CN105956515A - Stereo-hyperspectral human face recognition method based on auroral imaging - Google Patents
Stereo-hyperspectral human face recognition method based on auroral imaging Download PDFInfo
- Publication number
- CN105956515A CN105956515A CN201610247977.2A CN201610247977A CN105956515A CN 105956515 A CN105956515 A CN 105956515A CN 201610247977 A CN201610247977 A CN 201610247977A CN 105956515 A CN105956515 A CN 105956515A
- Authority
- CN
- China
- Prior art keywords
- face
- image
- human face
- pole
- regional area
- 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
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
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- 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 Processing (AREA)
Abstract
The invention provides a stereo-hyperspectral human face recognition method based on auroral imaging, for solving the technical problems of poor recognition result stability and low recognition efficiency of conventional multi-spectral human face recognition methods. The stereo-hyperspectral human face recognition method comprises the following steps of: 1, using an auroral spectrum system to collect human face Information; 2, carrying out stereo reconstruction of a human face according to a two-dimensional auroral image in the human face information; 3, diving the two-dimensional human face into regional blocks, and dividing the three-dimensional human face and human face spectrum information into regional blocks according to the division result; 4, extracting comprehensive feature vectors of the regional blocks; 5, for many human faces and the to-be-recognized human face, adopting the step 1 to the step 4 to obtain sample library data and to-be-recognized human face data; 6, training the sample library data to get a classifier; 7, inputting the to-be-recognized human face data to the classifier, and identifying the human face. The stereo-hyperspectral human face recognition method of the invention has the characteristics of good stability and high identification efficiency, and can be used to authenticate an identity for financial transaction, security defense, customs clearance and the like.
Description
Technical field
The invention belongs to biometrics identification technology field, relate to a kind of face identification method, be specifically related to one
Solid-EO-1 hyperion face identification method based on pole imaging, can be used for financial transaction, safe defence, clearance peace
The identity authentication of inspection etc..
Background technology
Live 6,700,000,000 appearances are different, the colour of skin the is different mankind in the world, and face is as an important biology
Feature, has critical role in interpersonal exchange.In society, face has become as a kind of weight
The feature for identity authentication wanted.Face identification system can be divided into three kinds by function: towards security type, pipe
Reason supervision class and multimedia recreation class, the requirement to recognition accuracy reduces the most successively.It is reported, every in the U.S.
Year has government's welfare fund of billions of dollars illegally to be falsely claimed as one's own, it addition, because lacking perfect identification system,
Also the multiple social safety problem including that the attack of terrorism threatens is occurred in that.The continuous attack of terrorism occurred in recent years
Tragedy, while strengthening security implementations, border control, immigration to various countries are investigated ability and have also been beaten alarm bell.
Therefore, research has clear and definite, urgent grinding towards the face recognition technology more accurate, efficient of security type
Study carefully meaning.
Face recognition process is generally described as, a given static state or dynamic facial image, utilizes existing
Face database determine the identity of one or more people in image.It mainly solves three tasks: body
Part certification, identification, monitoring.Authentication is to judge that whether face in image is the identity that he is stated,
Have only to the template image phase comparison of this identity stored in the facial image of input and data base, be a pair
The comparison of one；Identification is to utilize face database to determine the identity of face in image, needs input
Facial image is compared with all of identity template image in storehouse and provides similarity, differentiates that input face is
Which identity (similarity is the highest) in storehouse, is the comparison of one-to-many.Supervision refers to find in monitoring range
Face is also followed the tracks of and is split, and judge this face whether on watch-list, if it does, enter one again
Step confirms their identity.
The basic procedure of recognition of face mainly includes that data acquisition, Face datection, face registration, face characteristic carry
Take and the process such as recognition of face.Imageing sensor such as CCD or CMOS etc. is used to obtain facial image or video,
Improve picture quality by pretreatment, then carry out Face datection, Attitude estimation and positioning feature point, i.e. from figure
Be partitioned into human face region interested in Xiang, human face region registrated and normalization, then to normalization after
Facial image carries out feature extraction, is input in grader mate with template image feature vector in face database
Calculate, finally export recognition result.
