CN106529398B - Fast accurate method for detecting human face based on cascade structure - Google Patents
Fast accurate method for detecting human face based on cascade structure Download PDFInfo
- Publication number
- CN106529398B CN106529398B CN201610847733.8A CN201610847733A CN106529398B CN 106529398 B CN106529398 B CN 106529398B CN 201610847733 A CN201610847733 A CN 201610847733A CN 106529398 B CN106529398 B CN 106529398B
- Authority
- CN
- China
- Prior art keywords
- face
- image
- frame
- pixel
- face frame
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—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
- G06V40/164—Detection; Localisation; Normalisation using holistic features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The present invention relates to field of neural networks, and it discloses a kind of fast accurate method for detecting human face based on cascade structure, include the following steps:(A)Fast human-face model detects;(B)Screen face frame;(C)Accurate faceform's detection;(D)Merge face frame and export.The beneficial effects of the invention are as follows:The speed and accuracy of detection have been taken into account simultaneously;Use the pixel comparative feature on gray level image and the high speed detector of soft cascade cascade decision trees, face frame to be selected is quickly selected, further uses the accurate detector for employing self study feature, screens face frame to be selected, with the presence or absence of face, the accuracy of face frame ensure that.
Description
【Technical field】
The present invention relates to field of face identification more particularly to a kind of fast accurate Face datection sides based on cascade structure
Method.
【Background technology】
Face datection is first step of recognition of face, and its task is to any given a sub-picture or a group picture
As sequence, judged automatically using machine with the presence or absence of face in image or the sequence, if there are face, find out its position and big
It is small.Facial image because Different Individual the internal factors such as age, sex, race, weight influence and there are larger differences, add
Upper illumination is blocked, the influence of the external factors such as angle, distance, collecting device, further increases the complexity of facial image.
Therefore, facial image and inhuman face image (ambient image) are distinguished very well, that is, improve the recall rate of face, reduced non-
Facial image mistake is divided into the quantity of facial image, needs to use extremely complex face characteristic and detection model.But it uses multiple
After miscellaneous face characteristic and detection model, detection efficiency will be caused to substantially reduce.For example it is directly carried out using deep neural network
Face datection, although being currently that detection result is best, its detection efficiency is relatively low, can not be satisfied with some to requirement of real-time
Higher system.And the method for detecting human face of some speed, the method such as PICO of pixel comparative feature is such as used, although examining
Degree of testing the speed is fast, but the face recall rate under some dynamic scenes, very low.
The detection of facial image under dynamic scene, the low detection method of model complexity, face recall rate and accuracy
It is not high, and the detection method that model complexity is high, then with high computing cost.Current method for detecting human face, in dynamic
Off field, higher computational efficiency can not be kept simultaneously, and ensures recall rate and the accuracy of face.Although have using harr,
The human-face detector application of the simple detector cascade deep neural network of SURF, HOG feature, but because these feature sheets of front
Body is more complicated, and calculating speed is not especially fast, in addition there is deep neural network below, whole detection speed is limited;In addition
During cascade, all face windows that can generally detect simple detector are sent into deep neural network, this also results in rear end
Face datection takes more.
【Invention content】
In order to solve the problems in the prior art, the present invention provides a kind of fast accurate face inspections based on cascade structure
Survey method, solution can not keep higher computational efficiency simultaneously in the prior art, and ensure recall rate and the accuracy of face
Problem.
The present invention is achieved by the following technical solutions:It designs, manufactured a kind of fast accurate based on cascade structure
Method for detecting human face includes the following steps:(A) fast human-face model detects;(B) face frame is screened;(C) precisely faceform's inspection
It surveys;(D) merge face frame and export.
As a further improvement on the present invention:In the step (A), accelerated model uses more special based on gradation of image
The Face datection algorithm of sign, decision tree and soft cascade structures carries out model foundation.
As a further improvement on the present invention:The step (A) is specially:(A1) original RGB color image is obtained;(A2)
Coloured image is converted to single pass gray level image;(A3) filtering process;(A4) size and location of frame is specified search for, is not required to
Pyramid sampling is done to image, only search box is amplified and diminution is handled;(A5) by the image in search box, it is sent into instruction
In each decision tree in the Face datection model perfected and obtained score value is handled.
As a further improvement on the present invention:In the step (C), accurate faceform is depth convolutional neural networks,
For the neural network using the 3 channel RGB images of 40X40 as input, which includes 2 convolutional layers, 2 sample levels and 2
A full articulamentum.
As a further improvement on the present invention:In the step (C), in the training process of model, with last time training
Complete model is detected negative sample picture, and the picture frame that mistake is divided into face cuts to build negative sample sectional drawing, is instructed with the last time
The negative sample sectional drawing practiced merges, for the training of this network.
As a further improvement on the present invention:In the step (D), merge face frame, in being detected to accurate faceform
It is judged as that the face window of face merges, exports final face frame.
As a further improvement on the present invention:In the step (A3), the noise level of image is characterized with Graded Density,
Using the average gradient of image as threshold value, bianry image is constructed, if the Grad of pixel is more than threshold value, 1 is put, otherwise sets to 0;By two
It is worth the corresponding Grad summation of pixel for 1 in image;In binary map, it is scanned with the window of fixed size, step-length
For the length of side of window, ensure that each pixel will not be repetitively scanned;In scanning window, if it is 1 to have pixel value, own in window
Pixel value be all assigned a value of 1, after the completion of scanning, in statistical picture pixel value be 1 pixel number;With Grad and remove
With the pixel number counted on, Graded Density is obtained.
As a further improvement on the present invention:In the step (A5), if the image in search box is after one tree
Cumulative score, less than threshold value, then abandon the search box, be transferred to step (A3) and handled;If higher than threshold value, continue under
Judged on one tree;Image in a search box, has passed through all decision trees, then exports the frame, forms one
A face frame to be selected, then screens face frame to be selected.
As a further improvement on the present invention:In the step (B), put by face frame score and the synthesis for being overlapped number
Evaluation index is believed to screen face frame.
The beneficial effects of the invention are as follows:The speed and accuracy of detection have been taken into account simultaneously;Use the pixel on gray level image
Comparative feature and the high speed detector of soft cascade cascade decision trees, quickly select face frame to be selected, further use and adopt
With the accurate detector of self study feature, face frame to be selected is screened, if there are faces, ensure that the accuracy of face frame.
【Description of the drawings】
Fig. 1 is flow diagram of the present invention;
Fig. 2 is the flow diagram of fast human-face model of the present invention detection;
Fig. 3 is the present invention accurate faceform's structure diagram that precisely faceform detects.
【Specific embodiment】
The present invention is further described for explanation and specific embodiment below in conjunction with the accompanying drawings.
A kind of fast accurate method for detecting human face based on cascade structure, includes the following steps:(A) fast human-face model is examined
It surveys;(B) face frame is screened;(C) precisely faceform's detection;(D) merge face frame and export.
In the step (A), accelerated model is used based on gradation of image comparative feature, decision tree and soft cascade knots
The Face datection algorithm of structure carries out model foundation.
The step (A) is specially:(A1) original RGB color image is obtained;(A2) coloured image is converted to single channel
Gray level image;(A3) filtering process;(A4) size and location of frame is specified search for, does not need to do image pyramid sampling,
Only search box is amplified and diminution is handled;(A5) it by the image in search box, is sent into trained Face datection model
Each decision tree in and obtained score value is handled.
In the step (C), accurate faceform is depth convolutional neural networks, and the neural network is with 3 channels of 40X40
RGB image includes 2 convolutional layers, 2 sample levels and 2 full articulamentums as input, the neural network.
In the step (C), in the training process of model, with the model trained of last time, to negative sample picture into
Row detection, the picture frame that mistake is divided into face cut to build negative sample sectional drawing, and the negative sample sectional drawing of training merges with the last time, uses
In the training of this network.
In the step (D), merge face frame, be judged as that the face window of face carries out in being detected to accurate faceform
Merge, export final face frame.
In the step (A3), the noise level of image is characterized with Graded Density, using the average gradient of image as threshold value,
Bianry image is constructed, if the Grad of pixel is more than threshold value, 1 is put, otherwise sets to 0;It will be 1 pixel correspondence in bianry image
Grad summation;It in binary map, is scanned with the window of fixed size, step-length is the length of side of window, ensures each picture
Element will not be repetitively scanned.In scanning window, if it is 1 to have pixel value, all pixel values are all assigned a value of 1 in window, scanning
After the completion, in statistical picture the pixel that pixel value is 1 number, with Grad and divided by the pixel number that counts on,
Obtain Graded Density.
In the step (A5), if image in search box in the cumulative score after one tree, less than threshold value, is then lost
The search box is abandoned, step (A3) is transferred to and is handled;If higher than threshold value, continuation is judged in next tree;When one
Image in search box has passed through all decision trees, then exports the frame, forms a face frame to be selected, and then screening is treated
It chooses face frame.
The face frame to be selected includes confidence score, upper left angular coordinate (x, y), the height height and width of frame
Spend width;Face frame is screened by the synthesis confidence evaluation index of face frame score and overlapping number.
To judging whether image is strong noise image, the evaluation method of other usual means progress may be used;Height is made an uproar
The filtering mode of acoustic image might have other usual ways.
In one embodiment, overall procedure such as Fig. 1 institutes of a kind of fast accurate method for detecting human face based on cascade structure
Show:Original image by the human-face detector of accelerated model, quickly obtains face frame to be selected (size and location of frame), then
It using the human-face detector of accurate model, further confirms that whether the image in face frame to be selected is face, is that then the output phase should
Face frame.Whole human-face detector uses cascade structure, by classification capacity is weaker but fireballing naive model Face datection
Device is strong with classification capacity but slow-footed accurate model human-face detector is used in combination, and is got rid of by the former most of non-face
Image-region accurately filters out the image-region containing face by the latter, finally merges the figure containing face using NMS algorithms
As region, final face frame is exported.
Be detected by fast human-face model, as shown in Fig. 2, accelerated model use based on gradation of image comparative feature,
Decision tree and the Face datection algorithm of soft cascade structures.Specially:
2.1 obtain the original rgb coloured images of 3 channels;
Coloured image is converted to single pass gray level image by 2.2;
2.3 judge whether the image is strong noise image using Graded Density method, are that image is filtered, abatement
The influence of noise on image.When light is weaker, the noise performance of image capture device in itself is apparent, to based on pixel compare
The human-face detector of feature is affected.The present invention characterizes the noise level of image with Graded Density.With the average ladder of image
It spends for threshold value, constructs bianry image, if the Grad of pixel is more than threshold value, put 1, otherwise set to 0.To be 1 in bianry image
The corresponding Grad summation of pixel.In binary map, it is scanned, step-length with the window (such as 2X2 pixel) of fixed size
For the length of side of window, ensure that each pixel will not be repetitively scanned.In scanning window, if it is 1 to have pixel value, own in window
Pixel value be all assigned a value of 1, after the completion of scanning, in statistical picture pixel value be 1 pixel number.Graded Density is shown in formula
(1)
Wherein Sx,yRefer to the Grad for being more than average gradient in image, N refers to of the pixel counted on scanning window
Number.When image is larger, Graded Density value only need to be sought image local area.Sx,yCalculating formula such as formula (2) shown in,
Sx,y=(| f (x+1, y)-f (x, y) |+| f (x, y+1)-f (x, y) |)/f (x, y) formula (2)
Wherein, f (x, y) represents image x, the pixel value at y-coordinate.Pass through ρsThe size of value can determine whether the noise water of image
It is flat.If being judged as strong noise image, image is filtered, the present invention also has other filtering sides certainly using medium filtering
Formula, such as gaussian filtering.
2.4 specify search for the size and location of frame, and this method is different from general Face datection algorithm, does not need to image
Pyramid sampling is done, only all search boxes is needed to be amplified and reduce.The reason is that algorithm use is characterized in pixel grey scale
It is worth comparative feature;
2.5 by the image in search box, is sent into each decision tree in trained Face datection model, so as to
To obtain a score value, which characterizes similarity of the image with face of decision tree assessment.The depth of decision tree is
M, the number of decision tree is n.In a certain range, m values are bigger, and the classification capacity of single tree is stronger, and n is bigger, the classification of detector
Ability is stronger, but speed is also slower.M takes 6, n to take 500 in the present invention.Score of the image on single tree needs and preceding one tree
Score add up;
If image in 2.6 search boxes in the cumulative score after one tree, less than threshold value, then abandons the search box,
Turn 2.3;If higher than threshold value, continuation is judged in next tree;
2.7, when the image in a search box, have passed through all decision trees, have then exported the frame, composition one is to be selected
Face frame.
2.8 screening face frames to be selected.Face frame to be selected, comprising confidence (score), upper left angular coordinate (x, y),
The height (height) of frame and width (width), 5 information.Around face, it generally will appear multiple face frames to be selected, weight
Folded degree is very high, and around non-face, if there is face frame to be selected, the negligible amounts of frame, and degree of overlapping is than relatively low.It establishes
Comprehensive confidence evaluation index V, judges whether face frame to be selected is sent into the accurate faceform of next step and is examined such as formula (3)
Survey
V=α s+ β n formula (3), wherein α, β are constant, and s is the highest score of multiple high superposed frames, and n is height
The number of frame is overlapped, only when the degree of overlapping of two frames reaches certain threshold value such as 0.65, the two frames are just judged as high superposed frame.Two
Formula (4) Ov=s is shown in the calculating of frame degree of overlapping Ovo/(s1+s2) formula (4)
soFor the overlapping area of two frames, s1、s2For the respective area of two frames.
After the comprehensive evaluation index V of face frame to be selected reaches requirement, then it is further to export the accurate human-face detector of feeding
Confirm.
It is detected by accurate faceform, as shown in figure 3, accurate faceform, is depth convolutional neural networks.It should
Neural network is using the 3 channel RGB images of 40X40 as input.Including 2 convolutional layers, 2 sample levels and 2 full articulamentums.Its
In, first convolutional layer, using the convolution kernel of 16 12X12, second convolutional layer uses the convolution kernel of 16 5X5.Each
There are one sample levels after convolutional layer, and using the mode of max-pooling, the region of 2X2 is handled for step-length with 2.
In the training process of model, the mode for employing bootstrap is trained, i.e., the model trained with the last time, to bearing sample
This picture is detected, and the picture frame that mistake is divided into face cuts to build negative sample sectional drawing, is cut with the negative sample of last time training
Figure merges, for the training of this network.During model training, network weight is updated using the mode of backpropagation.
When merging face frame, using NMS methods (non-maximum value compression), it is judged as the face window of face to exact classification device
Mouth merges, and exports final face frame.
The present invention has cascaded the accurate human-face detector of fast face detector and self study feature based on manual feature.
Wherein, manual features detector is compared based on pixel, and self study is characterized in what neural network learning arrived.Compare for pixel
Feature is easily affected by noise, the situation that Face datection effect is caused to decline, and proposes to judge whether image is high using Graded Density
Noise image, then decide whether to carry out image filtering to eliminate the influence of noise;This method, to normal picture and strong noise image
It is handled differently, avoids and all images are all filtered, improve detection efficiency, also avoid caused by filtering just
The problem of recall rate of normal image declines;By the filtering process to strong noise picture, fast face detector is improved herein
Face recall rate in class image.
Using the highest score and number of high superposed frame, to screen the face window for inputting accurate human-face detector,
While ensureing recall rate, Face datection efficiency is improved.
The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments, it is impossible to assert
The specific implementation of the present invention is confined to these explanations.For those of ordinary skill in the art to which the present invention belongs, exist
Under the premise of not departing from present inventive concept, several simple deduction or replace can also be made, should all be considered as belonging to the present invention's
Protection domain.
Claims (8)
1. a kind of fast accurate method for detecting human face based on cascade structure, it is characterised in that:Include the following steps:(A)Quickly
Faceform detects;(B) face frame is screened;(C)Accurate faceform's detection;(D)Merge face frame and export;The step
(B)In, face frame is screened by the synthesis confidence evaluation index of face frame score and overlapping number, is specially:It establishes comprehensive
Confidence evaluation index V is closed, judges whether face frame to be selected is sent into the accurate faceform of next step by formula V=α * s+ β * n
It is detected, wherein α, β are constant, and s is the highest score of multiple high superposed frames, and n is the number of high superposed frame;Two frame weights
The calculation formula of folded degree Ov is Ov=S0/( S1+ S2), wherein S0For the overlapping area of two frames, S1、S2For the respective face of two frames
Product;Two frames are judged as high superposed frame when the degree of overlapping of two frames reaches certain threshold;When the synthesis confidence of face frame to be selected is commented
After valency index V reaches requirement, then export the accurate human-face detector of feeding and further confirm that.
2. the fast accurate method for detecting human face according to claim 1 based on cascade structure, it is characterised in that:The step
Suddenly(A)In, accelerated model is calculated using the Face datection based on gradation of image comparative feature, decision tree and soft cascade structures
Method carries out model foundation.
3. the fast accurate method for detecting human face according to claim 1 based on cascade structure, it is characterised in that:The step
Suddenly(A)Specially:(A1)Obtain original RGB color image;(A2)Coloured image is converted to single pass gray level image;(A3)
Filtering process;(A4)The size and location of frame is specified search for, does not need to do image pyramid sampling, only search box is put
Big and diminution processing;(A5)The image in search box is sent into each decision tree in trained Face datection model
And obtained score value is handled.
4. the fast accurate method for detecting human face according to claim 1 based on cascade structure, it is characterised in that:The step
Suddenly(C)In, accurate faceform is depth convolutional neural networks, and the neural network is using the 3 channel RGB images of 40X40 as defeated
Enter, which includes 2 convolutional layers, 2 sample levels and 2 full articulamentums.
5. the fast accurate method for detecting human face according to claim 1 based on cascade structure, it is characterised in that:The step
Suddenly(C)In, in the training process of model, with the last model trained, negative sample picture is detected, mistake is divided into
The picture frame of face cuts to build negative sample sectional drawing, and the negative sample sectional drawing of training merges with the last time, for the instruction of this network
Practice.
6. the fast accurate method for detecting human face according to claim 1 based on cascade structure, it is characterised in that:The step
Suddenly(D)In, merge face frame, be judged as that the face window of face merges in being detected to accurate faceform, output is final
Face frame.
7. the fast accurate method for detecting human face according to claim 3 based on cascade structure, it is characterised in that:The step
Suddenly(A3)In, the noise level of image is characterized with Graded Density, using the average gradient of image as threshold value, constructs bianry image,
If the Grad of pixel is more than threshold value, 1 is put, otherwise is set to 0;It will be the 1 corresponding Grad summation of pixel in bianry image;
It in binary map, is scanned with the window of fixed size, step-length is the length of side of window, ensures that each pixel will not be repeated and sweeps
It retouches;In scanning window, if it is 1 to have pixel value, all pixel values are all assigned a value of 1, after the completion of scanning in window, statistical picture
Middle pixel value is the number of 1 pixel;With Grad and divided by the pixel number that counts on, obtain Graded Density.
8. the fast accurate method for detecting human face according to claim 3 based on cascade structure, it is characterised in that:The step
Suddenly(A5)In, if image in search box in the cumulative score after one tree, less than threshold value, then abandons the search box, turns
Enter step(A3)It is handled;If higher than threshold value, continuation is judged in next tree;Figure in a search box
Picture has passed through all decision trees, then exports the frame, forms a face frame to be selected, then screens face frame to be selected.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610847733.8A CN106529398B (en) | 2016-09-23 | 2016-09-23 | Fast accurate method for detecting human face based on cascade structure |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610847733.8A CN106529398B (en) | 2016-09-23 | 2016-09-23 | Fast accurate method for detecting human face based on cascade structure |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106529398A CN106529398A (en) | 2017-03-22 |
CN106529398B true CN106529398B (en) | 2018-06-29 |
Family
ID=58344140
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610847733.8A Active CN106529398B (en) | 2016-09-23 | 2016-09-23 | Fast accurate method for detecting human face based on cascade structure |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106529398B (en) |
Families Citing this family (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107292274A (en) * | 2017-06-28 | 2017-10-24 | 北京飞搜科技有限公司 | A kind of method and system of the detection of the fast face suitable for video |
CN107609465A (en) * | 2017-07-25 | 2018-01-19 | 北京联合大学 | A kind of multi-dimension testing method for Face datection |
CN107766806A (en) * | 2017-09-30 | 2018-03-06 | 北京飞搜科技有限公司 | A kind of method for detecting human face and system based on change of scale |
CN108875502B (en) * | 2017-11-07 | 2021-11-16 | 北京旷视科技有限公司 | Face recognition method and device |
CN108563982B (en) * | 2018-01-05 | 2020-01-17 | 百度在线网络技术(北京)有限公司 | Method and apparatus for detecting image |
CN108509895B (en) * | 2018-03-28 | 2022-09-27 | 百度在线网络技术(北京)有限公司 | Method and device for detecting face image |
CN110390344B (en) * | 2018-04-19 | 2021-10-26 | 华为技术有限公司 | Alternative frame updating method and device |
CN108629330A (en) * | 2018-05-22 | 2018-10-09 | 上海交通大学 | Face dynamic based on multi-cascade grader captures and method for quickly identifying and system |
CN108876809B (en) * | 2018-06-17 | 2021-07-20 | 天津理工大学 | TLD image tracking algorithm based on Kalman filtering |
CN109472223A (en) * | 2018-10-26 | 2019-03-15 | 博康智能信息技术有限公司 | A kind of face identification method and device |
CN109829371B (en) * | 2018-12-26 | 2022-04-26 | 深圳云天励飞技术有限公司 | Face detection method and device |
CN113887541A (en) * | 2021-12-06 | 2022-01-04 | 北京惠朗时代科技有限公司 | Multi-region employee number detection method applied to company management |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7634142B1 (en) * | 2005-01-24 | 2009-12-15 | Adobe Systems Incorporated | Detecting objects in images using a soft cascade |
CN102129572A (en) * | 2011-02-25 | 2011-07-20 | 杭州海康威视软件有限公司 | Face detection method and device adopting cascade classifier |
CN103886308A (en) * | 2014-04-15 | 2014-06-25 | 中南大学 | Pedestrian detection method through soft cascade classifiers according to polymerization channel characteristics |
CN105894506A (en) * | 2016-03-30 | 2016-08-24 | 深圳云天励飞技术有限公司 | Face image fuzziness computing method and device |
-
2016
- 2016-09-23 CN CN201610847733.8A patent/CN106529398B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7634142B1 (en) * | 2005-01-24 | 2009-12-15 | Adobe Systems Incorporated | Detecting objects in images using a soft cascade |
CN102129572A (en) * | 2011-02-25 | 2011-07-20 | 杭州海康威视软件有限公司 | Face detection method and device adopting cascade classifier |
CN103886308A (en) * | 2014-04-15 | 2014-06-25 | 中南大学 | Pedestrian detection method through soft cascade classifiers according to polymerization channel characteristics |
CN105894506A (en) * | 2016-03-30 | 2016-08-24 | 深圳云天励飞技术有限公司 | Face image fuzziness computing method and device |
Also Published As
Publication number | Publication date |
---|---|
CN106529398A (en) | 2017-03-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106529398B (en) | Fast accurate method for detecting human face based on cascade structure | |
CN106960195B (en) | Crowd counting method and device based on deep learning | |
US11263434B2 (en) | Fast side-face interference resistant face detection method | |
CN110348319B (en) | Face anti-counterfeiting method based on face depth information and edge image fusion | |
US5245672A (en) | Object/anti-object neural network segmentation | |
CN108830225B (en) | Method, device, equipment and medium for detecting target object in terahertz image | |
CN107944427B (en) | Dynamic face recognition method and computer readable storage medium | |
CN115205274A (en) | Cloth flaw detection method based on lightweight cascade network | |
CN110929635B (en) | False face video detection method and system based on face cross-over ratio under trust mechanism | |
CN101447023B (en) | Method and system for detecting human head | |
CN109543760A (en) | Confrontation sample testing method based on image filters algorithm | |
CN107292913A (en) | A kind of multistage based on video falls down detection algorithm | |
CN107766806A (en) | A kind of method for detecting human face and system based on change of scale | |
CN101364263A (en) | Method and system for detecting skin texture to image | |
CN103218615B (en) | Face judgment method | |
CN110390673A (en) | Cigarette automatic testing method based on deep learning under a kind of monitoring scene | |
CN106886763A (en) | The system and method for real-time detection face | |
CN104268595B (en) | general object detection method and system | |
CN109190456A (en) | Pedestrian detection method is overlooked based on the multiple features fusion of converging channels feature and gray level co-occurrence matrixes | |
CN105740752B (en) | Sensitive picture filtering method and system | |
CN113076860B (en) | Bird detection system under field scene | |
CN107123109A (en) | A kind of window sliding algorithm detected for Bridge Crack | |
CN112464864A (en) | Face living body detection method based on tree-shaped neural network structure | |
CN116109840B (en) | Cherry spore identification method based on machine vision | |
CN106943116A (en) | A kind of infant eyesight automatic testing method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | 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 |