CN106548128A - Based on the regioselective facial image feature extracting method of multi-layer and device - Google Patents
Based on the regioselective facial image feature extracting method of multi-layer and device Download PDFInfo
- Publication number
- CN106548128A CN106548128A CN201610849893.6A CN201610849893A CN106548128A CN 106548128 A CN106548128 A CN 106548128A CN 201610849893 A CN201610849893 A CN 201610849893A CN 106548128 A CN106548128 A CN 106548128A
- Authority
- CN
- China
- Prior art keywords
- class
- regional
- regional area
- ensemble
- carried out
- 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
- G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
-
- 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
-
- 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
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/467—Encoded features or binary features, e.g. local binary patterns [LBP]
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Human Computer Interaction (AREA)
- Collating Specific Patterns (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides one kind and is based on the regioselective facial image feature extracting method of multi-layer and device, including：Multiple scanning windows are carried out into traverse scanning to pending image, regional ensemble is obtained；Regional area subset is obtained to the regional area sampling in regional ensemble；Localized region subset carries out parallel processing, obtains optimal candidate regional ensemble；Screening Treatment is carried out to optimal candidate regional ensemble, best region set is obtained；Face regional area is obtained according to best region set, and extracts corresponding face local feature.The present invention provides one kind and is based on the regioselective facial image feature extracting method of multi-layer and device, by the regional area for choosing the various yardsticks of image diverse location, the calculating between class distance vector in class is carried out successively, best region is selected to calculate optimal solution, face local feature is extracted to best region, so as to complete the sparse expression to face characteristic, the rate of precision of efficiency that face regional area selects and recognition of face is improve.
Description
Technical field
The present invention relates to technical field of image processing, more particularly to it is a kind of special based on the regioselective facial image of multi-layer
Levy extracting method and device.
Background technology
In image procossing and computer vision field, in the existing feature extracting method towards recognition of face, principle phase
To simple, the feature extracting method based on regional area such as the low local binary patterns of computation complexity (LBP), widely should
For field of face identification.And for such feature extracting method based on region, optimum is obtained from facial image
Regional area, is the important channel for realizing feature effective expression.Currently used regional selection method has following several：
Stress and strain model method, which carries out regularly stress and strain model to a width facial image, identical and non-overlapping copies in yardstick
Feature is extracted in local area image.The information in a more complete region is easily divided by this regional area division methods
To two or even in multiple independent regions, so as to cause local message to lack integrality.
In order to avoid causing the integrality of local message, it is proposed that the regional area selection method based on AdaBoost.But
There are two subject matters in the method when regional choice being carried out using AdaBoost learning algorithms：First, in the training stage and
Setting of the test phase to region importance is inconsistent.As AdaBoost Algorithm for Training is realized by iterative process, often
A regional area is selected after one wheel iteration, the region that the region for first choosing can be chosen than after is important；But
Test phase, the importance in all regions be defaulted as being identical, thus results in test result not fully up to expectations.Second, the party
The computational efficiency of method is relatively low.In each wheel iteration, the regional area of all candidates is compared and can just obtain one by needs
Best region, thus results in time complexity higher.
The content of the invention
The present invention provides one kind and is based on the regioselective facial image feature extracting method of multi-layer and device, for solving
The problem that precision is low and matching efficiency is low of features of human face images matching in prior art.
In a first aspect, the present invention provides one kind is based on the regioselective facial image feature extracting method of multi-layer, including：
Set up scanning window set of different sizes；
The scanning window in scanning window set is carried out into traverse scanning to pending image respectively, is obtained comprising M office
The regional ensemble in portion region；
Regional area in regional ensemble is sampled t time using formula sample mode is put back at random, t regional area is obtained
Collection；
Localized region subset carries out parallel processing, obtains optimal candidate regional ensemble；
Screening Treatment is carried out to optimal candidate regional ensemble, best region set is obtained；
Corresponding face regional area is obtained according to best region set, corresponding face office is extracted to face regional area
Portion's feature.
Preferably, the scanning window includes multiple length and widths within preset range, and the length of window two-by-two
There is the rectangular window of preset ratio with width.
Preferably, it is described that Screening Treatment is carried out to optimal candidate regional ensemble, best region set is obtained, including：
Regional area in optimal candidate regional ensemble is carried out in feature extraction, and zoning class and the card side between class
Distance vector；
Distance matrix is obtained according to the card side's distance vector in class and between class；
Linear model is generated according to the card side's distance vector in class and between class and distance matrix, linear model includes：
Wherein, Y is the interior card side's distance vector and between class of class；A is distance matrix；X+And X-Respectively in the class of region and between class
Card side's distance；β is assessment vector；
The optimal solution for obtaining assessment vector is calculated using Lasso regularized regressions algorithm according to linear block；
Screening Treatment is carried out according to the optimal solution of assessment vector to optimal candidate regional ensemble, best region set is obtained.
Preferably, feature extraction is carried out using LBP algorithms to the regional area in optimal candidate regional ensemble.
Second aspect, the present invention provide one kind and are based on the regioselective facial image feature deriving means of multi-layer, including：
Generation module, for setting up scanning window set of different sizes；
Scan module, for the scanning window in scanning window set is carried out traverse scanning to pending image respectively,
Obtain the regional ensemble comprising M regional area；
Sampling module, for being sampled t time using formula sample mode is put back at random to the regional area in regional ensemble, obtains
T regional area subset；
First processing module, carries out parallel processing for localized region subset, obtains optimal candidate regional ensemble；
Second processing module, for Screening Treatment is carried out to optimal candidate regional ensemble, obtains best region set；
Extraction module, for obtaining corresponding face regional area according to best region set, carries to face regional area
Take corresponding face local feature.
Preferably, the scanning window includes multiple length and widths within preset range, and the length of window two-by-two
There is the rectangular window of preset ratio with width.
Preferably, the Second processing module specifically for：
Regional area in optimal candidate regional ensemble is carried out in feature extraction, and zoning class and the card side between class
Distance vector；
Distance matrix is obtained according to the card side's distance vector in class and between class；
Linear model is generated according to the card side's distance vector in class and between class and distance matrix, linear model includes：
Wherein, Y is the interior card side's distance vector and between class of class；A is distance matrix；X+And X-Respectively in the class of region and between class
Card side's distance；β is assessment vector；
The optimal solution for obtaining assessment vector is calculated using Lasso regularized regressions algorithm according to linear block；
Screening Treatment is carried out according to the optimal solution of assessment vector to optimal candidate regional ensemble, best region set is obtained.
Preferably, feature extraction is carried out using LBP algorithms to the regional area in optimal candidate regional ensemble.
As shown from the above technical solution, the present invention provides a kind of based on the regioselective facial image feature extraction of multi-layer
Method and device, by the regional area for choosing the various yardsticks of diverse location in facial image, carries out regional area class successively
The calculating of interior and between class distance vector, the method choice for calculating optimal solution by Lasso regularizations goes out all to be had to position and yardstick
There is the best region of adaptivity, face local feature is extracted to best region, so as to complete the sparse expression to face characteristic,
Improve the rate of precision of efficiency that face regional area selects and recognition of face.
Description of the drawings
Fig. 1 is the stream based on the regioselective facial image feature extracting method of multi-layer that the embodiment of the present invention 1 is provided
Journey schematic diagram；
Fig. 2 is the acquisition schematic flow sheet of the best local zone that the embodiment of the present invention 1 is provided；
Fig. 3 is the acquisition image schematic flow sheet of the best local zone that the embodiment of the present invention 1 is provided；
Fig. 4 is the distribution schematic diagram of the best region that the embodiment of the present invention 1 is provided；
Fig. 5 is the knot based on the regioselective facial image feature deriving means of multi-layer that the embodiment of the present invention 2 is provided
Structure schematic diagram.
Specific embodiment
With reference to the accompanying drawings and examples, the specific embodiment of the present invention is described in further detail.Hereinafter implement
Example is for illustrating the present invention, but is not limited to the scope of the present invention.
It is a kind of based on the regioselective facial image feature extraction side of multi-layer that Fig. 1 shows that the embodiment of the present invention 1 is provided
Method, including：
S11, set up scanning window set of different sizes.
In this step, it should be noted that when setting up scanning window, can set up length and width preset range it
It is interior, to preset the rectangular window of the different size yardstick of step change.Therefore, scanning window set includes multiple length and width
Degree is within preset range, and the length and width of window has the rectangular window of preset ratio two-by-two.These rectangular windows
Collection is combined into S, { S1, S2, S3…Sn, wherein n is the number of window in set.
Length and width such as one of rectangular window is respectively 2 and 3, then the square for changing according to preset ratio and setting up
The length and width of shape window is respectively 3 and 5,4 and 7,5 and 9 ....
S12, the scanning window in scanning window set is carried out into traverse scanning to pending image respectively, obtained comprising M
The regional ensemble of individual regional area.
In this step, it should be noted that by each scanning window in scanning window set successively for whole
Pending image carry out traverse scanning, so as to obtain including the regional ensemble U of M regional area.
If rectangular window is S1, it is M which is scanned the regional area number for obtaining to pending image1。
If rectangular window is S2, it is M which is scanned the regional area number for obtaining to pending image2。
The like ...
If rectangular window is Sn, it is M which is scanned the regional area number for obtaining to pending imagen。
Then regional ensemble U may include M1+M2+…+MnIndividual regional area.That is M1+M2+…+Mn=M.
S13, in regional ensemble regional area using put back at random formula sample mode sample t time, obtain t partial zones
Domain subset.
In this step, it should be noted that all of regional area in regional ensemble U carries out first time sampling
Afterwards, obtain first regional area subset B1.The regional area that first time obtains is put back in regional ensemble U, the is being carried out
Secondary acquisition, obtains second regional area subset B2.The like, t collection is carried out, t regional area subset is obtained, i.e.,
For { B1, B2, B3…Bt}。
S14, localized region subset carry out parallel processing, obtain optimal candidate regional ensemble.
In this step, it should be noted that being directed to regional area subset { B1, B2, B3…BtParallel processing can be adopted
Sparse study preference pattern carries out parallel processing, and selection obtains optimal candidate regional ensemble C, as { C1, C2, C3…Ct}。
S15, Screening Treatment is carried out to optimal candidate regional ensemble, obtain best region set.
In this step, it should be noted that as shown in Figures 2 and 3, optimal candidate regional ensemble is carried out at screening
The step of reason, acquisition best region set, may include：
S151, feature extraction is carried out to the regional area in optimal candidate regional ensemble, and in the class of zoning and between class
Card side's distance vector.
In this step, it should be noted that being carried out using LBP algorithms to the regional area in optimal candidate regional ensemble
Feature extraction.In Digital Image Processing and area of pattern recognition, LBP refers to local binary patterns, is auxiliary Image Warping,
Carry out feature extraction.Therefore it is more ripe technology to carry out feature extraction using LBP algorithms to image local, be will not be described here.
Local features to getting carry out calculating the card side's distance vector in the class for obtaining regional area and between class.
S152, distance matrix is obtained according to the card side distance vector in class and between class.
In this step, it should be noted that calculating in the class for obtaining regional area and the card side's distance between class, respectively
X+And X-.Due between class distance X-Vectorial number q is much larger than inter- object distance X+Vectorial number p, is distance vector and class in balanced class
Between distance vector gap quantitatively, using Bootstrap methods in X-Collection up-sampling obtains P sample conductCollection, it is whole
Individual X+Collection conductAnd byWithComposition distance matrix A.
S153, linear model is generated according to the card side's distance vector in class and between class and distance matrix, linear model includes：
Wherein, Y is the interior card side's distance vector and between class of class；A is distance matrix；X+And X-Respectively in the class of region and between class
Card side's distance；β is assessment vector, for evaluating the importance of each element in distance vector.
S154, the optimal solution for calculating acquisition assessment vector using Lasso regularized regressions algorithm according to linear block.
S155, Screening Treatment is carried out according to the optimal solution of assessment vector to optimal candidate regional ensemble, obtain best region
Set.
In this step, it should be noted that seeking optimal solution by Lasso Regularization Problems：
Can obtain selecting evaluation coefficient β by solution formula.As openness effect selects the β in evaluation coefficient big
Majority is all zero, and for select coefficient for, negative value be it is nonsensical, so in β on the occasion of the area corresponding to element
Domain is the best region for choosing.The best region schematic diagram being illustrated in figure 4 in pending image.
S16, corresponding face regional area is obtained according to best region set, corresponding people is extracted to face regional area
Face local feature.
The embodiment of the present invention 1 provides a kind of based on the regioselective facial image feature extracting method of multi-layer, by choosing
The regional area of the various yardsticks of diverse location in facial image is taken, regional area class is carried out successively interior and between class distance vector
Calculate, the method choice for calculating optimal solution by Lasso regularizations goes out the optimal zone for all having adaptivity to position and yardstick
Domain, extracts face local feature to best region, so as to complete the sparse expression to face characteristic, improves face regional area
The efficiency selected and the rate of precision of recognition of face.
Fig. 5 shows that one kind that the embodiment of the present invention 2 is provided is based on the regioselective facial image feature extraction of multi-layer
Device, including generation module 21, scan module 22, sampling module 23, first processing module 24, Second processing module 25 and extraction
Module 26, wherein：
Generation module 21, for setting up scanning window set of different sizes；
Scan module 22, sweeps for the scanning window in scanning window set is carried out traversal to pending image respectively
Retouch, obtain the regional ensemble comprising M regional area；
Sampling module 23, for being sampled t time using formula sample mode is put back at random to the regional area in regional ensemble, obtains
To t regional area subset；
First processing module 24, carries out parallel processing for localized region subset, obtains optimal candidate regional ensemble；
Second processing module 25, for Screening Treatment is carried out to optimal candidate regional ensemble, obtains best region set；
Extraction module 26, for obtaining corresponding face regional area according to best region set, to face regional area
Extract corresponding face local feature.
In extraction process, generation module 21 sets up scanning window set of different sizes, and scanning window is sent to
Scan module 22.Scanning window in scanning window set is carried out traverse scanning to pending image by scan module 22 respectively,
The regional ensemble comprising M regional area is obtained, and set is sent to into sampling module 23.Sampling module 23 is in regional ensemble
Regional area using put back at random formula sample mode sample t time, obtain t regional area subset.First processing module 24 pairs
Regional area subset carries out parallel processing, obtains optimal candidate regional ensemble.Second processing module 25 is to optimal candidate set of regions
Conjunction carries out Screening Treatment, obtains best region set.Extraction module 26 obtains corresponding face local according to best region set
Region, extracts corresponding face local feature to face regional area.
A kind of concrete work based on the regioselective facial image feature deriving means of multi-layer in the embodiment of the present invention 4
Make process, may be referred to it is above-mentioned based on the content described by the regioselective facial image feature extracting method of multi-layer,
This no longer repeats one by one.
It should be noted that can be by hardware processor (hardware processor) come real in the embodiment of the present invention
Existing related function module.
The embodiment of the present invention 2 provides a kind of based on the regioselective facial image feature deriving means of multi-layer, by choosing
The regional area of the various yardsticks of diverse location in facial image is taken, regional area class is carried out successively interior and between class distance vector
Calculate, the method choice for calculating optimal solution by Lasso regularizations goes out the optimal zone for all having adaptivity to position and yardstick
Domain, extracts face local feature to best region, so as to complete the sparse expression to face characteristic, improves face regional area
The efficiency selected and the rate of precision of recognition of face.
Although additionally, it will be appreciated by those of skill in the art that some embodiments described herein include other embodiments
In some included features rather than further feature, but the combination of the feature of different embodiments means in of the invention
Within the scope of and form different embodiments.For example, in the following claims, embodiment required for protection appoint
One of meaning can in any combination mode using.
It should be noted that above-described embodiment the present invention will be described rather than limits the invention, and ability
Field technique personnel can design alternative embodiment without departing from the scope of the appended claims.In the claims,
Any reference symbol between bracket should not be configured to limitations on claims.Word "comprising" is not excluded the presence of not
Element listed in the claims or step.Word "a" or "an" before element does not exclude the presence of multiple such
Element.The present invention can come real by means of the hardware for including some different elements and by means of properly programmed computer
It is existing.If in the unit claim for listing equipment for drying, several in these devices can be by same hardware branch
To embody.The use of word first, second, and third does not indicate that any order.These words can be explained and be run after fame
Claim.
One of ordinary skill in the art will appreciate that：Various embodiments above only to illustrate technical scheme, and
It is non-which is limited；Although being described in detail to the present invention with reference to foregoing embodiments, one of ordinary skill in the art
It should be understood that：Which still can be modified to the technical scheme described in foregoing embodiments, or to which part or
All technical characteristic carries out equivalent；And these modifications or replacement, do not make the essence of appropriate technical solution depart from this
Bright claim limited range.
Claims (8)
1. it is a kind of to be based on the regioselective facial image feature extracting method of multi-layer, it is characterised in that to include：
Set up scanning window set of different sizes；
The scanning window in scanning window set is carried out into traverse scanning to pending image respectively, is obtained comprising M partial zones
The regional ensemble in domain；
Regional area in regional ensemble is sampled t time using formula sample mode is put back at random, t regional area subset is obtained；
Localized region subset carries out parallel processing, obtains optimal candidate regional ensemble；
Screening Treatment is carried out to optimal candidate regional ensemble, best region set is obtained；
Corresponding face regional area is obtained according to best region set, corresponding face local is extracted to face regional area special
Levy.
2. method according to claim 1, it is characterised in that the scanning window includes multiple length and widths default
Within the scope of, and the length and width of window has the rectangular window of preset ratio two-by-two.
3. method according to claim 1, it is characterised in that described that Screening Treatment is carried out to optimal candidate regional ensemble,
Best region set is obtained, including：
Regional area in optimal candidate regional ensemble is carried out in feature extraction, and zoning class and the card side's distance between class
Vector；
Distance matrix is obtained according to the card side's distance vector in class and between class；
Linear model is generated according to the card side's distance vector in class and between class and distance matrix, linear model includes：
Wherein, Y is the interior card side's distance vector and between class of class；A is distance matrix；X+And X-Card respectively in the class of region and between class
Square distance；β is assessment vector；
The optimal solution for obtaining assessment vector is calculated using Lasso regularized regressions algorithm according to linear block；
Screening Treatment is carried out according to the optimal solution of assessment vector to optimal candidate regional ensemble, best region set is obtained.
4. method according to claim 3, it is characterised in that the regional area in optimal candidate regional ensemble is adopted
LBP algorithms carry out feature extraction.
5. it is a kind of to be based on the regioselective facial image feature deriving means of multi-layer, it is characterised in that to include：
Generation module, for setting up scanning window set of different sizes；
Scan module, for the scanning window in scanning window set is carried out to pending image traverse scanning respectively, obtains
Regional ensemble comprising M regional area；
Sampling module, for sampling t time using formula sample mode is put back at random to the regional area in regional ensemble, obtains t
Regional area subset；
First processing module, carries out parallel processing for localized region subset, obtains optimal candidate regional ensemble；
Second processing module, for Screening Treatment is carried out to optimal candidate regional ensemble, obtains best region set；
Extraction module, for obtaining corresponding face regional area according to best region set, extracts right to face regional area
The face local feature answered.
6. device according to claim 5, it is characterised in that the scanning window includes multiple length and widths default
Within the scope of, and the length and width of window has the rectangular window of preset ratio two-by-two.
7. device according to claim 5, it is characterised in that the Second processing module specifically for：
Regional area in optimal candidate regional ensemble is carried out in feature extraction, and zoning class and the card side's distance between class
Vector；
Distance matrix is obtained according to the card side's distance vector in class and between class；
Linear model is generated according to the card side's distance vector in class and between class and distance matrix, linear model includes：
Wherein, Y is the interior card side's distance vector and between class of class；A is distance matrix；X+And X-Card respectively in the class of region and between class
Square distance；β is assessment vector；
The optimal solution for obtaining assessment vector is calculated using Lasso regularized regressions algorithm according to linear block；
Screening Treatment is carried out according to the optimal solution of assessment vector to optimal candidate regional ensemble, best region set is obtained.
8. device according to claim 7, it is characterised in that the regional area in optimal candidate regional ensemble is adopted
LBP algorithms carry out feature extraction.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610849893.6A CN106548128A (en) | 2016-09-26 | 2016-09-26 | Based on the regioselective facial image feature extracting method of multi-layer and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610849893.6A CN106548128A (en) | 2016-09-26 | 2016-09-26 | Based on the regioselective facial image feature extracting method of multi-layer and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106548128A true CN106548128A (en) | 2017-03-29 |
Family
ID=58368019
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610849893.6A Pending CN106548128A (en) | 2016-09-26 | 2016-09-26 | Based on the regioselective facial image feature extracting method of multi-layer and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106548128A (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101127076A (en) * | 2007-09-27 | 2008-02-20 | 上海交通大学 | Human eye state detection method based on cascade classification and hough circle transform |
| CN103049733A (en) * | 2011-10-11 | 2013-04-17 | 株式会社理光 | Human face detection method and human face detection equipment |
-
2016
- 2016-09-26 CN CN201610849893.6A patent/CN106548128A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101127076A (en) * | 2007-09-27 | 2008-02-20 | 上海交通大学 | Human eye state detection method based on cascade classification and hough circle transform |
| CN103049733A (en) * | 2011-10-11 | 2013-04-17 | 株式会社理光 | Human face detection method and human face detection equipment |
Non-Patent Citations (1)
| Title |
|---|
| YUE ZHAO ET AL: "New Sparse Facial Feature Description model Based on Salience Evaluation of Regions and Features", 《INTERNATIONAL JOURNAL OF PATTERN RECOGNITION AND ARTIFICIAL INTELLIGENCE》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Fiaschi et al. | Learning to count with regression forest and structured labels | |
| CN108830209B (en) | Remote sensing image road extraction method based on generation countermeasure network | |
| Wu et al. | Classification of defects with ensemble methods in the automated visual inspection of sewer pipes | |
| CN109670429A (en) | A kind of the monitor video multiple target method for detecting human face and system of Case-based Reasoning segmentation | |
| CN104392463A (en) | Image salient region detection method based on joint sparse multi-scale fusion | |
| Dobrescu et al. | Understanding deep neural networks for regression in leaf counting | |
| CN109508644A (en) | Facial paralysis grade assessment system based on the analysis of deep video data | |
| CN108537751B (en) | Thyroid ultrasound image automatic segmentation method based on radial basis function neural network | |
| CN104346617A (en) | Cell detection method based on sliding window and depth structure extraction features | |
| Ylipaavalniemi et al. | Analyzing consistency of independent components: An fMRI illustration | |
| Li et al. | Separating the EoR signal with a convolutional denoising autoencoder: a deep-learning-based method | |
| Sali et al. | Celiacnet: Celiac disease severity diagnosis on duodenal histopathological images using deep residual networks | |
| CN106096655A (en) | A kind of remote sensing image airplane detection method based on convolutional neural networks | |
| CN115546605A (en) | Training method and device based on image labeling and segmentation model | |
| Sadik et al. | SpecMEn-DL: Spectral mask enhancement with deep learning models to predict COVID-19 from lung ultrasound videos | |
| Wu et al. | A new sampling algorithm for high-quality image matting | |
| CN103903015A (en) | Cell mitosis detection method | |
| Reyes et al. | Vetting asteroseismic Δν measurements using neural networks | |
| Yu et al. | Salient feature extraction for hyperspectral image classification | |
| Somervuo | Time–frequency warping of spectrograms applied to bird sound analyses | |
| Aryal et al. | Context-aware graph-based self-supervised learning of whole slide images | |
| CN110930369B (en) | Pathological section identification method based on group et-variable neural network and conditional probability field | |
| KR20140053259A (en) | Techniques for feature extraction | |
| CN106548128A (en) | Based on the regioselective facial image feature extracting method of multi-layer and device | |
| CN105139019A (en) | Iris image screening method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170329 |
|
| RJ01 | Rejection of invention patent application after publication |