Recognition of face can be divided into two big classes from the method describing face: based on geometric properties face representation method
With face representation method based on texture.Recognition methods based on face geometry utilizes face to form in structure
On characteristic information, ask for edge and profile as feature from face two dimensional image, and then utilize aspect ratio pair
Similarity between mode judging characteristic.Substantially, the information extracted from two-dimension picture is to lost true three
Subset after dimension face depth dimension information, and the resemblance of face instability, easily by illumination and expression
The impact of change, therefore geometry recognition methods stability based on two-dimension human face image is the highest, in large area
Shadow effect under even can lose efficacy.Thus, over the past two years based on three-dimensional or the recognition of face of depth geometry feature
Method is suggested, and obtains the human face three-dimensional model with depth information by the technology such as laser scanning, three-dimensional imaging,
And from model, extract effective feature, such as use the information such as forehead curvature, nose height composition to describe three-dimensional
The characteristic vector of face, and then be identified by the method for template matching.The introducing of Spatial Multi-Dimensional degree information becomes
The face characteristic that solves of merit is affected by illumination variation, makes recognition accuracy reach more than 90%.
On the other hand, researchers both domestic and external have also carried out substantial amounts of recognition of face side based on texture information
Method research.This type of method ties up information from the spectrum of face, utilizes spectrum in the regularity of face surface distributed
With stationarity as the feature description of face, and then Land use models is known method for distinguishing and is completed recognition of face.Act on
The light of face is divided into actively and passively light source, and the wavelength band of light source light spectrum also expands to near-infrared from visible ray,
The resolving range of spectrum is from polychrome, multispectral again to EO-1 hyperion.Rely on theory and the skill of statistical-simulation spectrometry
The development of art, emerges substantial amounts of face recognition algorithms based on texture information.Nineteen nineties is carried
" eigenface (Eignface) " that go out is representative recognition methods.So far " eigenface " method is still
Old is for one of benchmark algorithm evaluating and testing new algorithm in field of face identification, is applied to each spectral band
Recognition of face.Along with method quilts such as rise and the fast development of machine learning techniques, neutral net, degree of depth study
It is applied in Face datection and is successful, hard along with high-performance computer the most over the past two years
The popularization of part, artificial intelligence technology has had great breakthrough, utilizes machine learning to obtain after extensive sample training
Obtain grader well, and then facial image or video are detected and identifies.Although, " machine learning "
Application development like a raging fire, but processing in human face detection and tracing problem or existing the most not enough
: need substantial amounts of Training data as support, to improve the generalization ability of system；The most existing
Face database has the little of complete tag information, limits major part supervised learning algorithm；Model training
Time and resource consumption are very big, as wanted adjusting training strategy and parameter, need the cycle grown very much.
At present, the face identification method that researcher is used is developed to higher dimensional space by low-dimensional from the most,
Develop to broadband or even height (many) spectrum by monochrome greyscale image from the most.On the one hand, by swashing
The technology such as photoscanning, three-dimensional imaging obtains with the human face three-dimensional model of depth information, and extracts from model and have
The geometric properties of effect, the introducing of Spatial Dimension information solves face characteristic to be affected by illumination and expression shape change.
On the other hand, utilize the optical instrument of glistening light of waves sheet or Amici prism to obtain the spectral information of face, and from obtaining
Extracting the spectral signature of face local in face spectral information, the extension of spectrum dimensional information solves recognition of face
Affected by shade and partial occlusion factor.Such as, Chinese patent application, Authorization Notice No. is CN102831400
B, the patent of invention of entitled " a kind of multispectral face identification method and system thereof ", this invention includes spectrum
Imaging system, color camera, face recognition module, data memory module, Central Control Module and spectrogrph,
Multi-optical spectrum imaging system by the face image data output of shooting to face recognition module, face recognition module according to
The information of the standard faces data base in data memory module is identified, and then carries out defeated by the result of identification
Going out, Central Control Module controls image capture and the identification of face recognition module of multi-optical spectrum imaging system.This
Although obvious between class distance can be provided during identifying bright, but still suffering from is owing to making less than one at two
Face image data is gathered, it is impossible to obtain the stereochemical structure information of face, make identification tie with multi-optical spectrum imaging system
The impact that fruit is easily subject to illumination variation, coats and block, causes recognition result poor stability, additionally this invention
Owing to needing manual operation when intercepting human face region sample, reduce the efficiency of recognition of face.
Summary of the invention
It is an object of the invention to the defect overcoming above-mentioned prior art to exist, it is proposed that a kind of based on pole imaging
Solid-EO-1 hyperion face identification method, is used for solving identification knot present in existing multispectral face identification method
Really poor stability and the low technical problem of recognition efficiency.
For achieving the above object, the technical scheme that the present invention takes, comprise the steps:
Step 1, utilizes auroral spectrum system acquisition human face data, obtains the face two-dimentional pole figure at each spectral band
The spectral information that on picture and this two dimension pole image, pixel is corresponding；
Step 2, filters out the most much higher spectral band correspondence of quality from the two-dimentional pole image of the face obtained
Pole image, the multiple poles image filtered out is merged, and to merge after pole image carry out affine transformation,
Obtain the image of three different points of view of face；
Step 3, utilizes multi-viewpoint three-dimensional matching process, founds the image obtaining three different points of view of face
Body weight structure, obtains the three dimensional structure of face；
Step 4, divides the full two-dimensional face in the pole image after described fusion, obtain face with
The two-dimentional regional area block of different tag along sorts；
Step 5, according to the face obtained with the two-dimentional regional area block of different tag along sorts, to described face
The three dimensional structure spectral information corresponding with pixel on described two dimension pole image carry out regional area block division,
To the multiple regional area blocks with tag along sort comprising three dimensional structure and spectral information；
Step 6, to each regional area block in the multiple regional area blocks obtained, calculates the spy of three dimensional structure
Levy vector, extract the characteristic vector of spectral information simultaneously, and the two characteristic vector is merged, obtain people
Multiple Local synthesis characteristic vectors with tag along sort of face；
Multiple sample faces are used step 1 to step 6, obtain comprising multiple sample face many by step 7
The face Sample Storehouse of the individual Local synthesis characteristic vector with tag along sort；
The face Sample Storehouse data obtained are input in linear SVM model be trained by step 8,
Obtain the grader for different target type；
Step 9, uses face to be identified step 1 to step 6, obtains the multiple with dividing of face to be identified
The Local synthesis characteristic vector of class label；
Step 10, the multiple Local synthesis characteristic vectors with tag along sort of face to be identified that will obtain,
It is input in described grader classify, obtains the target class belonging to face to be identified multiple regional area block
Type, and whether belong to same target type according to multiple regional area blocks of face to be identified, it is judged that this face is
No it is consistent with Sample Storehouse.
The present invention compared with prior art, has the advantage that
1, due to the fact that the auroral spectrum system that make use of when gathering human face data, and according to obtaining two dimension pole figure
As face is carried out three-dimensionalreconstruction, it is possible to realize face characteristic obtains while three dimensional structure and spectral information,
With prior art utilizes multi-optical spectrum imaging system gather human face data, and obtain face two-dimensional geometry feature and
Spectral information is compared, recognition result not by illumination variation, coat and affected with blocking, have higher stability.
2, due to the fact that when full two-dimensional face is divided, it is not necessary to manually participate in, with existing skill
The method of the Manual interception human face region sample that art uses is compared, and improves the efficiency of recognition of face.
Accompanying drawing explanation
Fig. 1 is the FB(flow block) of the present invention；
Fig. 2 is the auroral spectrum system on human face data acquisition schematic diagram in the present invention；
Fig. 3 is pole imaging Multi-visual point image schematic diagram；
Fig. 4 is face multiple regional area block schematic diagram in the present invention.
Detailed description of the invention
Below in conjunction with the drawings and specific embodiments, the invention will be further described.
With reference to Fig. 1, solid-EO-1 hyperion face identification method based on pole imaging, comprise the steps:
Step 1, utilizes auroral spectrum system acquisition human face data, as in figure 2 it is shown, obtain face at each spectrum ripple
The spectral information that on the two-dimentional pole image of section and this two dimension pole image, pixel is corresponding, gathers each spectral band
Human face data can more fully reflect face characteristic, makes recognition result more accurate；
Step 2, filters out the most much higher spectral band correspondence of quality from the two-dimentional pole image of the face obtained
Pole image, the most much higher pole image of the quality filtered out is merged, and to merge after pole image enter
Row affine transformation, extracts the RADIAL (polar curve) in the image of pole after merging when affine transformation, and level is stacked
Become a new image, according to the pole imaging Multi-visual point image principle shown in Fig. 3, new images is pressed in the horizontal direction
Ratio is divided into three parts, three viewpoints of the most corresponding hyperspectral imager, thus obtains three different points of view of face
Image；
The pole image filtering out the most much higher spectral band of quality corresponding realizes in accordance with the following steps:
Step 2a, utilize Y-PSNR function (Peak Signal-to-Noise Ratio, PSNR) and
Evaluation function based on structural similarity (Structural Similarity Image Measurement,
SSIM), the pole image of spectral band each to face carries out quality evaluation, obtains face each spectral band pole image
Y-PSNR and scoring based on structural similarity, Y-PSNR function can obtain EO-1 hyperion efficiently
The less wave band of noise in image, and evaluation function based on structural similarity from the brightness of image, contrast and
The content quality of three factors evaluation high spectrum images of structure, the purpose of fusion mass higher band is to obtain spy
Levy a little obvious single width two dimensional image, provide sufficient characteristic point to be used for Stereo matching for the reconstruct of the degree of depth below；
Step 2b, the Y-PSNR of spectral band pole each to the face obtained image and based on structural similarity
Scoring carry out quantifying and linear superposition, obtain the quality score of each spectral band pole image, and from each spectrum ripple
Filtering out, in the image of section pole, the pole image that the high spectral band of at least two quality score is corresponding, the present invention filters out
Three pole images, if filtering out too much image can not reach high-quality effect, screening image very little then can
Make fusion image too unification, it is impossible to complete description human face data；
Step 3, utilizes multi-viewpoint three-dimensional matching process, uses the GPU parallel algorithm face to obtaining three not
Carry out stereo reconstruction with the image of viewpoint, obtain the depth information that pixel in two-dimension human face is corresponding, i.e. obtain people
The three dimensional structure of face；
Step 4, owing to the circular central region of face two dimension pole image is that face passes through the true of hyperspectral imager
The picture that real viewpoint becomes, this area image can accurately show two-dimension human face, by dividing the full two-dimensional in this region
Facial image, can be to divide face three dimensional structure and spectral information offer partitioning standards the most accurately below,
So the full two-dimensional face in the two-dimentional pole image after described fusion is divided, obtain face with difference
The two-dimentional regional area block of tag along sort；
The realization in accordance with the following steps of division full two-dimensional face:
Step 4a, after the face fusion that will obtain, the circular image comprising complete face in the image of pole is partitioned into
Come, obtain complete two-dimension human face image；
Step 4b, is sharpened process to the complete two-dimension human face image obtained, and after detecting Edge contrast
The marginal point of two-dimension human face image, obtains two-dimension human face image marginal point；
Step 4c, clusters the two-dimension human face image marginal point obtained, obtains in two-dimension human face image many
The individual classification comprising marginal point, cluster here is to divide classification by the difference of Different Organs on face,
Due to the organ notable difference to spectral reflectivity such as eyes, nose, lip on face, so being easy to pass through
Facial image carrys out zoning block, and face is divided in the present invention 9 classifications；
Step 4d, the central point of multiple classifications comprising marginal point in the most calculated two-dimension human face image
Position, and determine its tag along sort and area size according to the position of central point of all categories, such as, for one
Class center point is in leftmost classification, and the present invention gives this classification eyes label, and sets its classification district
Territory is shaped as rectangle, and size is the average long and wide of normal person's eyes, obtains face with different tag along sorts
Two-dimentional regional area block；
Step 5, according to the face obtained with the two-dimentional regional area block of different tag along sorts, to described face
Three dimensional structure (two dimensional image combine depth information composition face steric information) and described two-dimentional pole image on picture
The spectral information that vegetarian refreshments is corresponding carries out regional area block division, this spectral information when dividing only to face pole image
The complete face of central circular divides, obtain comprising three dimensional structure and spectral information with contingency table
The multiple regional area blocks signed, as shown in Figure 4；
Step 6, in multiple regional area blocks of the face obtained, utilizes average height and the ladder of regional area
The characteristic vector of degree composition three dimensional structure, uses each district of HySime algorithm rapid extraction based on GPU simultaneously
The spectrum n-dimensional subspace n of territory block, it is thus achieved that the characteristic vector of spectral information, and two spies to each regional area block
Levy vector to merge, obtain multiple Local synthesis characteristic vectors with tag along sort of face；
Multiple sample faces are used step 1 to step 6, obtain comprising multiple sample face many by step 7
The face Sample Storehouse of the individual Local synthesis characteristic vector with tag along sort；
Step 8, the data volume of the Local synthesis characteristic vector of face three dimensional structure and spectral information is big, people's work point
Class region unit is time-consuming, laborious, unrealistic, and the present invention uses linear SVM model, by face Sample Storehouse
In have the Local synthesis characteristic of tag along sort, be input in linear SVM model be trained,
Holding vector machine is substantially a kind of two disaggregated models, i.e. finds the segmentation of optimum to surpass in two kind feature spaces
Plane, and in general application, the local feature of target scene segmentation is often more than two kinds, it is therefore desirable to use and prop up
" one to one " pattern holding vector machine carries out the process of many subseries, obtains the classification for different target types
Device, this grader judges the target type of the Local synthesis characteristic vector with tag along sort, i.e. can be determined that defeated
Which face that the region units such as the eyes entered or nose belong in Sample Storehouse；
Step 9, uses face to be identified step 1 to step 6, obtains the multiple with dividing of face to be identified
The Local synthesis characteristic vector of class label；
Step 10, the multiple Local synthesis characteristic vectors with tag along sort of face to be identified that will obtain,
It is input in described grader classify, obtains the target class belonging to face to be identified multiple regional area block
Type, if multiple regional area block is all categorized as same target type, recognition of face the most to be identified is this target
Type, face the most to be identified is not inconsistent with Sample Storehouse, is i.e. not belonging to Sample Storehouse.
With reference to Fig. 2, the system that the present invention uses by hyperspectral imager, round platform internal reflector, scanning turntable,
Synchronous bracket forms with fixture.Hyperspectral imager light spectrum image-forming scope, from 400nm to 1000nm, covers
Visible ray-near infrared band, owing to argent remains at a relatively high in visible ray-near infrared spectrum range
Reflectance (r > 0.96), beneficially imaging spectrometer are reflected by internally reflective the radial direction spectra collection of mirror, therefore select
Silver (Ag) is as the Coating Materials of round platform reflecting mirror inwall.Spatial discrimination due to hyperspectral imager scan line
Rate is 1632 pixels, and the average minimum resolution of spectrum is 0.64nm, i.e. has 936 bands, the highest
The having good stability of spatially and spectrally resolution requirement scanning process support, it is contemplated that the present invention will be by imager
Combine composition " auroral spectrum " acquisition system with internal reflector, need transformation to scan turntable, pass through
Processing synchronous bracket and fixture control the position of internal reflector, and the control mode transforming motor makes Gao Guang
The scanning process of spectrum imager and internal reflector keeps synchronizing.After hyperspectral imager is cascaded with internal reflector,
Need to utilize laser imaging optical path is calibrated and tests.And according to the image-forming principle of pole imaging, determine whole
The angle of visual field of " auroral spectrum " imaging system and EFFECTIVE RANGE.
Auroral spectrum system exists certain position relationship, in figure between round platform internal reflector and hyperspectral imager
R represents internal reflector left-hand end port radius, and θ represents imager lens centre and opens internal reflector left-hand end mouth
The half at angle, β represents internal reflector lens barrel incisal plane and horizontal plane angle, owing to face to be ensured is in internal reflection
Only carrying out Polaroid in mirror, so lens barrel horizontal length l meets some requirements, l must meet
L=2r cos (β) cos (θ-2 β) csc (θ-3 β).
With reference to Fig. 3, light, after truncated cones internal reflector, forms multi-view image in hyperspectral imager,
Length, subtended angle and a range of light of radius restriction front and back according to pre-designed reflecting mirror only pass through
Primary event can arrive imaging focal plane, thus completes the imaging of " an object point multiple views ", in operating distance
Interior space is divided into three mesh districts, binocular district and other blind areas.Within the position of face is in three mesh districts, should
Three mesh districts refer to any point meeting three viewpoint imagings diametrically simultaneously on face, in these three viewpoint i.e. figure
A, B, C 3 point, some B is the photocentre of hyperspectral imager imaging len, and some A and some C is that a B exists respectively
The picture become about internal reflector in the radial direction.Figure midpoint P is to be in pole imaging system in target scene space
A bit in three mesh districts, after internal reflector, forms A, B, C on the focal plane of imaging spectrometer respectively
The picture of three viewpoints, B spot projection is in the central circular of pole image, A and C spot projection is in pole image
In peripheral circle ring area.
With reference to Fig. 4, face is divided into multiple regional area blocks such as eyebrow, eyes, nose, lip, each
Regional area block all includes three dimensional structure (two dimensional image combines depth information) and spectral information (spectrum dimension
Data), each some reflectance to each spectral band during wherein spectral information comprises complete face.
Claims (3)
1. solid-EO-1 hyperion face identification method based on pole imaging, it is characterised in that include walking as follows
Rapid:
(1) utilize auroral spectrum system acquisition human face data, obtain the face two-dimentional pole image at each spectral band
And the spectral information that on this two dimension pole image, pixel is corresponding；
(2) from the two-dimentional pole image of the face obtained, the most much higher spectral band of quality is filtered out corresponding
Pole image, merges the multiple poles image filtered out, and the pole image after merging is carried out affine transformation,
Obtain the image of three different points of view of face；
(3) utilize multi-viewpoint three-dimensional matching process, the image obtaining three different points of view of face is carried out solid
Reconstruct, obtains the three dimensional structure of face；
(4) the full two-dimensional face in the pole image after described fusion is divided, obtain face with not
Two-dimentional regional area block with tag along sort；
(5) face that basis obtains is with the two-dimentional regional area block of different tag along sorts, to described face
The spectral information that three dimensional structure is corresponding with pixel on the image of described two dimension pole carries out regional area block division, obtains
Comprise the multiple regional area blocks with tag along sort of three dimensional structure and spectral information；
(6) to each regional area block in the multiple regional area blocks obtained, the feature of three dimensional structure is calculated
Vector, extracts the characteristic vector of spectral information simultaneously, and merges the two characteristic vector, obtain face
Multiple Local synthesis characteristic vectors with tag along sort；
(7) use step (1) to step (6) multiple sample faces, obtain comprising multiple sample face
The face Sample Storehouse of multiple Local synthesis characteristic vectors with tag along sort；
(8) it is input in linear SVM model be trained by the face Sample Storehouse data obtained,
To the grader for different target type；
(9) face to be identified is used step (1) to step (6), obtain face to be identified multiple with
The Local synthesis characteristic vector of tag along sort；
(10) the multiple Local synthesis characteristic vectors with tag along sort of face to be identified that will obtain, input
Classify in described grader, obtain the target type belonging to face to be identified multiple regional area block,
And whether belong to same target type according to multiple regional area blocks of face to be identified, it is judged that this face whether with
Sample Storehouse is consistent.
Solid-EO-1 hyperion face identification method based on pole imaging the most according to claim 1, its feature
It is: described in step (2), filter out the pole image that the most much higher spectral band of quality is corresponding, according to such as
Lower step realizes:
(2a) Y-PSNR function and evaluation function based on structural similarity, spectrum ripple each to face are utilized
The pole image of section carries out quality evaluation, obtains the Y-PSNR of face each spectral band pole image and based on structure
The scoring of similarity；
(2b) Y-PSNR of spectral band pole each to the face obtained image and based on structural similarity comment
Divide and carry out quantifying and linear superposition, obtain the quality score of each spectral band pole image, and therefrom filter out quality
The pole image that multiple spectral bands of highest scoring are corresponding.
Solid-EO-1 hyperion face identification method based on pole imaging the most according to claim 1, its feature
It is: the division described in step (4), realizes in accordance with the following steps:
(4a) after the face fusion that will obtain, the circular image comprising complete face in the image of pole splits,
Obtain complete two-dimension human face image；
(4b) the complete two-dimension human face image obtained is sharpened process, and detects two dimension after Edge contrast
The marginal point of facial image, obtains two-dimension human face image marginal point；
(4c) the two-dimension human face image marginal point obtained is clustered, obtain multiple bags in two-dimension human face image
Classification containing marginal point；
(4d) center position of multiple classifications comprising marginal point in the most calculated two-dimension human face image,
And determine its tag along sort and area size according to the position of central point of all categories, obtain face and classify with difference
The two-dimentional regional area block of label.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610247977.2A CN105956515B (en) | 2016-04-20 | 2016-04-20 | Solid-EO-1 hyperion face identification method based on pole imaging |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610247977.2A CN105956515B (en) | 2016-04-20 | 2016-04-20 | Solid-EO-1 hyperion face identification method based on pole imaging |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105956515A true CN105956515A (en) | 2016-09-21 |
| CN105956515B CN105956515B (en) | 2019-04-23 |
Family
ID=56917870
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610247977.2A Active CN105956515B (en) | 2016-04-20 | 2016-04-20 | Solid-EO-1 hyperion face identification method based on pole imaging |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105956515B (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106778681A (en) * | 2017-01-11 | 2017-05-31 | 厦门中控生物识别信息技术有限公司 | A kind of face identification method and face recognition device |
| CN106909944A (en) * | 2017-03-01 | 2017-06-30 | 西北大学 | A kind of method of face picture cluster |
| CN106991378A (en) * | 2017-03-09 | 2017-07-28 | 广东欧珀移动通信有限公司 | Facial orientation detection method, detection means and electronic installation based on depth |
| CN107194369A (en) * | 2017-06-09 | 2017-09-22 | 西安电子科技大学 | A kind of face identification method and face recognition device based on pole imaging volume camera |
| CN107239833A (en) * | 2017-05-12 | 2017-10-10 | 吉林师范大学 | A kind of information processing system and method based on deep learning |
| CN107748859A (en) * | 2017-08-10 | 2018-03-02 | 东南大学 | A kind of three-dimensional face identification method under partial occlusion based on RADIAL |
| CN107832793A (en) * | 2017-11-08 | 2018-03-23 | 深圳大学 | The sorting technique and system of a kind of high spectrum image |
| CN107991283A (en) * | 2017-12-26 | 2018-05-04 | 同方威视技术股份有限公司 | Raman spectrum detection device and Raman spectra detection process |
| CN108764328A (en) * | 2018-05-24 | 2018-11-06 | 广东工业大学 | The recognition methods of Terahertz image dangerous material, device, equipment and readable storage medium storing program for executing |
| CN108805862A (en) * | 2018-05-02 | 2018-11-13 | 南京大学 | A kind of tag discrimination methods based on improved structure similarity |
| CN110022462A (en) * | 2019-03-29 | 2019-07-16 | 江西理工大学 | A kind of safety defense monitoring system based on multispectral camera |
| CN110309692A (en) * | 2018-03-27 | 2019-10-08 | 杭州海康威视数字技术股份有限公司 | Face identification method, apparatus and system, model training method and device |
| CN111368681A (en) * | 2020-02-27 | 2020-07-03 | 深圳数联天下智能科技有限公司 | Live body discrimination method, device and equipment based on multipoint positioning and storage medium |
| CN111368644A (en) * | 2020-02-14 | 2020-07-03 | 深圳市商汤科技有限公司 | Image processing method, image processing device, electronic equipment and storage medium |
| WO2023000864A1 (en) * | 2021-07-19 | 2023-01-26 | 清华大学 | Face recognition method and system |
| CN116168346A (en) * | 2023-04-21 | 2023-05-26 | 独数一智(北京)信息科技有限公司 | Remote accompanying-reading monitoring system based on student behavior capturing |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7469060B2 (en) * | 2004-11-12 | 2008-12-23 | Honeywell International Inc. | Infrared face detection and recognition system |
| CN101398886A (en) * | 2008-03-17 | 2009-04-01 | 杭州大清智能技术开发有限公司 | Rapid three-dimensional face identification method based on bi-eye passiveness stereo vision |
| CN102324022A (en) * | 2011-09-05 | 2012-01-18 | 辽宁工程技术大学 | Composite gradient vector-based face recognition method |
| US20160086018A1 (en) * | 2013-04-26 | 2016-03-24 | West Virginia High Technology Consortium Foundation, Inc. | Facial recognition method and apparatus |
-
2016
- 2016-04-20 CN CN201610247977.2A patent/CN105956515B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7469060B2 (en) * | 2004-11-12 | 2008-12-23 | Honeywell International Inc. | Infrared face detection and recognition system |
| CN101398886A (en) * | 2008-03-17 | 2009-04-01 | 杭州大清智能技术开发有限公司 | Rapid three-dimensional face identification method based on bi-eye passiveness stereo vision |
| CN102324022A (en) * | 2011-09-05 | 2012-01-18 | 辽宁工程技术大学 | Composite gradient vector-based face recognition method |
| US20160086018A1 (en) * | 2013-04-26 | 2016-03-24 | West Virginia High Technology Consortium Foundation, Inc. | Facial recognition method and apparatus |
Cited By (27)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106778681B (en) * | 2017-01-11 | 2019-12-17 | 厦门中控智慧信息技术有限公司 | face recognition method and face recognition equipment |
| CN106778681A (en) * | 2017-01-11 | 2017-05-31 | 厦门中控生物识别信息技术有限公司 | A kind of face identification method and face recognition device |
| CN106909944A (en) * | 2017-03-01 | 2017-06-30 | 西北大学 | A kind of method of face picture cluster |
| CN106909944B (en) * | 2017-03-01 | 2020-05-26 | 西北大学 | Face picture clustering method |
| CN106991378A (en) * | 2017-03-09 | 2017-07-28 | 广东欧珀移动通信有限公司 | Facial orientation detection method, detection means and electronic installation based on depth |
| CN107239833A (en) * | 2017-05-12 | 2017-10-10 | 吉林师范大学 | A kind of information processing system and method based on deep learning |
| CN107239833B (en) * | 2017-05-12 | 2018-04-10 | 吉林师范大学 | A kind of information processing system and method based on deep learning |
| CN107194369A (en) * | 2017-06-09 | 2017-09-22 | 西安电子科技大学 | A kind of face identification method and face recognition device based on pole imaging volume camera |
| CN107748859A (en) * | 2017-08-10 | 2018-03-02 | 东南大学 | A kind of three-dimensional face identification method under partial occlusion based on RADIAL |
| CN107748859B (en) * | 2017-08-10 | 2021-04-27 | 东南大学 | Three-dimensional face recognition method under local shielding based on radial lines |
| CN107832793B (en) * | 2017-11-08 | 2021-08-06 | 深圳大学 | Hyperspectral image classification method and system |
| CN107832793A (en) * | 2017-11-08 | 2018-03-23 | 深圳大学 | The sorting technique and system of a kind of high spectrum image |
| CN107991283A (en) * | 2017-12-26 | 2018-05-04 | 同方威视技术股份有限公司 | Raman spectrum detection device and Raman spectra detection process |
| CN107991283B (en) * | 2017-12-26 | 2023-09-22 | 同方威视技术股份有限公司 | Raman spectrum detection device and Raman spectrum detection method |
| CN110309692B (en) * | 2018-03-27 | 2023-06-02 | 杭州海康威视数字技术股份有限公司 | Face recognition method, device and system, and model training method and device |
| CN110309692A (en) * | 2018-03-27 | 2019-10-08 | 杭州海康威视数字技术股份有限公司 | Face identification method, apparatus and system, model training method and device |
| CN108805862A (en) * | 2018-05-02 | 2018-11-13 | 南京大学 | A kind of tag discrimination methods based on improved structure similarity |
| CN108805862B (en) * | 2018-05-02 | 2021-08-13 | 南京大学 | Label identification method based on improved structure similarity |
| CN108764328A (en) * | 2018-05-24 | 2018-11-06 | 广东工业大学 | The recognition methods of Terahertz image dangerous material, device, equipment and readable storage medium storing program for executing |
| CN110022462A (en) * | 2019-03-29 | 2019-07-16 | 江西理工大学 | A kind of safety defense monitoring system based on multispectral camera |
| CN111368644A (en) * | 2020-02-14 | 2020-07-03 | 深圳市商汤科技有限公司 | Image processing method, image processing device, electronic equipment and storage medium |
| CN111368644B (en) * | 2020-02-14 | 2024-01-05 | 深圳市商汤科技有限公司 | Image processing method, device, electronic equipment and storage medium |
| CN111368681A (en) * | 2020-02-27 | 2020-07-03 | 深圳数联天下智能科技有限公司 | Live body discrimination method, device and equipment based on multipoint positioning and storage medium |
| CN111368681B (en) * | 2020-02-27 | 2023-09-01 | 深圳数联天下智能科技有限公司 | Living body screening method, device, equipment and storage medium based on multi-point positioning |
| WO2023000864A1 (en) * | 2021-07-19 | 2023-01-26 | 清华大学 | Face recognition method and system |
| CN116168346B (en) * | 2023-04-21 | 2023-08-18 | 独数一智(北京)信息科技有限公司 | Remote accompanying-reading monitoring system based on student behavior capturing |
| CN116168346A (en) * | 2023-04-21 | 2023-05-26 | 独数一智(北京)信息科技有限公司 | Remote accompanying-reading monitoring system based on student behavior capturing |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105956515B (en) | 2019-04-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105956515A (en) | Stereo-hyperspectral human face recognition method based on auroral imaging | |
| US10204262B2 (en) | Infrared imaging recognition enhanced by 3D verification | |
| WO2022041627A1 (en) | Living body facial detection method and system | |
| CN101999900B (en) | Living body detecting method and system applied to human face recognition | |
| JP4187651B2 (en) | Near-infrared method and system for use in face detection | |
| US7174033B2 (en) | Methods and systems for detecting and recognizing an object based on 3D image data | |
| Raghavendra et al. | Exploring the usefulness of light field cameras for biometrics: An empirical study on face and iris recognition | |
| CN111046703A (en) | Face anti-counterfeiting detection method and device and multi-view camera | |
| CN104881632A (en) | Hyperspectral face recognition method | |
| CN112288663A (en) | Infrared and visible light image fusion method and system | |
| Medioni et al. | Identifying noncooperative subjects at a distance using face images and inferred three-dimensional face models | |
| CN103116749A (en) | Near-infrared face identification method based on self-built image library | |
| CN105426843B (en) | Single-lens lower palm vein and palm print image acquisition device and image enhancement and segmentation method | |
| WO2016175234A1 (en) | Color image processing method, color image processing program, and object recognition method and device | |
| CN106372629A (en) | Living body detection method and device | |
| CN108416291A (en) | Face datection recognition methods, device and system | |
| CN109255319A (en) | For the recognition of face payment information method for anti-counterfeit of still photo | |
| Sun et al. | Context based face spoofing detection using active near-infrared images | |
| RU2370817C2 (en) | System and method for object tracking | |
| CN111753652B (en) | Three-dimensional face recognition method based on data enhancement | |
| Yao et al. | rPPG-based spoofing detection for face mask attack using EfficientNet on weighted spatial-temporal representation | |
| Proença | Non-cooperative iris recognition: Issues and trends | |
| Bastias et al. | A method for 3D iris reconstruction from multiple 2D near-infrared images | |
| Huynh et al. | Hyperspectral imaging for skin recognition and biometrics | |
| GB2471192A (en) | Iris and Ocular Recognition using Trace Transforms |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |