CN108520215B - Single-sample face recognition method based on multi-scale joint feature encoder - Google Patents
Single-sample face recognition method based on multi-scale joint feature encoder Download PDFInfo
- Publication number
- CN108520215B CN108520215B CN201810262820.6A CN201810262820A CN108520215B CN 108520215 B CN108520215 B CN 108520215B CN 201810262820 A CN201810262820 A CN 201810262820A CN 108520215 B CN108520215 B CN 108520215B
- Authority
- CN
- China
- Prior art keywords
- face
- feature
- image
- matched
- scale
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 19
- 239000011159 matrix material Substances 0.000 claims abstract description 28
- 238000012549 training Methods 0.000 claims abstract description 28
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 5
- 238000005070 sampling Methods 0.000 claims description 23
- 239000013598 vector Substances 0.000 claims description 19
- 238000012545 processing Methods 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 3
- 238000009499 grossing Methods 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 2
- 238000010606 normalization Methods 0.000 claims description 2
- 238000012216 screening Methods 0.000 claims 1
- 238000005286 illumination Methods 0.000 abstract description 6
- 238000005516 engineering process Methods 0.000 abstract description 4
- 238000012360 testing method Methods 0.000 description 8
- 238000012986 modification Methods 0.000 description 5
- 230000004048 modification Effects 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
Images
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
-
- 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
Abstract
The invention discloses a single-sample face recognition method based on a multi-scale joint feature encoder, which comprises the following steps: preprocessing an image; dividing a face area; independent feature learning of multi-scale regions, wherein each face region is trained respectively to obtain an independent feature coding matrix; performing joint feature learning, namely performing decomposition and clustering learning on the multi-scale independent feature coding matrix of each region to obtain a joint feature coding matrix; extracting features, namely extracting the features of the face image through a multi-scale combined feature coding matrix; and (4) feature identification, namely comparing the features extracted from each region to obtain an identification result. The invention provides a combined feature learning method based on a sparse automatic coding technology, which can learn robust face features under the condition of limited face training data, solve the face recognition problem under the conditions of shielding and illumination interference and greatly expand the application scene of face recognition.
Description
Technical Field
The invention relates to the technical field of face recognition, in particular to a single-sample face recognition method based on a multi-scale joint feature encoder.
Background
Because of the uniqueness and the non-replicability of the human face, the human face is an important biological feature in identity recognition. Face recognition technology is widely used in various security fields due to its non-contact and easy acquisition. However, in practical applications, various complex scenes and single sample set limitations exist, so that efficient and ubiquitous face recognition algorithms are increasingly hot in current research.
The face recognition technology in an actual recognition application scene is often influenced by factors such as shielding, illumination and posture, the traditional face recognition method is often simple and low in efficiency, and the face recognition in a complex scene cannot be dealt with. Most targets to be identified in an actual identification application scene only have unique or limited face sample images, and the popular deep learning method at present needs a large number of training images. New methods are therefore urgently needed to address such problems.
Disclosure of Invention
The invention aims to: in order to solve the defects of the prior art and provide a more accurate and more convenient face recognition method aiming at the conditions of complex scenes and single-sample face data, the invention provides a single-sample face recognition method based on a multi-scale joint feature encoder.
The technical scheme of the invention is as follows: a single sample face recognition method based on a multi-scale joint feature encoder comprises the following steps:
training a multi-scale joint feature encoder:
carrying out image preprocessing on the face image, including size normalization, gray level image conversion, image enhancement and the like;
dividing the preprocessed face image into a plurality of face areas with the same size and without overlap;
traversing and sampling templates with different scales for each face region to obtain a training data set of each face region under different scales; based on a preset image feature coding mode (any conventional mode can be adopted for the image feature coding mode), performing feature coding training on a training data set of each face region under each scale to obtain independent feature coding matrixes of different face regions under different scales, and completing multi-scale region independent feature learning;
decomposing each independent feature coding matrix into a plurality of feature coding vectors based on the same decomposition number, namely the feature coding vectors obtained by decomposing each independent feature coding matrix are the same in number; the feature coding vectors of the same face region under the same scale form a feature coding vector set of different face regions under different scales, the feature coding vector set is called as the feature of each face region under different scales for short, then the features of all the face regions under the same scale are fused into a combined feature coding matrix under the current scale to obtain combined feature coding matrices of different scales, and a multi-scale combined feature encoder is obtained based on a preset image feature coding mode and the combined feature coding matrices of different scales;
carrying out feature extraction on the face image to be recognized and the face image to be matched:
after image preprocessing is carried out on the face image to be recognized and the face image to be matched (the image preprocessing mode is the same as that of the multi-scale joint feature encoder during training), the face image to be recognized and the face image to be matched are divided into a plurality of face regions to be recognized and face regions to be matched, wherein the number of the face regions is the same as that of the face regions divided during training of the multi-scale joint feature encoder;
traversing and sampling templates with different scales for each face region to obtain a data set to be recognized and a data set to be matched of each face region under different scales;
extracting features of a data set to be recognized based on a multi-scale joint feature encoder to obtain a feature matrix of a face image to be recognized under different scales;
extracting the features of the data set to be matched based on a multi-scale combined feature encoder to obtain a feature matrix of the face image to be matched under different scales;
calculating the matching degree of the face image to be recognized and the face image to be matched:
based on the feature matrix under each scale, calculating the similarity of the image to be recognized and the image to be matched in the same face region based on the vector distance, wherein the optimal similarity calculation formula is as follows:
wherein s is t (i, j) represents the similarity of the face image i to be recognized and the face image j to be matched in the face area t,
representing the characteristic distance extracted from the face image i to be recognized and the face image j to be matched in the face area t, namely
Where d () represents a distance function which,
respectively representing the characteristics of a face image i to be recognized and a face image j to be matched in a face area t, M representing the number of the face images to be matched, k representing an identifier of the face image to be matched, and epsilon representing a smoothing parameter, wherein the value is usually a number greater than 0, and the preferred value is 0.1;
and calculating the overall similarity of the face image to be recognized and the face image to be matched according to the preset weights of different face regions, namely the matching degree of the face image to be recognized and the face image to be matched is the weighted sum of the similarity of all the face regions.
Preferably, the weight setting mode of each face region is as follows:
in order to further improve the recognition accuracy, after the face region of the face image to be recognized is screened, the overall similarity is calculated: based on a face region classifier trained to identify whether the face region is a normal face region, inputting each face region of a face image to be identified into the trained face region classifier, discarding an abnormal face region (such as a face region under the influence of shielding, illumination and expression), and only weighting and summing the similarity of the face regions on the normal face region to obtain the overall similarity.
In addition, a similarity threshold value can be set, the face image to be matched which is lower than the similarity threshold value is used as an abnormal matching result, and only the matching result which is not lower than the similarity threshold value is output.
In summary, due to the adoption of the technical solutions, one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:
the face region features under different scales are learned through a region independent feature coding technology to obtain face information under different scales, then all independent coding matrixes are decomposed, and clustering reconstruction is carried out to obtain global combined features, so that the single-sample face information representation capability is stronger. Through the calculation of the similarity of each region, each region is weighted, and the beneficial region suppression interference region is enhanced, so that the face recognition rate in a complex scene is improved.
Drawings
FIG. 1: the invention discloses a schematic diagram of a training principle of a multi-scale joint feature encoder.
FIG. 2: the testing principle in the embodiment of the invention is shown schematically.
FIG. 3: the human face region division schematic diagram in the embodiment of the invention.
FIG. 4: the sampling scale in the embodiment of the invention is shown schematically.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the embodiments and the accompanying drawings.
Examples
1. Referring to fig. 1, the specific implementation steps of training the multi-scale joint feature encoder in the present invention are as follows:
s1: the specific steps of training set face image preprocessing are as follows:
s11: converting the image from a color space to a gray scale space;
s12: and carrying out image enhancement operation.
S2: the method comprises the following specific steps of training set face image region division and dividing a face into non-overlapping regions:
s21: the face data of the training set are label-free data, and the face of the training set can be related or unrelated to the category of the face data of the test set;
s22: dividing the preprocessed face image into T non-overlapping regions of the same size, referring to fig. 3;
s23: traversing and sampling in each area by taking the size of m multiplied by m pixels as a template to obtain a training data set, wherein the data set of the t-th area is
Wherein t represents a sampling region, i represents the number of samples in the t region, x represents the sampled data, and the sampling template refers to fig. 4;
s24: traversing and sampling in each area by taking the size of m multiplied by 2m pixels as a template to obtain a training data set, wherein the data set of the t-th area is
Wherein t represents a sampling region, i represents the number of samples in the region t, x represents the sampled data, and the sampling template refers to fig. 4;
s25: traversing and sampling in each area by taking the size of 2m multiplied by m pixels as a template to obtain a training data set, wherein the data set of the t-th area is
Wherein t represents a sampling region, i represents the number of samples in the t region, x represents the sampled data, and the sampling template refers to fig. 4;
s3: independent region feature learning, namely training each face region based on a preset feature coding mode (such as sparse automatic coding) to obtain an independent feature coding matrix;
s31: data for each region at m × m scale
Respectively training to obtain independent feature coding matrix
S32: data set for each region at m × 2m scale
Respectively training to obtain independent feature coding matrix
S33: data set for each region at 2m x m scale
Respectively training to obtain independent feature coding matrix
S4: performing joint feature learning, namely performing decomposition and clustering learning on feature coding matrixes of which the regions are independent from each other to obtain a joint feature coding matrix;
s41: decomposing all independent feature coding matrixes into a plurality of feature coding vectors, and using W to simplify description t Representing the independent characteristic coding matrix to be decomposed currently, and the decomposition steps are as follows:
wherein, W t An independent feature encoding matrix representing the tth individual face region at the current scale,
a set of feature code vectors representing the tth personal face region, i representing a decomposed feature code vector identifier, k representing the number of decomposed feature code vectors, and ω representing a feature code vector;
s42: and (3) fusing all feature coding vectors into a joint feature coding matrix through a clustering algorithm, such as:
wherein the content of the first and second substances,
features indicating the tth individual face regionSyndrome vector set, W u Representing the joint feature coding matrix at the current scale, in this embodiment, using
Representing the joint feature coding matrix at the m × m,2m × m and m × 2m scales, respectively.
By
And obtaining the multi-scale joint feature encoder by a preset feature encoding mode.
2. Referring to fig. 2, the single-sample face recognition processing based on the multi-scale joint feature encoder is implemented by the following steps:
s1: the sample set and the specific steps of the pretreatment of the tested face image are as follows:
s11: converting the image from a color space to a gray scale space;
s12: and carrying out image enhancement operation.
S2: the method comprises the following specific steps of sample set and test face image region division, wherein the specific steps of dividing a face into non-overlapping regions are as follows:
s21: dividing the preprocessed face image into T non-overlapping regions of the same size, referring to fig. 3;
s22: traversing and sampling in each area by taking the size of m multiplied by m pixels as a template to obtain a training data set, wherein the data set of the t-th area is
Wherein t represents a sampling region, i represents the number of samples in the region t, x represents the sampled data, and the sampling template refers to fig. 4;
s23: traversing and sampling in each area by taking the size of m multiplied by 2m pixels as a template to obtain a training data set, wherein the data set of the t-th area is
Wherein t represents a sampling region, i represents the number of samples in the region t, x represents the sampled data, and the sampling template refers to fig. 4;
s24: traversing and sampling in each area by taking the size of 2m multiplied by m pixels as a template to obtain a training data set, wherein the data set of the t-th area is
Wherein t represents a sampling region, i represents the number of samples in the t region, x represents the sampled data, and the sampling template refers to fig. 4;
s3: and extracting the face characteristics of the sample set by using the multi-scale joint characteristic coding matrix, and extracting the characteristics of the sampled data of each region under the corresponding scale by using the trained joint characteristic coding matrix under each scale.
S4: calculating the characteristic distance of corresponding area between the tested face and all sample faces
S41: calculating the characteristic distance extracted from the corresponding face area of the sample face i and the test face j
Wherein
Representing the characteristics of the face i in the t area;
s5: calculating the similarity of corresponding areas between the test face and all sample faces, wherein the similarity calculation formula is as follows:
wherein s is t (i, j) represents the similarity of the faces i and j in the t area,
the feature distance extracted in the t region is represented, M represents the total number of classes included in the sample when the sample number to be compared is identified, for example, in the case of image class identification, and ∈ represents a smoothing parameter, and is usually 0.1.
S6: judging each area as a normal face area through the classifier, referring to fig. 2, discarding an abnormal face area, such as a face area under the influence of shielding, illumination and expression, and weighting and calculating the whole face similarity of the remaining areas, wherein a calculation formula of the similarity is as follows:
wherein, w t Is the weight occupied by the t region similarity in the recognition, s t (i, j) represents the similarity of the faces i and j in the t area, and N represents the number of normal face areas.
In this embodiment, the image category is subjected to recognition processing, in order to further improve recognition accuracy, a similarity threshold is trained, if the similarity threshold is lower than the similarity threshold, it is determined that the test image classification does not exist in the sample set, otherwise, the sample class with the largest similarity to the test face is the recognition result.
The face recognition algorithm tests a local shielding face sequence, an illumination change face sequence and an expression change face sequence contained in an AR face library on the premise of a single sample, and the average recognition rate reaches 95.85 percent, while the average recognition rate of the current similar method is 94.38 percent; on the premise of a single sample, the illumination change face sequence contained in the Yale _ B face library is tested, and the average recognition rate reaches 93.66%, while the average recognition rate of the current similar method is 82.22%. The face recognition method improves the accuracy of single-sample face recognition in a complex scene, and has good popularization and application prospects.
While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all such alterations and modifications as fall within the scope of the invention.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.
Claims (3)
1. A single sample face recognition method based on a multi-scale joint feature encoder is characterized by comprising the following steps:
training a multiscale joint feature encoder:
carrying out image preprocessing on the face image;
dividing the preprocessed face image into a plurality of face areas which have the same size and are not overlapped;
traversing and sampling templates with different scales for each face region to obtain a training data set of each face region under different scales; performing feature coding training on a training data set of each face region under each scale based on a preset image feature coding mode to obtain independent feature coding matrixes of different face regions under different scales; the different scale template sizes include: m × m, m × 2m,2m × m, where m represents the width of the template;
decomposing each independent feature coding matrix into a plurality of feature coding vectors based on the same decomposition number, namely the feature coding vectors obtained by decomposing each independent feature coding matrix are the same in number; feature coding vectors of the same face region under the same scale form feature coding vector sets of different face regions under different scales, and then the feature coding vector sets of all the face regions under the same scale are fused into a combined feature coding matrix under the current scale to obtain combined feature coding matrices under different scales;
obtaining a multi-scale joint feature encoder based on a preset image feature encoding mode and joint feature encoding matrixes with different scales;
carrying out feature extraction on the face image to be recognized and the face image to be matched:
after image preprocessing is carried out on a face image to be recognized and a face image to be matched, the face image to be recognized and the face image to be matched are divided into a plurality of face regions to be recognized and face regions to be matched, wherein the number of the face regions is the same as the number of the face regions divided when a multi-scale combined feature encoder is trained;
traversing and sampling templates with different scales for each face region to obtain a data set to be recognized and a data set to be matched of each face region under different scales;
performing feature extraction on a data set to be recognized based on a multi-scale joint feature encoder to obtain a feature matrix of a face image to be recognized under different scales;
extracting the features of the data set to be matched based on a multi-scale combined feature encoder to obtain a feature matrix of the face image to be matched under different scales;
calculating the matching degree of the face image to be recognized and the face image to be matched:
based on the feature matrix under each scale, calculating the similarity of the image to be recognized and the image to be matched in the same face region based on the vector distance:
wherein s is t (i, j) represents the similarity of the face image i to be recognized and the face image j to be matched in the face area t,
representing the characteristic distance extracted from the face image i to be recognized and the face image j to be matched in the face area t, namely
Where d () represents a distance function,
f t j respectively representing the characteristics of a face image i to be recognized and a face image j to be matched in a face area t, wherein M represents the number of the face images to be matched, k represents an identifier of the face image to be matched, and epsilon represents a smoothing parameter;
according to the formula
Calculating the weight w of the face region t t ；
The similarity of each face area is weighted to obtain the matching degree of the face image to be recognized and the face image to be matched, and the face image to be matched with the maximum matching degree is used as a recognition result.
2. The method according to claim 1, further comprising performing normal face region screening processing on the face image to be recognized, and performing face region similarity weighting summation only on the normal face region to obtain the matching degree between the face image to be recognized and the face image to be matched.
3. The method of claim 1, wherein image pre-processing comprises: size normalization, grey scale map conversion and image enhancement processing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810262820.6A CN108520215B (en) | 2018-03-28 | 2018-03-28 | Single-sample face recognition method based on multi-scale joint feature encoder |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810262820.6A CN108520215B (en) | 2018-03-28 | 2018-03-28 | Single-sample face recognition method based on multi-scale joint feature encoder |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108520215A CN108520215A (en) | 2018-09-11 |
| CN108520215B true CN108520215B (en) | 2022-10-11 |
Family
ID=63434356
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810262820.6A Expired - Fee Related CN108520215B (en) | 2018-03-28 | 2018-03-28 | Single-sample face recognition method based on multi-scale joint feature encoder |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108520215B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109766857A (en) * | 2019-01-16 | 2019-05-17 | 嘉兴学院 | A kind of three-dimensional face identification method based on semantic alignment multizone template fusion |
| CN109951289B (en) * | 2019-01-25 | 2021-01-12 | 北京三快在线科技有限公司 | Identification method, device, equipment and readable storage medium |
| CN109919888B (en) * | 2019-02-26 | 2023-09-19 | 腾讯科技（深圳）有限公司 | Image fusion method, model training method and related device |
| CN110210321B (en) * | 2019-05-08 | 2020-11-24 | 华南理工大学 | Under-sample face recognition method based on multi-dimensional scale transformation network and block weighting method |
| CN111274883B (en) * | 2020-01-10 | 2023-04-25 | 杭州电子科技大学 | Synthetic sketch face recognition method based on multi-scale HOG features and deep features |
| CN115601484B (en) * | 2022-11-07 | 2023-03-28 | 广州趣丸网络科技有限公司 | Virtual character face driving method and device, terminal equipment and readable storage medium |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102521561A (en) * | 2011-11-16 | 2012-06-27 | 湖南大学 | Face identification method on basis of multi-scale weber local features and hierarchical decision fusion |
| CN103310236A (en) * | 2013-06-27 | 2013-09-18 | 上海数据分析与处理技术研究所 | Mosaic image detection method and system based on local two-dimensional characteristics |
| CN103632132A (en) * | 2012-12-11 | 2014-03-12 | 广西工学院 | Face detection and recognition method based on skin color segmentation and template matching |
| CN103745207A (en) * | 2014-01-27 | 2014-04-23 | 中国科学院深圳先进技术研究院 | Feature extraction method and device for human face identification |
| CN104680158A (en) * | 2015-03-31 | 2015-06-03 | 盐城工学院 | Face recognition method based on multi-scale block partial multi-valued mode |
| CN104992223A (en) * | 2015-06-12 | 2015-10-21 | 安徽大学 | Dense population estimation method based on deep learning |
| CN105975931A (en) * | 2016-05-04 | 2016-09-28 | 浙江大学 | Convolutional neural network face recognition method based on multi-scale pooling |
| CN106845362A (en) * | 2016-12-27 | 2017-06-13 | 湖南长城信息金融设备有限责任公司 | A kind of face identification method of the rarefaction representation based on multi-scale transform |
| CN106897700A (en) * | 2017-02-27 | 2017-06-27 | 苏州大学 | A kind of single sample face recognition method and system |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102592136B (en) * | 2011-12-21 | 2013-10-16 | 东南大学 | Three-dimensional human face recognition method based on intermediate frequency information in geometry image |
| CN103514432B (en) * | 2012-06-25 | 2017-09-01 | 诺基亚技术有限公司 | Face feature extraction method, equipment and computer program product |
| WO2015127394A1 (en) * | 2014-02-23 | 2015-08-27 | Northeastern University | System for beauty, cosmetic, and fashion analysis |
| US9858502B2 (en) * | 2014-03-31 | 2018-01-02 | Los Alamos National Security, Llc | Classification of multispectral or hyperspectral satellite imagery using clustering of sparse approximations on sparse representations in learned dictionaries obtained using efficient convolutional sparse coding |
| CN103902977B (en) * | 2014-03-31 | 2017-04-05 | 华为技术有限公司 | Face identification method and device based on Gabor binary patterns |
| EP3029606A3 (en) * | 2014-11-14 | 2016-09-14 | Thomson Licensing | Method and apparatus for image classification with joint feature adaptation and classifier learning |
| CN105447441B (en) * | 2015-03-19 | 2019-03-29 | 北京眼神智能科技有限公司 | Face authentication method and device |
| CN107273824B (en) * | 2017-05-27 | 2020-04-07 | 西安电子科技大学 | Face recognition method based on multi-scale and multi-direction local binary pattern |
-
2018
- 2018-03-28 CN CN201810262820.6A patent/CN108520215B/en not_active Expired - Fee Related
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102521561A (en) * | 2011-11-16 | 2012-06-27 | 湖南大学 | Face identification method on basis of multi-scale weber local features and hierarchical decision fusion |
| CN103632132A (en) * | 2012-12-11 | 2014-03-12 | 广西工学院 | Face detection and recognition method based on skin color segmentation and template matching |
| CN103310236A (en) * | 2013-06-27 | 2013-09-18 | 上海数据分析与处理技术研究所 | Mosaic image detection method and system based on local two-dimensional characteristics |
| CN103745207A (en) * | 2014-01-27 | 2014-04-23 | 中国科学院深圳先进技术研究院 | Feature extraction method and device for human face identification |
| CN104680158A (en) * | 2015-03-31 | 2015-06-03 | 盐城工学院 | Face recognition method based on multi-scale block partial multi-valued mode |
| CN104992223A (en) * | 2015-06-12 | 2015-10-21 | 安徽大学 | Dense population estimation method based on deep learning |
| CN105975931A (en) * | 2016-05-04 | 2016-09-28 | 浙江大学 | Convolutional neural network face recognition method based on multi-scale pooling |
| CN106845362A (en) * | 2016-12-27 | 2017-06-13 | 湖南长城信息金融设备有限责任公司 | A kind of face identification method of the rarefaction representation based on multi-scale transform |
| CN106897700A (en) * | 2017-02-27 | 2017-06-27 | 苏州大学 | A kind of single sample face recognition method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108520215A (en) | 2018-09-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108520215B (en) | Single-sample face recognition method based on multi-scale joint feature encoder | |
| CN110443143B (en) | Multi-branch convolutional neural network fused remote sensing image scene classification method | |
| Gao et al. | Automatic change detection in synthetic aperture radar images based on PCANet | |
| CN110348319B (en) | Face anti-counterfeiting method based on face depth information and edge image fusion | |
| CN110633708A (en) | Deep network significance detection method based on global model and local optimization | |
| CN109801305B (en) | SAR image change detection method based on deep capsule network | |
| US20200134382A1 (en) | Neural network training utilizing specialized loss functions | |
| CN110827265B (en) | Image anomaly detection method based on deep learning | |
| CN113011357A (en) | Depth fake face video positioning method based on space-time fusion | |
| CN116910752B (en) | Malicious code detection method based on big data | |
| WO2015146113A1 (en) | Identification dictionary learning system, identification dictionary learning method, and recording medium | |
| CN111241987B (en) | Multi-target model visual tracking method based on cost-sensitive three-branch decision | |
| CN110647897B (en) | Zero sample image classification and identification method based on multi-part attention mechanism | |
| CN107729863B (en) | Human finger vein recognition method | |
| CN110909678B (en) | Face recognition method and system based on width learning network feature extraction | |
| CN105844299B (en) | A kind of image classification method based on bag of words | |
| CN110969128A (en) | Method for detecting infrared ship under sea surface background based on multi-feature fusion | |
| US11715288B2 (en) | Optical character recognition using specialized confidence functions | |
| CN113033305B (en) | Living body detection method, living body detection device, terminal equipment and storage medium | |
| CN115511867A (en) | Weaving method and system for high-wear-resistance textile fabric | |
| CN112330562B (en) | Heterogeneous remote sensing image transformation method and system | |
| CN108154107B (en) | Method for determining scene category to which remote sensing image belongs | |
| CN110555342B (en) | Image identification method and device and image equipment | |
| Chakroun et al. | Hierarchical convolutional neural networks for leaf disease detection | |
| CN111931757A (en) | Finger vein quick sorting method and device based on MDLBP block histogram and PCA dimension reduction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20221011 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |