is positioned at for the face left and right sides, and background contrast region does not comprise ear and head
The region sent out, the height in background contrast region is identical with the height of face, and face width is d, in the background of the face left and right sides
The width of contrast district A is 1/8d~1/3d, and background contrast region A is 1/8d~1/3d with the distance at homonymy face edge.
Two-stage anti-fraud detection method based on multi-orientation Face and environmental information the most according to claim 1, its feature
It is, described step 4) in, the match is successful the rate of the background under the background of current face and unmanned environment in background contrast region
During more than 70%, it is determined that the match is successful, meeting background environment matching condition, otherwise it fails to match.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610614086.6A CN106295522B (en) | 2016-07-29 | 2016-07-29 | A kind of two-stage anti-fraud detection method based on multi-orientation Face and environmental information |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610614086.6A CN106295522B (en) | 2016-07-29 | 2016-07-29 | A kind of two-stage anti-fraud detection method based on multi-orientation Face and environmental information |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106295522A true CN106295522A (en) | 2017-01-04 |
CN106295522B CN106295522B (en) | 2019-09-10 |
Family
ID=57663431
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610614086.6A Expired - Fee Related CN106295522B (en) | 2016-07-29 | 2016-07-29 | A kind of two-stage anti-fraud detection method based on multi-orientation Face and environmental information |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106295522B (en) |
Cited By (25)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107093191A (en) * | 2017-03-06 | 2017-08-25 | 阿里巴巴集团控股有限公司 | A kind of verification method of image matching algorithm, device and computer-readable storage medium |
CN107229927A (en) * | 2017-08-03 | 2017-10-03 | 河北工业大学 | A kind of Face datection anti-fraud method |
CN107239735A (en) * | 2017-04-24 | 2017-10-10 | 复旦大学 | A kind of biopsy method and system based on video analysis |
CN107403146A (en) * | 2017-07-14 | 2017-11-28 | 广东欧珀移动通信有限公司 | detection method and related product |
CN107657519A (en) * | 2017-09-06 | 2018-02-02 | 深圳市恒创智达信息技术有限公司 | Emolument calculating, delivery system and its method |
CN107798742A (en) * | 2017-10-31 | 2018-03-13 | 深圳市中科智诚科技有限公司 | A kind of strong intelligent Time Attendance Device of recognition capability with automatic regulation function |
CN107862247A (en) * | 2017-10-13 | 2018-03-30 | 平安科技(深圳)有限公司 | A kind of human face in-vivo detection method and terminal device |
CN107993302A (en) * | 2017-12-18 | 2018-05-04 | 信利光电股份有限公司 | A kind of face is registered method, apparatus, equipment and computer-readable recording medium |
CN109087429A (en) * | 2018-09-19 | 2018-12-25 | 重庆第二师范学院 | The method of library ticket testimony of a witness consistency check based on face recognition technology |
CN109325413A (en) * | 2018-08-17 | 2019-02-12 | 深圳市中电数通智慧安全科技股份有限公司 | A kind of face identification method, device and terminal |
CN109543521A (en) * | 2018-10-18 | 2019-03-29 | 天津大学 | The In vivo detection and face identification method that main side view combines |
CN109636935A (en) * | 2018-12-07 | 2019-04-16 | 苏州科达科技股份有限公司 | Personnel register method, apparatus and storage medium |
CN109961587A (en) * | 2017-12-26 | 2019-07-02 | 天地融科技股份有限公司 | A kind of monitoring system of self-service bank |
CN110309798A (en) * | 2019-07-05 | 2019-10-08 | 中新国际联合研究院 | A kind of face cheat detecting method extensive based on domain adaptive learning and domain |
WO2019214401A1 (en) * | 2018-05-11 | 2019-11-14 | 京东方科技集团股份有限公司 | Visitor recognition method and apparatus, and access control system |
WO2020087341A1 (en) * | 2018-10-29 | 2020-05-07 | 安徽智传科技有限公司 | Method and system for improving accuracy of multi-view face recognition |
CN111723761A (en) * | 2020-06-28 | 2020-09-29 | 杭州海康威视***技术有限公司 | Method and device for determining abnormal face image and storage medium |
CN111914769A (en) * | 2020-08-06 | 2020-11-10 | 腾讯科技(深圳)有限公司 | User validity judging method, device, computer readable storage medium and equipment |
CN112069917A (en) * | 2020-08-14 | 2020-12-11 | 武汉轻工大学 | Face recognition system for fixed scene |
CN112330833A (en) * | 2020-10-30 | 2021-02-05 | 中国平安人寿保险股份有限公司 | Face recognition attendance data verification method and device and computer equipment |
CN113079161A (en) * | 2021-04-01 | 2021-07-06 | 武汉飓飚信息技术有限公司 | Method and system for sharing monitoring camera |
CN113177437A (en) * | 2017-06-13 | 2021-07-27 | 阿里巴巴集团控股有限公司 | Face recognition method and device |
CN113553990A (en) * | 2021-08-09 | 2021-10-26 | 深圳智必选科技有限公司 | Method and device for tracking and identifying multiple faces, computer equipment and storage medium |
CN114360036A (en) * | 2022-03-21 | 2022-04-15 | 深圳市海清视讯科技有限公司 | Image recognition method, device, server and storage medium |
US11501531B2 (en) | 2020-03-03 | 2022-11-15 | Cyberlink Corp. | Systems and methods for anti-spoofing protection using motion detection and video background analysis |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050276452A1 (en) * | 2002-11-12 | 2005-12-15 | Boland James M | 2-D to 3-D facial recognition system |
CN101702198A (en) * | 2009-11-19 | 2010-05-05 | 浙江大学 | Identification method for video and living body faces based on background comparison |
CN102426646A (en) * | 2011-10-24 | 2012-04-25 | 西安电子科技大学 | Multi-angle human face detection device and method |
CN102609695A (en) * | 2012-02-14 | 2012-07-25 | 上海博物馆 | Method and system for recognizing human face from multiple angles |
CN102819733A (en) * | 2012-08-09 | 2012-12-12 | 中国科学院自动化研究所 | Rapid detection fuzzy method of face in street view image |
CN103971092A (en) * | 2014-04-09 | 2014-08-06 | 中国船舶重工集团公司第七二六研究所 | Facial trajectory tracking method |
-
2016
- 2016-07-29 CN CN201610614086.6A patent/CN106295522B/en not_active Expired - Fee Related
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050276452A1 (en) * | 2002-11-12 | 2005-12-15 | Boland James M | 2-D to 3-D facial recognition system |
CN101702198A (en) * | 2009-11-19 | 2010-05-05 | 浙江大学 | Identification method for video and living body faces based on background comparison |
CN102426646A (en) * | 2011-10-24 | 2012-04-25 | 西安电子科技大学 | Multi-angle human face detection device and method |
CN102609695A (en) * | 2012-02-14 | 2012-07-25 | 上海博物馆 | Method and system for recognizing human face from multiple angles |
CN102819733A (en) * | 2012-08-09 | 2012-12-12 | 中国科学院自动化研究所 | Rapid detection fuzzy method of face in street view image |
CN103971092A (en) * | 2014-04-09 | 2014-08-06 | 中国船舶重工集团公司第七二六研究所 | Facial trajectory tracking method |
Cited By (33)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107093191A (en) * | 2017-03-06 | 2017-08-25 | 阿里巴巴集团控股有限公司 | A kind of verification method of image matching algorithm, device and computer-readable storage medium |
CN107239735A (en) * | 2017-04-24 | 2017-10-10 | 复旦大学 | A kind of biopsy method and system based on video analysis |
CN113177437A (en) * | 2017-06-13 | 2021-07-27 | 阿里巴巴集团控股有限公司 | Face recognition method and device |
CN107403146A (en) * | 2017-07-14 | 2017-11-28 | 广东欧珀移动通信有限公司 | detection method and related product |
CN107229927A (en) * | 2017-08-03 | 2017-10-03 | 河北工业大学 | A kind of Face datection anti-fraud method |
CN107229927B (en) * | 2017-08-03 | 2020-04-03 | 河北工业大学 | Face detection anti-cheating method |
CN107657519A (en) * | 2017-09-06 | 2018-02-02 | 深圳市恒创智达信息技术有限公司 | Emolument calculating, delivery system and its method |
CN107862247A (en) * | 2017-10-13 | 2018-03-30 | 平安科技(深圳)有限公司 | A kind of human face in-vivo detection method and terminal device |
WO2019071739A1 (en) * | 2017-10-13 | 2019-04-18 | 平安科技(深圳)有限公司 | Face living body detection method and apparatus, readable storage medium and terminal device |
CN107798742A (en) * | 2017-10-31 | 2018-03-13 | 深圳市中科智诚科技有限公司 | A kind of strong intelligent Time Attendance Device of recognition capability with automatic regulation function |
CN107993302A (en) * | 2017-12-18 | 2018-05-04 | 信利光电股份有限公司 | A kind of face is registered method, apparatus, equipment and computer-readable recording medium |
CN109961587A (en) * | 2017-12-26 | 2019-07-02 | 天地融科技股份有限公司 | A kind of monitoring system of self-service bank |
US11393267B2 (en) | 2018-05-11 | 2022-07-19 | Boe Technology Group Co., Ltd. | Method and device for identifying visitor and access control system |
WO2019214401A1 (en) * | 2018-05-11 | 2019-11-14 | 京东方科技集团股份有限公司 | Visitor recognition method and apparatus, and access control system |
CN109325413A (en) * | 2018-08-17 | 2019-02-12 | 深圳市中电数通智慧安全科技股份有限公司 | A kind of face identification method, device and terminal |
CN109087429A (en) * | 2018-09-19 | 2018-12-25 | 重庆第二师范学院 | The method of library ticket testimony of a witness consistency check based on face recognition technology |
CN109087429B (en) * | 2018-09-19 | 2020-12-04 | 重庆第二师范学院 | Method for checking consistency of library book-borrowing testimony of witness based on face recognition technology |
CN109543521A (en) * | 2018-10-18 | 2019-03-29 | 天津大学 | The In vivo detection and face identification method that main side view combines |
WO2020087341A1 (en) * | 2018-10-29 | 2020-05-07 | 安徽智传科技有限公司 | Method and system for improving accuracy of multi-view face recognition |
CN109636935A (en) * | 2018-12-07 | 2019-04-16 | 苏州科达科技股份有限公司 | Personnel register method, apparatus and storage medium |
CN110309798B (en) * | 2019-07-05 | 2021-05-11 | 中新国际联合研究院 | Face spoofing detection method based on domain self-adaptive learning and domain generalization |
CN110309798A (en) * | 2019-07-05 | 2019-10-08 | 中新国际联合研究院 | A kind of face cheat detecting method extensive based on domain adaptive learning and domain |
US11501531B2 (en) | 2020-03-03 | 2022-11-15 | Cyberlink Corp. | Systems and methods for anti-spoofing protection using motion detection and video background analysis |
CN111723761A (en) * | 2020-06-28 | 2020-09-29 | 杭州海康威视***技术有限公司 | Method and device for determining abnormal face image and storage medium |
CN111723761B (en) * | 2020-06-28 | 2023-08-11 | 杭州海康威视***技术有限公司 | Method, device and storage medium for determining abnormal face image |
CN111914769A (en) * | 2020-08-06 | 2020-11-10 | 腾讯科技(深圳)有限公司 | User validity judging method, device, computer readable storage medium and equipment |
CN111914769B (en) * | 2020-08-06 | 2024-01-26 | 腾讯科技(深圳)有限公司 | User validity determination method, device, computer readable storage medium and equipment |
CN112069917A (en) * | 2020-08-14 | 2020-12-11 | 武汉轻工大学 | Face recognition system for fixed scene |
CN112069917B (en) * | 2020-08-14 | 2024-02-02 | 武汉轻工大学 | Face recognition system for fixed scene |
CN112330833A (en) * | 2020-10-30 | 2021-02-05 | 中国平安人寿保险股份有限公司 | Face recognition attendance data verification method and device and computer equipment |
CN113079161A (en) * | 2021-04-01 | 2021-07-06 | 武汉飓飚信息技术有限公司 | Method and system for sharing monitoring camera |
CN113553990A (en) * | 2021-08-09 | 2021-10-26 | 深圳智必选科技有限公司 | Method and device for tracking and identifying multiple faces, computer equipment and storage medium |
CN114360036A (en) * | 2022-03-21 | 2022-04-15 | 深圳市海清视讯科技有限公司 | Image recognition method, device, server and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN106295522B (en) | 2019-09-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106295522A (en) | A kind of two-stage anti-fraud detection method based on multi-orientation Face and environmental information | |
KR102596897B1 (en) | Method of motion vector and feature vector based fake face detection and apparatus for the same | |
CN107194341B (en) | Face recognition method and system based on fusion of Maxout multi-convolution neural network | |
CN103440479B (en) | A kind of method and system for detecting living body human face | |
CN100397410C (en) | Method and device for distinguishing face expression based on video frequency | |
CN108182397B (en) | Multi-pose multi-scale human face verification method | |
CN102509138B (en) | Authentication system based on second-generation ID card and human face feature recognition and working method therefor | |
CN105138954A (en) | Image automatic screening, query and identification system | |
TW201137768A (en) | Face recognition apparatus and methods | |
CN106650693A (en) | Multi-feature fusion identification algorithm used for human face comparison | |
CN109685713B (en) | Cosmetic simulation control method, device, computer equipment and storage medium | |
CN108829900A (en) | A kind of Research on face image retrieval based on deep learning, device and terminal | |
CN109858375A (en) | Living body faces detection method, terminal and computer readable storage medium | |
CN110175501A (en) | More people's scene focus recognition methods based on recognition of face | |
Houshmand et al. | Facial expression recognition under partial occlusion from virtual reality headsets based on transfer learning | |
CN112016527B (en) | Panda behavior recognition method, system, terminal and medium based on deep learning | |
CN107066955B (en) | Method for restoring whole human face from local human face area | |
CN104143086A (en) | Application technology of portrait comparison to mobile terminal operating system | |
CN111008971B (en) | Aesthetic quality evaluation method of group photo image and real-time shooting guidance system | |
CN107480586B (en) | Face characteristic point displacement-based biometric photo counterfeit attack detection method | |
CN102184016B (en) | Noncontact type mouse control method based on video sequence recognition | |
CN109934047A (en) | Face identification system and its face identification method based on deep learning | |
CN109377429A (en) | A kind of recognition of face quality-oriented education wisdom evaluation system | |
CN110163567A (en) | Classroom roll calling system based on multitask concatenated convolutional neural network | |
CN106874877A (en) | A kind of combination is local and global characteristics without constraint face verification method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190910 |