CN106886763A - The system and method for real-time detection face - Google Patents
The system and method for real-time detection face Download PDFInfo
- Publication number
- CN106886763A CN106886763A CN201710065482.2A CN201710065482A CN106886763A CN 106886763 A CN106886763 A CN 106886763A CN 201710065482 A CN201710065482 A CN 201710065482A CN 106886763 A CN106886763 A CN 106886763A
- Authority
- CN
- China
- Prior art keywords
- face
- classification device
- main classification
- rear end
- end main
- 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
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000011897 real-time detection Methods 0.000 title claims abstract description 13
- 238000001514 detection method Methods 0.000 claims abstract description 118
- 244000287680 Garcinia dulcis Species 0.000 claims description 25
- 238000010586 diagram Methods 0.000 description 11
- 230000000694 effects Effects 0.000 description 5
- 238000013527 convolutional neural network Methods 0.000 description 4
- 229910002056 binary alloy Inorganic materials 0.000 description 3
- 230000008447 perception Effects 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 238000003066 decision tree Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 210000000056 organ Anatomy 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000000205 computational method Methods 0.000 description 1
- PCHJSUWPFVWCPO-UHFFFAOYSA-N gold Chemical compound [Au] PCHJSUWPFVWCPO-UHFFFAOYSA-N 0.000 description 1
- 239000010931 gold Substances 0.000 description 1
- 229910052737 gold Inorganic materials 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 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
- G06V40/172—Classification, e.g. identification
-
- 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)
- Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- General Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of system and method for real-time detection face.Face detection system includes primary detector and branch circuit detector.Further, primary detector includes front end Main classification device and rear end Main classification device, front end Main classification device and rear end Main classification device are all two classification devices, the face output end of front end Main classification device is connected with the input of rear end Main classification device, and the non-face output end of rear end Main classification device is connected with the input of branch circuit detector.Further, branch circuit detector includes front end branch road grader and rear end branch road grader.Correspondingly, present invention also offers a kind of method for detecting human face.By face detection system disclosed by the invention and method, can not only ensure that detection speed is sufficiently fast, moreover it is possible to make that recall rate is sufficiently high, false drop rate is sufficiently low.
Description
Technical field
The present invention relates to digital image processing field, more particularly to human face detection tech.
Background technology
In the last few years, human face detection tech was one of study hotspot in digital image processing field because it
All played an important role in various applications.Such as, when being taken a picture with digital camera, can be real by real-time detection face
The auto-focusing of existing camera, and then cause that face part is more clear in the photo for shooting.Additionally, human face detection tech is also
A kind of technology is had to use in recognition of face.Because only that being accurately positioned the region comprising face, Cai Nengcong in the picture
In extract the characteristic information of human face, and then realize the identification of face.
" Rapid Object Detection using a Boosted in " CVPR 2001 " international conference collection of thesis
A Cascade of Simple Features " texts describe a kind of method that utilization cascade classifier detects face, the method
It is to be proposed by Paul Viola and Michael Jones.Cascade classifier is that one kind is cascaded by multiple strong classifiers
And the grader for constituting, every grade of strong classifier therein is formed by Weak Classifier training using AdaBoost methods.This people
Than very fast, because when face is detected, most detection window is cascaded in grader most for the detection speed of face detecting method
What first grader is filtered out.In addition when the numerical value of Harr features is calculated, the method also uses integral image techniques,
Make the computational efficiency of Haar character numerical values also higher.But, the recall rate of this method for detecting human face is not high enough, especially detects that
The effect of a little faces being partly blocked and side face is undesirable.Since Paul Viola and Michael Jones propose this people
After face detecting method, it is attempted to further improve it in terms of two：(1) different characteristics of image is used；(2) change
Become the structure of cascade classifier.
" Face Detection Based on Multi-Block LBP in " ICB 2007 " international conference collection of thesis
The texts of Representation " describe a kind of method for detecting human face, and MB-LBP characteristics of image is introduced cascade sort by the method
In device." Locally Assembled Binary (LAB) feature with " CVPR 2008 " international conference collection of thesis
One is also described in a feature-centric cascade for fast and accurate face detection " texts
Method for detecting human face is planted, LAB characteristics of image is introduced cascade sort wherein by the method.Comparatively, both Face datection sides
The recall rate of method increases, but still not ideal enough.《IEEE TRANSACTIONS ON PATTERN ANALYSIS
AND MACHINE INTELLIGENCE》Academic journal the 2nd interim " A Fast and Accurate in 2016
The method for detecting human face introduced in a Unconstrained Face Detector " texts is also a kind of based on cascade classifier
Method for detecting human face.In this approach, NPD characteristics of image is employed, while also making to the decision tree in cascade classifier
Improve.The detection speed and recall rate of this method are all more satisfactory, but its rate of false alarm is of a relatively high.Chinese patent is announced
Number CN105718868A, the day for announcing is on 06 29th, 2016, a kind of entitled " face detection system for multi-pose Face
And method " in disclose a kind of method for detecting human face, the method is the cascade classifier based on LAB characteristics of image and based on SURF
The multi-layer perception (MLP) of characteristics of image is combined, and constructs a kind of human-face detector of funnel structure.The recall rate of this method
It is all more satisfactory with rate of false alarm, but its detection speed is slower.
In addition to the method for detecting human face based on cascade classifier, people also explore other kinds of Face datection side
Method.Such as, " A Convolutional Neural Network Cascade in " CVPR 2015 " international conference collection of thesis
A for Face Detection " texts describe a kind of method for detecting human face of (CNN) based on convolutional neural networks.“BTAS
2015 " " A Deep Pyramid Deformable Part Model for Face in international conference collection of thesis
The texts of Detection " describe a kind of method for detecting human face based on deformable organ model.Both approaches are all influence powers
Larger method, their recall rate is very high, rate of false alarm is very low, but detection speed is very slow, receives their practical value
To limitation.
In sum, in existing method for detecting human face, also without definitely gratifying method.Some methods are detected
Speed is slow；Some method recall rates are not ideal enough；Some method false drop rates are higher.
The content of the invention
It is an object of the invention to provide a kind of face detection system and its method that can overcome above-mentioned technical problem.
Realizing the technical scheme that one of the object of the invention is used is：A kind of face detection system, including primary detector and branch
Circuit detector, the primary detector includes front end Main classification device and rear end Main classification device, the front end Main classification device and rear end master
Grader is all two classification device, and the face output end of the front end Main classification device is connected with the input of rear end Main classification device
Connect, the non-face output end of the rear end Main classification device is connected with the input of branch circuit detector.
Preferably, the branch circuit detector includes front end branch road grader and rear end branch road grader, the front end branch road
Grader and rear end branch road grader are all two classification devices, face output end and the rear end branch road of the front end branch road grader
The input of grader is connected.
Preferably, the correct rejection ratio of the front end Main classification device more than or equal to 98.00% and less than or equal to 99.98%,
Correct recognition rata more than or equal to 98.50% and less than or equal to 99.5%, the correct rejection ratio of the rear end Main classification device more than etc.
In 99.60% and less than or equal to 99.99%, correct recognition rata more than or equal to 86.00% and less than or equal to 99.20%, the branch
The correct rejection ratio and correct recognition rata of circuit detector are both greater than equal to 99.9%.
Preferably, the correct rejection ratio of the front end branch road grader is more than or equal to 80.00% and is less than or equal to
99.50%th, correct recognition rata more than or equal to 99.20% and less than or equal to 99.80%, refuse by the correct of the rear end branch road grader
Exhausted rate and correct recognition rata are both greater than equal to 99.9%.
Preferably, the primary detector is a n ranks depth cascade classifier, wherein the 1st to m rank grader is used as institute
Front end Main classification device is stated, m+1 to n-th order grader is used as the rear end Main classification device, and the m and n is two integers, and
And m ＜ n.
Preferably, the primary detector is a n ranks depth cascade classifier, wherein the 1st to m rank grader is used as institute
Front end Main classification device is stated, m+1 to n-th order grader is used as the rear end Main classification device, and the m and n is two integers, and
And m ＜ n, the front end branch road grader includes 1 shallow cascade classifier, or shallow including more than 2 be cascaded
Cascade classifier.
Preferably, the front end Main classification device and rear end Main classification device are using the characteristics of image that can quickly calculate, the energy
The quick characteristics of image for calculating includes Haar features, LBP features, LAB features or global binary features.
Preferably, the front end Main classification device, rear end Main classification device and front end branch road grader use what can quickly be calculated
Characteristics of image, the characteristics of image that can quickly calculate includes that Haar features, LBP features, LAB features or global binary system are special
Levy, the front end Main classification device and front end branch road grader use different types of characteristics of image.
Preferably, the global binary features are a kind of characteristics of image based on gray level image grey scale pixel value, it
Numerical computations step is：
Step 1, from gray level image, obtains 1 threshold pixels and more than 2 gray values of binaryzation pixel, the threshold
Value pixel is any one pixel in image, and the binaryzation pixel is the pixel being sequentially connected in image；
Step 2, according to below equation, calculates the numerical value of global binary features：
In formula：GBF represents the numerical value of global binary features, and m represents the number of binaryzation pixel, IbkRepresent k-th two
The gray value of value pixel, ItRepresent the gray value of threshold pixels；
Realizing the technical scheme of two uses of the object of the invention is：A kind of method for detecting human face, comprises the following steps：
Step 1101, zoomed image forms image pyramid；
Step 1102, in each image of image pyramid, detection window is moved according to specified step-length, sets up detection
Window set；
Step 1103, using face detection system of the invention, judges that each detection window in detection window set is
It is no comprising face；
Step 1104, the detection window comprising face is placed in face window set；
Step 1105, merges the detection window in face window set,
The step 1103 is comprised the following steps：
Whether step 1201, judges include face in detection window by front end Main classification device；
Step 1202, step 1203 is performed if face is included in detection window, otherwise performs step 1209；
Whether step 1203, judges include face in detection window by rear end Main classification device；
Step 1204, step 1210 is performed if face is included in detection window, otherwise performs step 1205；
Whether step 1205, judges include face in detection window by front end branch road grader
Step 1206, step 1207 is performed if face is included in detection window, otherwise performs step 1209；
Whether step 1207, judges include face in detection window by rear end branch road grader；
Step 1208, step 1210 is performed if face is included in detection window, otherwise performs step 1209；
Step 1209, filters detection window, and perform step 1211；
Step 1210, detection window is placed in face window set；
Step 1211, terminates.
Due to using above-mentioned technical proposal, the face detection system and method that the present invention is provided have an advantageous effect in that：
Both can guarantee that detection speed was sufficiently fast, moreover it is possible to make that recall rate is sufficiently high, rate of false alarm is also sufficiently low.
Brief description of the drawings
Fig. 1 is illustrated that the schematic diagram of face detection system according to an embodiment of the invention；
Fig. 2 is illustrated that the schematic diagram of face detection system according to another embodiment of the invention；
Fig. 3 is illustrated that the schematic diagram of face detection system according to still a further embodiment；
Fig. 4 is illustrated that the schematic diagram of global binary features of the invention；
Fig. 5 is illustrated that the schematic diagram of four squares overall situation binary features of the invention；
Fig. 6 is illustrated that the schematic diagram of four horizontal line sections overall situation binary features of the invention；
Fig. 7 is illustrated that the schematic diagram of four vertical segments overall situation binary features of the invention；
Fig. 8 is illustrated that the schematic diagram of the global binary features of four oblique lines section of the invention；
Fig. 9 is illustrated that the schematic diagram of four backslash line segments overall situation binary features of the invention；
Figure 10 is illustrated that the schematic diagram of face detection system according to still another embodiment of the invention；
Figure 11 is illustrated that the flow chart of method for detecting human face according to an embodiment of the invention；
Figure 12 is illustrated that each detection window judged in detection window set using face detection system of the invention
Whether the flow chart of the method for face is included.
Specific embodiment
With reference to the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is clearly and completely retouched
State.Obviously, described embodiment is only a part of embodiment of the invention, rather than whole embodiments.Based in the present invention
Embodiment, the every other embodiment that those of ordinary skill in the art are obtained under the premise of creative work is not made,
Belong to the scope of protection of the invention.
Fig. 1 shows the schematic diagram of face detection system according to an embodiment of the invention.As shown in figure 1, of the invention
Face detection system includes primary detector 11 and branch circuit detector 12.Further, primary detector 11 includes 1 front end Main classification
Device 111 and 1 rear end Main classification device 112.Front end Main classification device 111 and rear end Main classification device 112 are all two classification devices, they
Two output ends be that (output does not include face for the face output end detection window of face (output comprising) and non-face output end
Detection window).The face output end of front end Main classification device 111 is connected with the input of rear end Main classification device 112, rear end master
The non-face output end of grader 112 is connected with the input of branch circuit detector 12.
When using face detection system of the invention, first have to according to certain scaling images, form image gold
Word tower.Then, in each image of image pyramid, with specified step-length, moved according to order from top to bottom, from left to right
Dynamic detection window.During mobile detection window, whether face is included in distinguishing detection window using face detection system.
Finally, reuse such as non-maximum and suppress (Non-maximum Suppression) method merging detection window.
The effect of front end Main classification device 111 is to filter the most detection window not comprising face, and it is a kind of operation
The grader that speed, correct rejection ratio are moderate, correct recognition rata is higher, such as correct rejection ratio are more than or equal to 98.00%
And less than or equal to 99.98%, correct recognition rata more than or equal to 98.50% and less than or equal to 99.5%.Wherein, correct rejection ratio and
Correct recognition rata is calculated according to below equation：
Rear end Main classification device 112 is that a kind of speed of service is very fast, the classification that correct rejection ratio is higher, correct recognition rata is moderate
Device, such as correct rejection ratio more than or equal to 99.60% and less than or equal to 99.99%, correct recognition rata more than or equal to 86.00% and
Less than or equal to 99.20%.It further filters the detection window not comprising face, and then correct detection of the output comprising face
Window.In those detection windows filtered by rear end Main classification device 112, both comprising non-face window, it is also possible to comprising some
Face window.Generally, the face included in the face window that these are filtered by rear end Main classification device 112 is all the difficult inspection of comparing
Survey, such as the face or side face being at least partially obscured.Branch circuit detector 12 be a kind of correct rejection ratio and correct recognition rata all
Grader very high, such as correct rejection ratio are more than or equal to 99.9% more than or equal to 99.9%, correct recognition rata, and it is used for from that
In the detection window for being filtered out by rear end Main classification device 112 a bit, the detection window comprising face is further selected.
As it was previously stated, front end Main classification device 111 is that a kind of speed of service is very fast, correct rejection ratio is moderate, correct recognition rata
Grader higher, therefore in the case where most of detection window not comprising face is filtered, it can make most of bag
Detection window containing face is by detection.Rear end Main classification device 112 is that a kind of speed of service is very fast, correct rejection ratio is higher, just
The moderate grader of true discrimination, it can be most of face windows detectings for being easier to detection out, while making to be examined by mistake
The ratio very little of the window (detection window not comprising face is regarded as the detection window comprising face) of survey.Branch road is classified
Device 12 is a kind of correct rejection ratio and correct recognition rata grader all very high.The effect of branch circuit detector 12 is, from those quilts
In those detection windows that rear end Main classification device 112 is filtered out, the detection window comprising face is detected.These detection windows
The face majority that mouth is included is the face of the difficult detection of comparing, such as the face being partly blocked and side face.Under normal circumstances, correctly
The speed of service of reject rate and correct recognition rata grader all very high is all slow, but due to front end Main classification device 111
Through having filtered the most detection window not comprising face, it is therefore desirable to which the number of windows detected by branch circuit detector 12 is ratio
Less, such as less than 200.As can be seen here, the face detection system for being designed according to such scheme, both can guarantee that detection speed
It is sufficiently fast, correct rejection ratio and correct recognition rata can be made sufficiently high again.
As shown in Fig. 2 according to another embodiment of the invention, branch circuit detector 12 further includes 1 front end branch
Road grader 121 and 1 rear end branch road grader 122.Front end branch road grader 121 and rear end branch road grader 122 are all two
Class grader, their two output ends are face output end (detection window of the output comprising face) and non-face output end
(detection window of the output not comprising face).The face output end of front end branch road grader 121 and rear end branch road grader 122
Input is connected.
The effect of front end Main classification device 111 and front end branch road grader 121 is identical, contributes to filter not comprising people
The detection window of face, and then reduce the quantity of the detection window processed by subsequent classifier.They both can be same type of
Grader, it is also possible to be not same type of grader.If they are same type of graders, it is necessary to using different figures
As characteristic set, could so go to investigate detection window from different angles.
Front end branch road grader 121 is that a kind of speed of service is very fast, correct rejection ratio is moderate, correct recognition rata is higher point
Class device, such as correct rejection ratio are more than or equal to 99.20% more than or equal to 80.00% and less than or equal to 99.50%, correct recognition rata
And less than or equal to 99.80%, it is used to further filter those detection windows filtered out by rear end Main classification device 112.Rear end
Branch road grader 122 is the classification of a kind of correct rejection ratio and correct recognition rata (being such as both greater than equal to 99.9%) all very high
Device.It is, for example possible to use following several graders are used as rear end branch road grader 122：(1) convolutional neural networks；(2) it is based on
The multi-layer perception (MLP) of SURF features.The human-face detector conduct that correct recognition rata is very high, correct rejection ratio is very high can also be used
Rear end branch road grader 122, such as the human-face detector based on deformable organ model.
As shown in figure 3, the further embodiment of face detection system of the invention, primary detector 11 is a n rank
Deep cascade classifier.Wherein, the 1st to m rank grader is taken as front end Main classification device 111 (m ＜ n), m+1 to n-th order
Grader is taken as rear end Main classification device 112.Front end branch road grader 121 both can be 1 shallow cascade classifier, or
Be cascaded the grader for constituting by the shallow cascade classifier of more than 2.Deep cascade classifier is the more cascade of exponent number point
Class device, such as exponent number are 17~30 cascade classifier.Conversely, shallow cascade classifier is the less cascade classifier of exponent number, than
Such as the cascade classifier that exponent number is 3~10.
In order to improve the speed of Face datection, the front end Main classification device in face detection system of the invention, main point of rear end
Class device and front end branch road grader are using the characteristics of image that can quickly calculate, such as Haar features, LBP features, LAB features
Or global binary features (Global Binary Feature, GBF).Global binary features are a kind of based on gray level image
The characteristics of image of grey scale pixel value, it is related to two kinds of pixel：(1) 1 threshold pixels；(2) m binaryzation pixel (m >=
2).Threshold pixels are any one pixels in image, and binaryzation pixel is the pixel being sequentially connected in image.Global binary system
The numerical value of feature is calculated according to following steps：
Step 1, from gray level image, obtains 1 threshold pixels and the m gray value of binaryzation pixel (m >=2).
As shown in figure 4, threshold pixels 401 are any one pixels in image, binaryzation pixel 402a to 402f is figure
The pixel being sequentially connected as in.
Step 2, according to below equation, calculates the numerical value of global binary features：
In formula：GBF represents the numerical value of global binary features；IbkRepresent k-th gray value of binaryzation pixel；ItRepresent
The gray value of threshold pixels；
When cascade classifier is trained, first have to build a characteristics of image set, then selected from characteristics of image set
The stronger characteristics of image of those resolution capabilities is selected, for the every first-level class device set up in cascade classifier.By global binary system
The computational methods of character numerical value understand that the difference of different global binary features is embodied in 3 aspects：(1) threshold pixels
Position；(2) quantity of binaryzation pixel；(3) position of binaryzation pixel.When global binary picture characteristic set is built,
If not specifying restrictive condition, the quantity of the characteristics of image that global binary picture characteristic set is included will be very huge,
Can so make the training time of cascade classifier long.Therefore, when global binary picture characteristic set is built, it is necessary in advance
Limit the quantity of binaryzation pixel, and the relative position relation between binaryzation pixel.Fig. 5 to Fig. 9 shows 5 types
Global binary features.As shown in figure 5, binaryzation pixel 402a to 402d is sequentially connected forms a square, therefore handle
This global binary features are referred to as four square overall situation binary features, abbreviation QGBF_4 features.As shown in fig. 6, two-value
Change pixel 402a to 402d is sequentially connected and forms a horizontal line section, therefore this global binary features are called four water
The global binary features of horizontal line section, abbreviation HLGBF_4 features.As shown in fig. 7, binaryzation pixel 402a to 402d is sequentially connected shape
Into a vertical segment, therefore this global binary features are called four vertical segment overall situation binary features, referred to as
VLGBF_4 features.As shown in figure 8, binaryzation pixel 402a to 402d is sequentially connected forms an oblique line section, therefore this
Global binary features are referred to as four global binary features of oblique line section, abbreviation SLGBF_4 features.As shown in figure 9, binaryzation picture
Plain 402a to 402d is sequentially connected and forms a backslash line segment, therefore this global binary features are called four backslashes
The global binary features of section, abbreviation BSLGBF_4 features.
As shown in Figure 10, another embodiment of face detection system of the invention, primary detector 11 is a base
In the 20 ranks depth cascade classifier of QGBF_4 features.Wherein, the 1st rank to the 9th rank grader is taken as front end Main classification device 111,
10th rank to the 20th rank grader is taken as rear end Main classification device 112.Front end branch road grader 121 is cascaded comprising 2
Cascade classifier, i.e. ground floor front end branch road grader 1211 and second layer front end branch road grader 1212.Wherein ground floor
Front end branch road grader 1211 is a shallow cascade classifier of 5 ranks based on VLGBF_4 features, second layer front end branch road grader
1212 is a shallow cascade classifier of 5 ranks based on HLGBF_4 features.Rear end branch road grader 122 is one special based on SURF
The multi-layer perception (MLP) levied.Primary detector 11, ground floor front end branch road grader 1211 and second layer front end branch road grader 1212
In every single order grader all be use Adaboost methods, by binary decision tree train.By training, make every single order
The correct rejection ratio of grader is more than or equal to 99.50% more than or equal to 45.00% and less than or equal to 55.00%, correct recognition rata
And less than or equal to 99.90%.
As shown in figure 11, the embodiment of face detection system of the invention, method for detecting human face bag of the invention
Include following steps：
Step 1101, zoomed image forms image pyramid；
Step 1102, in each image of image pyramid, detection window is moved according to specified step-length, sets up detection
Window set；
Step 1103, using face detection system of the invention, judges that each detection window in detection window set is
It is no comprising face；
Step 1104, the detection window comprising face is placed in face window set；
Step 1105, merges the detection window in face window set.
As shown in figure 12, further, step 1103 includes：
Whether step 1201, judges include face in detection window by front end Main classification device；
Step 1202, step 1203 is performed if face is included in detection window, otherwise performs step 1209；
Whether step 1203, judges include face in detection window by rear end Main classification device；
Step 1204, step 1210 is performed if face is included in detection window, otherwise performs step 1205；
Whether step 1205, judges include face in detection window by front end branch road grader
Step 1206, step 1207 is performed if face is included in detection window, otherwise performs step 1209；
Whether step 1207, judges include face in detection window by rear end branch road grader；
Step 1208, step 1210 is performed if face is included in detection window, otherwise performs step 1209；
Step 1209, filters detection window, and perform step 1211；
Step 1210, detection window is placed in face window set；
Step 1211, terminates.
Claims (10)
1. a kind of system of real-time detection face, it is characterised in that including primary detector and branch circuit detector, the primary detector
Including front end Main classification device and rear end Main classification device, the front end Main classification device and rear end Main classification device are all two classification devices,
The face output end of the front end Main classification device is connected with the input of rear end Main classification device, the rear end Main classification device it is non-
Face output end is connected with the input of branch circuit detector.
2. the system of real-time detection face according to claim 1, it is characterised in that the branch circuit detector includes front end
Branch road grader and rear end branch road grader, the front end branch road grader and rear end branch road grader are all two classification devices,
The face output end of the front end branch road grader is connected with the input of rear end branch road grader.
3. the system of real-time detection face according to claim 1 and 2, it is characterised in that the front end Main classification device
Correct rejection ratio more than or equal to 98.00% and less than or equal to 99.98%, correct recognition rata more than or equal to 98.50% and less than etc.
In 99.5%, the correct rejection ratio of the rear end Main classification device is more than or equal to 99.60% and less than or equal to 99.99%, correct knowledge
, more than or equal to 86.00% and less than or equal to 99.20%, the correct rejection ratio and correct recognition rata of the branch circuit detector are not all for rate
More than or equal to 99.9%.
4. the system of real-time detection face according to claim 2, it is characterised in that the front end branch road grader is just
True reject rate is more than or equal to 80.00% and less than or equal to 99.50%, correct recognition rata is more than or equal to 99.20% and is less than or equal to
99.80%, the correct rejection ratio and correct recognition rata of the rear end branch road grader are both greater than equal to 99.9%.
5. the system of real-time detection face according to claim 1, it is characterised in that the primary detector is a n rank
Deep cascade classifier, wherein the 1st to m rank grader is used as the front end Main classification device, m+1 to n-th order grader quilt
As the rear end Main classification device, the m and n is two integers, and m ＜ n.
6. the system of real-time detection face according to claim 2, it is characterised in that the primary detector is a n rank
Deep cascade classifier, wherein the 1st to m rank grader is used as the front end Main classification device, m+1 to n-th order grader quilt
As the rear end Main classification device, the m and n is two integers, and m ＜ n, the front end branch road grader include 1 it is shallow
Cascade classifier, or including the shallow cascade classifier being cascaded of more than 2.
7. the system of real-time detection face according to claim 1, it is characterised in that the front end Main classification device and rear end
Using the characteristics of image that can quickly calculate, the characteristics of image that can quickly calculate includes that Haar features, LBP are special to Main classification device
Levy, LAB features or global binary features.
8. the system of real-time detection face according to claim 2, it is characterised in that the front end Main classification device, rear end
, using the characteristics of image that can quickly calculate, the characteristics of image that can quickly calculate includes for Main classification device and front end branch road grader
Haar features, LBP features, LAB features or global binary features, the front end Main classification device and front end branch road grader are used
Different types of characteristics of image.
9. the system of the real-time detection face according to claim 7 or 8, it is characterised in that the global binary features
It is a kind of characteristics of image based on gray level image grey scale pixel value, its numerical computations step is：
Step 1, from gray level image, obtains 1 threshold pixels and more than 2 gray values of binaryzation pixel, the threshold value picture
Element is any one pixel in image, and the binaryzation pixel is the pixel being sequentially connected in image；
Step 2, according to below equation, calculates the numerical value of global binary features：
In formula：GBF represents the numerical value of global binary features, and m represents the number of binaryzation pixel, IbkRepresent k-th binaryzation
The gray value of pixel, ItRepresent the gray value of threshold pixels；
10. a kind of method for detecting human face, comprises the following steps：
Step 1101, zoomed image forms image pyramid；
Step 1102, in each image of image pyramid, detection window is moved according to specified step-length, sets up detection window
Set；
Step 1103, using face detection system of the invention, judges whether each detection window in detection window set wraps
Containing face；
Step 1104, the detection window comprising face is placed in face window set；
Step 1105, merges the detection window in face window set,
Characterized in that, the step 1103 is comprised the following steps：
Whether step 1201, judges include face in detection window by front end Main classification device；
Step 1202, step 1203 is performed if face is included in detection window, otherwise performs step 1209；
Whether step 1203, judges include face in detection window by rear end Main classification device；
Step 1204, step 1210 is performed if face is included in detection window, otherwise performs step 1205；
Whether step 1205, judges include face in detection window by front end branch road grader
Step 1206, step 1207 is performed if face is included in detection window, otherwise performs step 1209；
Whether step 1207, judges include face in detection window by rear end branch road grader；
Step 1208, step 1210 is performed if face is included in detection window, otherwise performs step 1209；
Step 1209, filters detection window, and perform step 1211；
Step 1210, detection window is placed in face window set；
Step 1211, terminates.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710065482.2A CN106886763B (en) | 2017-01-20 | 2017-01-20 | System and method for detecting human face in real time |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710065482.2A CN106886763B (en) | 2017-01-20 | 2017-01-20 | System and method for detecting human face in real time |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106886763A true CN106886763A (en) | 2017-06-23 |
| CN106886763B CN106886763B (en) | 2020-02-18 |
Family
ID=59178939
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710065482.2A Expired - Fee Related CN106886763B (en) | 2017-01-20 | 2017-01-20 | System and method for detecting human face in real time |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106886763B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107341477A (en) * | 2017-07-10 | 2017-11-10 | 北京联合大学 | A kind of fast cascaded formula is without constraint method for detecting human face |
| CN108256451A (en) * | 2018-01-05 | 2018-07-06 | 百度在线网络技术（北京）有限公司 | For detecting the method and apparatus of face |
| CN109214328A (en) * | 2018-08-29 | 2019-01-15 | 成都睿码科技有限责任公司 | Face grasping system based on face recognition engine |
| CN109993061A (en) * | 2019-03-01 | 2019-07-09 | 珠海亿智电子科技有限公司 | A kind of human face detection and tracing method, system and terminal device |
| CN110472570A (en) * | 2019-08-14 | 2019-11-19 | 旭辉卓越健康信息科技有限公司 | A kind of recognition of face multipath deep neural network method based on adaptive weighting |
| CN112801233A (en) * | 2021-04-07 | 2021-05-14 | 杭州海康威视数字技术股份有限公司 | Internet of things equipment honeypot system attack classification method, device and equipment |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103049733A (en) * | 2011-10-11 | 2013-04-17 | 株式会社理光 | Human face detection method and human face detection equipment |
| US20140286527A1 (en) * | 2013-03-20 | 2014-09-25 | Qualcomm Incorporated | Systems and methods for accelerated face detection |
| CN104850818A (en) * | 2014-02-17 | 2015-08-19 | 华为技术有限公司 | Face detector training method, face detection method and device |
| CN105138956A (en) * | 2015-07-22 | 2015-12-09 | 小米科技有限责任公司 | Face detection method and device |
| CN105335684A (en) * | 2014-06-25 | 2016-02-17 | 小米科技有限责任公司 | Face detection method and device |
| CN105718868A (en) * | 2016-01-18 | 2016-06-29 | 中国科学院计算技术研究所 | Face detection system and method for multi-pose faces |
| CN105912990A (en) * | 2016-04-05 | 2016-08-31 | 深圳先进技术研究院 | Face detection method and face detection device |
-
2017
- 2017-01-20 CN CN201710065482.2A patent/CN106886763B/en not_active Expired - Fee Related
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103049733A (en) * | 2011-10-11 | 2013-04-17 | 株式会社理光 | Human face detection method and human face detection equipment |
| US20140286527A1 (en) * | 2013-03-20 | 2014-09-25 | Qualcomm Incorporated | Systems and methods for accelerated face detection |
| CN104850818A (en) * | 2014-02-17 | 2015-08-19 | 华为技术有限公司 | Face detector training method, face detection method and device |
| CN105335684A (en) * | 2014-06-25 | 2016-02-17 | 小米科技有限责任公司 | Face detection method and device |
| CN105138956A (en) * | 2015-07-22 | 2015-12-09 | 小米科技有限责任公司 | Face detection method and device |
| CN105718868A (en) * | 2016-01-18 | 2016-06-29 | 中国科学院计算技术研究所 | Face detection system and method for multi-pose faces |
| CN105912990A (en) * | 2016-04-05 | 2016-08-31 | 深圳先进技术研究院 | Face detection method and face detection device |
Non-Patent Citations (1)
| Title |
|---|
| 徐科 等: "基于全局二值模式的特征提取方法及其应用", 《模式识别与人工智能》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107341477A (en) * | 2017-07-10 | 2017-11-10 | 北京联合大学 | A kind of fast cascaded formula is without constraint method for detecting human face |
| CN108256451A (en) * | 2018-01-05 | 2018-07-06 | 百度在线网络技术（北京）有限公司 | For detecting the method and apparatus of face |
| CN108256451B (en) * | 2018-01-05 | 2022-09-27 | 百度在线网络技术（北京）有限公司 | Method and device for detecting human face |
| CN109214328A (en) * | 2018-08-29 | 2019-01-15 | 成都睿码科技有限责任公司 | Face grasping system based on face recognition engine |
| CN109993061A (en) * | 2019-03-01 | 2019-07-09 | 珠海亿智电子科技有限公司 | A kind of human face detection and tracing method, system and terminal device |
| CN109993061B (en) * | 2019-03-01 | 2021-12-07 | 珠海亿智电子科技有限公司 | Face detection and recognition method, system and terminal equipment |
| CN110472570A (en) * | 2019-08-14 | 2019-11-19 | 旭辉卓越健康信息科技有限公司 | A kind of recognition of face multipath deep neural network method based on adaptive weighting |
| CN112801233A (en) * | 2021-04-07 | 2021-05-14 | 杭州海康威视数字技术股份有限公司 | Internet of things equipment honeypot system attack classification method, device and equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106886763B (en) | 2020-02-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106886763A (en) | The system and method for real-time detection face | |
| CN111639692B (en) | Shadow detection method based on attention mechanism | |
| CN110956094B (en) | RGB-D multi-mode fusion personnel detection method based on asymmetric double-flow network | |
| CN111415316B (en) | Defect data synthesis method based on generation countermeasure network | |
| CN108334848B (en) | Tiny face recognition method based on generation countermeasure network | |
| WO2019169895A1 (en) | Fast side-face interference resistant face detection method | |
| CN107491762B (en) | A kind of pedestrian detection method | |
| CN107330390B (en) | People counting method based on image analysis and deep learning | |
| CN107169415A (en) | Human motion recognition method based on convolutional neural networks feature coding | |
| CN107463920A (en) | A kind of face identification method for eliminating partial occlusion thing and influenceing | |
| CN107451994A (en) | Object detecting method and device based on generation confrontation network | |
| CN107092884B (en) | Rapid coarse-fine cascade pedestrian detection method | |
| KR102320985B1 (en) | Learning method and learning device for improving segmentation performance to be used for detecting road user events using double embedding configuration in multi-camera system and testing method and testing device using the same | |
| CN113536972B (en) | Self-supervision cross-domain crowd counting method based on target domain pseudo label | |
| JP2007047965A (en) | Method and device for detecting object of digital image, and program | |
| CN106529398A (en) | Quick and accurate face detection method based on cascade structure | |
| CN107194946B (en) | FPGA-based infrared salient object detection method | |
| CN110472634A (en) | Change detecting method based on multiple dimensioned depth characteristic difference converged network | |
| CN107944437B (en) | A kind of Face detection method based on neural network and integral image | |
| CN107066963A (en) | A kind of adaptive people counting method | |
| CN112766186A (en) | Real-time face detection and head posture estimation method based on multi-task learning | |
| CN107392089A (en) | A kind of eyebrow movement detection method and device and vivo identification method and system | |
| CN110532959B (en) | Real-time violent behavior detection system based on two-channel three-dimensional convolutional neural network | |
| CN110503049B (en) | Satellite video vehicle number estimation method based on generation countermeasure network | |
| CN112926667B (en) | Method and device for detecting saliency target of depth fusion edge and high-level feature |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20200218 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |