CN106971168A - Multi-direction multi-level dual crossing robustness recognition methods based on human face structure feature - Google Patents
Multi-direction multi-level dual crossing robustness recognition methods based on human face structure feature Download PDFInfo
- Publication number
- CN106971168A CN106971168A CN201710207911.5A CN201710207911A CN106971168A CN 106971168 A CN106971168 A CN 106971168A CN 201710207911 A CN201710207911 A CN 201710207911A CN 106971168 A CN106971168 A CN 106971168A
- Authority
- CN
- China
- Prior art keywords
- dcp
- image
- face
- robustness
- dual crossing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—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
- G06V40/171—Local features and components; Facial parts ; Occluding parts, e.g. glasses; Geometrical relationships
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/32—Normalisation of the pattern dimensions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
Abstract
The invention discloses a kind of multi-direction multi-level dual crossing robustness identification based on human face structure feature, this method is a kind of completely new approach for extracting facial expression feature, studies two big main components of face recognition:Extraction and face's matching that face is characterized.We are divided to two groups of coded system to extract face sign using on eight directions of inside and outside circle, diesis intersects, and while view face characteristic, simplify amount of calculation.Add multidirectional gradient filtering and multi-level face is characterized, Gauss single order operator derivative converting gradation facial image is used for multidirectional gradient image, mitigate illumination, it is image blurring, block, the influence of posture and these factors of expressing one's feelings so that illumination variation has more robustness.According to the Optimal gradient wave filter of three criterions, i.e. signal to noise ratio（SNR）Maximize, edge positioning-precision is preserved, it is single in response to single edges so that the suffered interference reduction when extracting countenance, so as to accurately extract countenance feature, save and calculate cost.
Description
Technical field
The present invention relates to technical field of image processing, the multi-direction multi-level dual crossing of human face structure feature is based particularly on
Robustness recognition methods.
Background technology
Recognition of face, is a kind of biological identification technology that the facial feature information based on people carries out identification.With shooting
Machine or camera collection image or video flowing containing face, and automatic detect and track face in the picture, and then to detection
The face that arrives carries out a series of correlation techniques of face, generally also referred to as Identification of Images, face recognition.
The research of face identification system is started from the 1960s, with computer technology and optical imagery skill after the eighties
The development of art is improved, and actually enters the primary application stage then 90 year later stage, and with the U.S., Germany and Japan
Based on technology is realized；The successful key of face identification system is the core algorithm for whether possessing tip, and has recognition result
There are practical discrimination and recognition speed；" face identification system " is integrated with artificial intelligence, machine recognition, machine learning, mould
A variety of professional techniques such as type theory, expert system, Computer Vision, while theory and realization that median is handled need to be combined,
It is the more recent application of living things feature recognition, the realization of its core technology presents weak artificial intelligence to the conversion of strong artificial intelligence.
Face identification system mainly includes four parts, is respectively：Man face image acquiring and detection, facial image pretreatment, people
Face characteristic feature is extracted and matching and identification.
Current existing traditional algorithm, because many influences the appearance factor of face-image, is produced aobvious when carrying out feature extraction
Write individual difference, appearance factor for example, illumination variation, postural change, block, image blurring and expression shape change etc. so that people
The accuracy rate of face identification is difficult to improve；In addition, the amount of calculation of existing recognition of face is still huge, the speed of recognition of face has been dragged slowly
Rate, this delay often carrys out very big error to data band, and system gives unfriendly impression.
The content of the invention
The technical problems to be solved by the invention overcome the deficiencies in the prior art and provided based on human face structure feature
Multi-direction multi-level dual crossing robustness recognition methods, the present invention can effectively reduce amount of calculation, improve the accurate of recognition of face
Rate and recognition speed.
The present invention uses following technical scheme to solve above-mentioned technical problem：
According to a kind of multi-direction multi-level dual crossing robustness identification side based on human face structure feature proposed by the present invention
Method, comprises the following steps：
Step 1, input facial image H (xk, yk), H (xk, yk) it is gray level image, facial image training set Pk：{Pk|
1 ..., M }, wherein, M is training set image pattern sum, and k is integer and 1≤k≤M；
If it is d to input distance between two of facial image, using two midpoints as origin of coordinates O (0,0), left and right two
Coordinate value is respectively (- 0.5d, 0), (0.5d, 0), into step 2；
Step 2, the facial image to step 1, on the basis of O, using left and right it is each away from O as d, vertical two line directions it is upper
Direction takes composition rectangular area between 0.5d and lower direction 1.5d to be sheared, and obtains rectangular area, and the rectangular area is expression
Subregion, the image of the rectangular area after shearing is expression sub-district area image；
Step 3, the expression sub-district area image to step 2 carry out entering step 4 after change of scale, uniform sizes and specification；
Step 4, to each pixel in the expression sub-district area image after uniform sizes, inside and outside two are drawn using it as the center of circle
Individual circle, respectively take 0 on inside and outside circle,π、WithThis eight directions；
Step 5, the texture information on eight directions of pixel in step 4 is quantified, distribute one unique ten
System number：Defined function S (y)：Y is to become
Amount, Io、It is point 0, A respectivelyi、BiGray value；Ai、BiPixel inner circle in step 4 respectively, it is cylindrical on point,
DCPiIt is the texture information amount on the i-th direction；The directions of 0- the 7th refer to 0 respectively,π、With
Step 6, for the DCP in step 5i, define { DCP0, DCP2, DCP4, DCP6It is used as the first subset, { DCP1,
DCP3, DCP5, DCP7It is used as yield in the second subset；The two subsets build the shape of a cross, constitute dual crossing pattern, occur most
Big combination entropy；
Step 7, two subsets in step 6 are grouped into two cross encoders, are respectively designated as DCP-1, DCP-2,
The code of pixel in expression sub-district area image is expressed as：
Two cross encoders constitute total descriptor DCP:DCP={ DCP-1, DCP-2 }, produces two by total descriptor and is mapped respectively
Into the trellis diagram of Non-overlapping Domain；All histograms in each region are calculated in this trellis diagram, and will be all straight
Square figure is connected, and overall face representation framework is formed, into step 8；
Step 8, by the overall face representation framework eventually formed in step 7 by Gaussian filter, using central point O as original
Point, uses the first derivative formula of Gauss operator：N=(cos θ, sin θ) is to represent filtering direction
Normal vector, θ is the angle for filtering direction,It is dimensional Gaussian Multi-aspect filtering device, xk、ykIt is face
Image H (xk, yk) coordinate value, σ is the variance of the function；
Step 9, using being used as training data, initialization training sample by the face representation overall framework of wave filter in step 8
This weights distribution, i.e., be endowed identical weights i.e. when each training sample most starts：D1=(w11, w12…
w1j…,w1M),D1It is training set initialization weights distribution, w1jFor j-th of sample power of initialization
Weight；
Step 10, the weight to training sample in step 9 carry out many wheel iteration；
101st, m is set as iterations, and during initialization iterations m=1, it is D that weights are initialized this moment1；
102nd, it is distributed D using with weightsmTraining dataset study, DmFor the sample weights after the m times iteration；Selection allows mistake
The minimum threshold value of rate designs basic classification device Gm(x)：Gm(x):X → { -1,1 }, x is variable, Gm(x) on training dataset
Error in classification rate em, emIt is by Gm(x) sample weights of misclassification and, and
wmjJ-th of sample weight when being the m times iteration, I (Gm(xk)≠yk) it is Gm(xk)≠ykThe gray value at place；
If the 103, m ＜ M, m=m+1, otherwise repeat step 102 stops circulation；
G in step 11, calculation procedure 10m(x) coefficient, obtains basic classification device Gm(x) institute in final classification device G (x)
The weight α accounted form：αm≥0；So as to obtain final classification device G (x)：Multiple face feature points are finally given, so as to form multi-direction multistage dual crossing pattern Shandong
Rod recognition of face.
It is used as a kind of multi-direction multi-level dual crossing robustness identification side based on human face structure feature of the present invention
, there is maximum combined entropy in the further prioritization scheme of method, the dual crossing packet mode based on facial symmetry feature that step 6 is used,
Robustness mode is formed, into step 7 dual crossing coding.
It is used as a kind of multi-direction multi-level dual crossing robustness identification side based on human face structure feature of the present invention
The further prioritization scheme of method, step 8 uses gaussian filtering,It is dimensional Gaussian Multi-aspect filtering device, makes
Obtain gray scale face-image and be converted to the multi-direction gradient image for having robustness to illumination change, signal noise ratio (snr) of image is maximized, increase
Image robustness.
It is used as a kind of multi-direction multi-level dual crossing robustness identification side based on human face structure feature of the present invention
In the further prioritization scheme of method, step 8, θ ∈ (0, π).
It is used as a kind of multi-direction multi-level dual crossing robustness identification side based on human face structure feature of the present invention
In the further prioritization scheme of method, step 10, it is 200 to take training set image pattern sum M.
The present invention uses above technical scheme compared with prior art, with following technique effect：
(1) weak classifier set it is determined that when face feature point, using iterative algorithm, constituted most strong point by the present invention
Class device, mitigates on error in classification, " focusing on " those more difficult point samples；
(2) present invention is summarised eight directions, successfully make use of the symmetrical of face using the coding method of dual crossing pattern
Architectural characteristic, simplifies algorithm steps, reduces amount of calculation；
(3) present invention utilizes Gaussian filter transition diagram picture, solves traditional algorithm when carrying out feature extraction because being permitted
The factor influences for making the outward appearance of face-image produce notable individual difference more, for example, illumination variation, postural change, block, image
Fuzzy and expression shape change etc. so that gray scale face-image is converted to the multi-direction gradient image for having robustness to illumination change,
Maximize signal noise ratio (snr) of image, marginal position precision keeps and becomes unilateral from single response；
Brief description of the drawings
Fig. 1 is the flow of the multi-direction multi-level dual crossing robustness recognition methods of the invention based on human face structure feature
Figure.
Fig. 2 is face's geometrical normalization center.
Fig. 3 is local sampling point set.
Fig. 4 is dual crossing block encoding；Wherein, (a) be by 0,π,Cross that this four direction is constituted
Collection, (b) serves as reasonsAnother cross subset that this four direction is constituted.
Fig. 5 is multiple characteristic points of image face.
Embodiment
Technical scheme is described in further detail below in conjunction with the accompanying drawings：
In specific implementation, Fig. 1 is the multi-direction multi-level dual crossing robustness recognition methods based on human face structure feature
Flow, input facial image H (xk, yk), H (xk, yk) it is gray level image, facial image training set Pk：{Pk| 1 ..., M }, its
In, M is training set image pattern sum, and k is integer and 1≤k≤M；If it is d to input distance between two of facial image, with
Two midpoints are origin of coordinates O (0,0), and the coordinate value that left and right is two is respectively (- 0.5d, 0), (0.5d, 0).
To the above-mentioned facial image for having marked coordinate value, on the basis of O, using left and right respectively away from O as d, vertical two companies
The upper direction in line direction takes composition rectangular area between 0.5d and lower direction 1.5d to be sheared, and obtains rectangular area, the rectangle
Expression subregion is in region, and the image of the rectangular area after shearing is expression sub-district area image, as shown in Figure 2.So more have
Beneficial to the extraction of expressive features.
Then expression sub-district area image is carried out after change of scale, uniform sizes and specification, in expression sub-district area image
Pixel, inside and outside two circles are drawn using it as the center of circle, respectively take 0 on inside and outside circle,π、WithThis
Eight directions.
Then the texture information on eight directions of pixel is quantified, distributes a unique decimal number：Defined function S (y)：Y is variable, Io、It is point 0, A respectivelyi、BiGray value；Ai、BiPixel inner circle in step 4 respectively, it is cylindrical on point, DCPiIt is i-th
Texture information amount on direction；0-7 directions refer to 0 respectively,π、WithAs shown in Figure 3.
For above-mentioned DCPi, define { DCP0, DCP2, DCP4, DCP6It is used as the first subset, { DCP1, DCP3, DCP5,
DCP7It is used as yield in the second subset；The two subsets build the shape of a cross, constitute dual crossing pattern, maximum combined entropy occur,
As shown in (a), (b) in Fig. 4.
The two subsets are grouped into two cross encoders, DCP-1, DCP-2 is respectively designated as, in expression subregion figure
The code of pixel as in is expressed as：Two intersect volume
Code device constitutes total descriptor DCP:DCP={ DCP-1, DCP-2 }, produces two by total descriptor and is mapped to non-overlapped area respectively
The trellis diagram in domain；All histograms in each region are calculated in this trellis diagram, and all histograms are connected
Come, form overall face representation framework.
Above-mentioned row into overall face representation framework by Gaussian filter, using central point O as origin, calculated using Gauss
The first derivative formula of son：N=(cos θ, sin θ) is the normal vector for representing filtering direction, and θ is filtering
The angle in direction,It is dimensional Gaussian Multi-aspect filtering device, xk、ykIt is facial image H mentioned above
(xk, yk) coordinate value, σ is the variance of the function.By wave filter so that gray scale face-image is converted to illumination change
There is the multi-direction gradient image of robustness, maximize signal noise ratio (snr) of image.
Then it assign the face representation overall framework by wave filter as training data, the weights point of initialization training sample
Cloth, i.e., be endowed identical weights i.e. when each training sample most starts：D1=(w11, w12…w1j…,w1M),D1It is training set initialization weights distribution, w1jTo initialize j-th of sample weights；
Then the weight to above-mentioned training sample carries out many wheel iteration：
(1) m, is set as iterations, during note primary iteration number of times m=1, the D in step 91To initialize weights this moment,
That is DmValue as m=1；
(2), it is distributed D using with weightsmTraining dataset study, DmFor the sample weights after the m times iteration；Selection allows mistake
The minimum threshold value of rate designs basic classification device Gm(x)：Gm(x):X → { -1,1 }, x is variable, Gm(x) on training dataset
Error in classification rate em, emIt is by Gm(x) sample weights of misclassification and, and
wmjJ-th of sample weight when being the m times iteration, I (Gm(xk)≠yk) it is Gm(xk)≠ykThe gray value at place；
(3) if, m ＜ M, m=m+1, repeat step (2) otherwise stops circulation.
Finally, G is calculatedm(x) coefficient, obtains basic classification device Gm(x) weight shared in final classification device G (x)
αm：αm≥0；So as to obtain final classification device G (x)：
Multiple face feature points are finally given, so as to form multi-direction multistage dual crossing pattern robustness recognition of face, as shown in Figure 5.
The present invention not only solves traditional algorithm when carrying out feature extraction because many produces the outward appearance of face-image
The factor influence of notable individual difference, for example, illumination variation, postural change, block, image blurring and expression shape change etc..Also
Amount of calculation is alleviated, the accuracy rate and speed of recognition of face is improved.
Claims (5)
1. a kind of multi-direction multi-level dual crossing robustness recognition methods based on human face structure feature, it is characterised in that including
Following steps：
Step 1, input facial image H (xk, yk), H (xk, yk) it is gray level image, facial image training set Pk：{Pk|1,...,
M }, wherein, M is training set image pattern sum, and k is integer and 1≤k≤M；
If it is d to input distance between two of facial image, using two midpoints as origin of coordinates O (0,0), the coordinate that left and right is two
Value is respectively (- 0.5d, 0), (0.5d, 0), into step 2；
Step 2, the facial image to step 1, on the basis of O, using left and right respectively away from O as d, the upper direction in vertical two line directions
Take composition rectangular area between 0.5d and lower direction 1.5d to be sheared, obtain rectangular area, the rectangular area is expression sub-district
Domain, the image of the rectangular area after shearing is expression sub-district area image；
Step 3, the expression sub-district area image to step 2 carry out entering step 4 after change of scale, uniform sizes and specification；
Step 4, to each pixel in the expression sub-district area image after uniform sizes, inside and outside two are drawn using it as the center of circle
Circle, respectively take 0 on inside and outside circle,π、WithThis eight directions；
Step 5, the texture information on eight directions of pixel in step 4 is quantified, distribute a unique decimal system
Number：Defined function S (y)：Y is variable, Io、It is point 0, A respectivelyi、BiGray value；Ai、BiPixel inner circle in step 4 respectively, it is cylindrical on point, DCPiIt is i-th
Texture information amount on direction；The directions of 0- the 7th refer to 0 respectively,π、WithStep 6, for step
DCP in rapid 5i, define { DCP0, DCP2, DCP4, DCP6It is used as the first subset, { DCP1, DCP3, DCP5, DCP7It is used as second
Subset；The two subsets build the shape of a cross, constitute dual crossing pattern, maximum combined entropy occur；
Step 7, two subsets in step 6 are grouped into two cross encoders, DCP-1, DCP-2 are respectively designated as, in table
The code of pixel in feelings sub-district area image is expressed as：Two
Individual cross encoder constitutes total descriptor DCP:DCP={ DCP-1, DCP-2 }, produces two by total descriptor and is mapped to respectively
The trellis diagram of Non-overlapping Domain；Calculate all histograms in each region in this trellis diagram, and by all Nogatas
Figure is connected, and overall face representation framework is formed, into step 8；
Step 8, by the overall face representation framework eventually formed in step 7 by Gaussian filter, using central point O as origin,
Use the first derivative formula of Gauss operator：It is the normal direction for representing filtering direction
Amount, θ is the angle for filtering direction,It is dimensional Gaussian Multi-aspect filtering device, xk、ykIt is facial image H
(xk, yk) coordinate value, σ is the variance of the function；
Step 9, using in step 8 by the face representation overall framework of wave filter as training data, initialization training sample
Weights are distributed, i.e., identical weights are endowed when each training sample most starts i.e.：D1=(w11, w12…w1j…,
w1M),D1It is training set initialization weights distribution, w1jTo initialize j-th of sample weights；
Step 10, the weight to training sample in step 9 carry out many wheel iteration；
101st, m is set as iterations, and during initialization iterations m=1, it is D that weights are initialized this moment1；
102nd, it is distributed D using with weightsmTraining dataset study, DmFor the sample weights after the m times iteration；Selection allows error rate
Minimum threshold value designs basic classification device Gm(x)：Gm(x):X → { -1,1 }, x is variable, Gm(x) classification on training dataset
Error rate em, emIt is by Gm(x) sample weights of misclassification and, and
wmjJ-th of sample weight when being the m times iteration, I (Gm(xk)≠yk) it is Gm(xk)≠ykThe gray value at place；
If the 103, m ＜ M, m=m+1, otherwise repeat step 102 stops circulation；
G in step 11, calculation procedure 10m(x) coefficient, obtains basic classification device Gm(x) weight shared in final classification device G (x)
αm：αm≥0；So as to obtain final classification device G (x)：
Multiple face feature points are finally given, so as to form multi-direction multistage dual crossing pattern robustness recognition of face.
2. a kind of multi-direction multi-level dual crossing robustness identification side based on human face structure feature according to claim 1
, there is maximum combined entropy in method, it is characterised in that the dual crossing packet mode based on facial symmetry feature that step 6 is used, is formed
Robustness mode, into step 7 dual crossing coding.
3. a kind of multi-direction multi-level dual crossing robustness identification side based on human face structure feature according to claim 1
Method, it is characterised in that step 8 uses gaussian filtering,It is dimensional Gaussian Multi-aspect filtering device so that
Gray scale face-image is converted to the multi-direction gradient image for having robustness to illumination change, and signal noise ratio (snr) of image is maximized, increase figure
As robustness.
4. a kind of multi-direction multi-level dual crossing robustness identification side based on human face structure feature according to claim 1
Method, it is characterised in that in step 8, θ ∈ (0, π).
5. a kind of multi-direction multi-level dual crossing robustness identification side based on human face structure feature according to claim 1
Method, it is characterised in that in step 10, it is 200 to take training set image pattern sum M.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710207911.5A CN106971168A (en) | 2017-03-31 | 2017-03-31 | Multi-direction multi-level dual crossing robustness recognition methods based on human face structure feature |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710207911.5A CN106971168A (en) | 2017-03-31 | 2017-03-31 | Multi-direction multi-level dual crossing robustness recognition methods based on human face structure feature |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106971168A true CN106971168A (en) | 2017-07-21 |
Family
ID=59336387
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710207911.5A Pending CN106971168A (en) | 2017-03-31 | 2017-03-31 | Multi-direction multi-level dual crossing robustness recognition methods based on human face structure feature |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106971168A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107577994A (en) * | 2017-08-17 | 2018-01-12 | 南京邮电大学 | A kind of pedestrian based on deep learning, the identification of vehicle auxiliary product and search method |
| CN112464901A (en) * | 2020-12-16 | 2021-03-09 | 杭州电子科技大学 | Face feature extraction method based on gradient face local high-order main direction mode |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103824052A (en) * | 2014-02-17 | 2014-05-28 | 北京旷视科技有限公司 | Multilevel semantic feature-based face feature extraction method and recognition method |
| CN105893941A (en) * | 2016-03-28 | 2016-08-24 | 电子科技大学 | Facial expression identifying method based on regional images |
-
2017
- 2017-03-31 CN CN201710207911.5A patent/CN106971168A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103824052A (en) * | 2014-02-17 | 2014-05-28 | 北京旷视科技有限公司 | Multilevel semantic feature-based face feature extraction method and recognition method |
| CN105893941A (en) * | 2016-03-28 | 2016-08-24 | 电子科技大学 | Facial expression identifying method based on regional images |
Non-Patent Citations (2)
| Title |
|---|
| CHANGXING DING等: "Multi-Directional Multi-Level Dual-Cross Patterns for Robust Face Recognition", 《IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE》 * |
| 严政: "人脸表情运动单元识别系统", 《中国优秀硕士论文全文数据库信息科技辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107577994A (en) * | 2017-08-17 | 2018-01-12 | 南京邮电大学 | A kind of pedestrian based on deep learning, the identification of vehicle auxiliary product and search method |
| CN112464901A (en) * | 2020-12-16 | 2021-03-09 | 杭州电子科技大学 | Face feature extraction method based on gradient face local high-order main direction mode |
| CN112464901B (en) * | 2020-12-16 | 2024-02-02 | 杭州电子科技大学 | Face feature extraction method based on gradient face local high-order main direction mode |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105975931B (en) | A kind of convolutional neural networks face identification method based on multiple dimensioned pond | |
| CN108520216B (en) | Gait image-based identity recognition method | |
| Zhu et al. | Lesion detection of endoscopy images based on convolutional neural network features | |
| CN104517104B (en) | A kind of face identification method and system based under monitoring scene | |
| CN103605972B (en) | Non-restricted environment face verification method based on block depth neural network | |
| CN104778457B (en) | Video face identification method based on multi-instance learning | |
| CN107316031A (en) | The image characteristic extracting method recognized again for pedestrian | |
| Zhang et al. | GAN-based image augmentation for finger-vein biometric recognition | |
| CN108304826A (en) | Facial expression recognizing method based on convolutional neural networks | |
| CN103927511B (en) | image identification method based on difference feature description | |
| CN105718889B (en) | Based on GB (2D)2The face personal identification method of PCANet depth convolution model | |
| Herdiyeni et al. | Combination of morphological, local binary pattern variance and color moments features for indonesian medicinal plants identification | |
| CN106529504B (en) | A kind of bimodal video feeling recognition methods of compound space-time characteristic | |
| CN107122712B (en) | Palm print image identification method based on CNN and bidirectional VLAD | |
| CN107992850B (en) | Outdoor scene three-dimensional color point cloud classification method | |
| CN103679136B (en) | Hand back vein identity recognition method based on combination of local macroscopic features and microscopic features | |
| Liu et al. | Finger vein recognition with superpixel-based features | |
| CN106778768A (en) | Image scene classification method based on multi-feature fusion | |
| CN108615007B (en) | Three-dimensional face identification method, device and storage medium based on characteristic tensor | |
| CN107886558A (en) | A kind of human face expression cartoon driving method based on RealSense | |
| CN115830652B (en) | Deep palm print recognition device and method | |
| CN109800677A (en) | A kind of cross-platform palm grain identification method | |
| Khan et al. | Texture representation through overlapped multi-oriented tri-scale local binary pattern | |
| CN106203448A (en) | A kind of scene classification method based on Nonlinear Scale Space Theory | |
| CN110188646B (en) | Human ear identification method based on fusion of gradient direction histogram and local binary pattern |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170721 |
|
| RJ01 | Rejection of invention patent application after publication |