CN108615002A - A kind of palm vein authentication method based on convolutional neural networks - Google Patents
A kind of palm vein authentication method based on convolutional neural networks Download PDFInfo
- Publication number
- CN108615002A CN108615002A CN201810363951.3A CN201810363951A CN108615002A CN 108615002 A CN108615002 A CN 108615002A CN 201810363951 A CN201810363951 A CN 201810363951A CN 108615002 A CN108615002 A CN 108615002A
- Authority
- CN
- China
- Prior art keywords
- convolutional
- layer
- neural networks
- image
- convolutional neural
- 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
-
- 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
- G06V40/14—Vascular patterns
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Collating Specific Patterns (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The palm vein authentication method based on convolutional neural networks that the invention discloses a kind of, the palm vein authentication method based on convolutional neural networks of being somebody's turn to do are as follows：S1, according to training sample set image, convolutional neural networks are trained, S2, user's registration image are input to convolutional network model and generate feature vector, and S3, input images to be recognized, identification is compared with the template characteristic vector in template memory module in S4, feature vector to be identified；The probability value that S5, comparison result obtain is maximized, and most probable value is more than certain threshold value, then certification success, otherwise authentification failure.The present invention passes through improved model, significantly compact model scale, and the method for a variety of data set joint trainings is provided to solve the disadvantage that sample size is small in authentication application, the likelihood probability value that multi-layer perception (MLP) calculates its matching identification is used in the feature verification stage, the parameter of multi-layer perception (MLP) on-line training and can automatically update, and improve certification speed and precision.
Description
Technical field
The present invention relates to vein identification technology field more particularly to a kind of palm vein authentications based on convolutional neural networks
Method.
Background technology
With the development of technology, biometrics identification technology is more and more universal, gradually replaces traditional password
Etc. identification authentication modes, brought great convenience.Traditional fingerprint recognition has had been applied to security protection, gate inhibition, gold
The every field such as melt.Vein identification technology is also gradually paid close attention to by researcher and commercial company due to its natural antifalsification.
Vena metacarpea is distributed under skin surface, belongs to the internal physiological feature of palm live body, is had very high safety, uniqueness and is prevented
The features such as puppet is strong.
Palm vein is quiet compared to finger vena and the back of the hand to have more application advantage, such as：(1) palm vein possesses more rich
Veinprint information, texture structure is more complicated, have more uniqueness；(2) acquisition of palm vein is easier, is more friendly
Good, without fixed palm position, light source design is more prone to reliable.Current hand vein recognition algorithm be mainly based upon characteristic point and
It is compared, some significant key points that wherein characteristic point is primarily referred to as in veinprint, is had very based on textural characteristics
Strong identification, such as endpoint, bifurcation, intersection point.Extraction common practice for these characteristic points is using SIFT, SURF etc.
Method, generally when characteristic point quantity is more, effect is fine for these methods, but time-consuming also high, is not suitable for answering in embedded device
With.In addition, the method based on textural characteristics is also more universal at present, such as LBP features, HOG features, these methods are clear to image
Clear degree is no longer sensitive, but its ability to express is also very limited, cannot cope with changeable practical application scene.
Convolutional neural networks are a new technologies for image processing field for rising in recent years, image classification,
The breakthrough in the fields such as semantic segmentation, target detection makes various image application product performances be increased substantially.It is right
In the application based on big data, convolutional neural networks can obtain the spy of the higher level of abstraction in various images by advance study
Sign, the resolution capability of these features is extremely strong, and minute differences even if in image s can distinguish.At present in authentication class
Using, be such as all the application of small sample a small range in fingerprint, vein etc., it is difficult to by pre-training obtain one it is preferable
Identification model.
To solve the above problems, the present invention proposes the palm vein authentication method based on channel packet convolutional neural networks,
Traditional convolutional neural networks are improved, aspect ratio pair is used for using multi-layer perception (MLP), is compared compared to traditional distance
Calculation amount is also greatly reduced while keeping authentication precision in mode accuracy rate higher.
Invention content
The shortcomings that aiming to overcome that existing convolutional neural networks of the present invention provides a kind of based on convolutional neural networks
Palm vein authentication method.
To achieve the above object, present invention employs following technical solutions：
A kind of palm vein authentication method based on convolutional neural networks is somebody's turn to do the palm vein based on convolutional neural networks and is recognized
Card method is as follows：
S1, according to training sample set image, convolutional neural networks are trained, a kind of volume based on channel packet is obtained
Product network model；
S2, user's registration image are input to convolutional network model and generate feature vector, and store to template and deposit as template
Store up module；
S3, input images to be recognized generate feature vector to be identified through the extraction of the convolutional network aspect of model；
Identification is compared with the template characteristic vector in template memory module in S4, feature vector to be identified；
The probability value that S5, comparison result obtain is maximized, and most probable value is more than certain threshold value, then certification success, no
Then authentification failure.
Preferably, the convolutional network model in step S1 includes multiple convolutional layers, non-linear layer, full articulamentum, SoftMax
Classification layer and multi-layer perception (MLP), it is specific as follows：
Convolutional layer in convolutional network model is divided into channel packet convolutional layer and channel fusion convolutional layer, channel packet convolution
Layer refers to is grouped independent convolution to each characteristic plane of input feature vector using the convolution kernel of 5 × 5 sizes, and convolution is merged in channel
Layer, which refers to, uses the convolution kernel of 1 × 1 size into row of channels integration in each channel of the output feature of the channel packet convolutional layer；
Convolutional network model has used average down-sampled technology, allows the image for inputting arbitrary size, last in convolutional layer
The characteristic plane of output is equally divided into 8 × 8 equal portions, is maximized the output valve as this part to the pixel in every portion, passes through
The input picture vectors that finally all output is tieed up for 8 × 8 of average down-sampled rear arbitrary size；
Multi-layer perception (MLP) is divided into three layers, and first layer and the second layer are characterized extract layer, last layer is classification layer, the network
Parameter needs trained in user's registration, the neuron number of last layer changes also with the number of users of registration.
Preferably, the method that the training sample set in step S1 uses a variety of vein image combined trainings, including palm are quiet
Arteries and veins image, finger venous image and hand back vein image, these three images are obtained by shooting, collecting, to have under near infrared light
Similar patterned feature can be such that training sample set image data is expanded, wherein containing VERA-Palmvein 100
The public databases such as classification, FINGER VEIN 192 classifications of USM and the vein image data collected certainly.
Preferably, the training sample set in step S1 additionally uses a series of data augmentation strategy, makes the training sample be in
Existing diversity, it is specific as follows：
Greyscale transformation is carried out to training sample set image, simulates the image effect under various brightness；
Random site cutting is carried out to training sample set image；
Random angles rotation is carried out to training sample set image；
Random proportional sizes compression, amplification are carried out to training sample set image.
Preferably, being trained to convolutional neural networks in step S1, includes the following steps：
When training, full articulamentum and SoftMax classification layers are added in convolutional neural networks, training convolutional is come with classification task
The weight parameter of layer；
Convolutional network is trained on ImageNet data sets in advance, then utilizes pre-training parameter in training sample set
In palm vein database, finger vena database, be trained in hand back vein database and weight parameter fine tuning；
Pass through the training classification task on different data sets so that convolutional layer learns to obtain extraction veinprint feature
Ability.
Compared with prior art, the beneficial effects of the invention are as follows：By improved model, significantly compact model scale,
And the method for a variety of data set joint trainings is provided to solve the disadvantage that sample size is small in authentication application, recognize in feature
The card stage uses the likelihood probability value that multi-layer perception (MLP) calculates its matching identification, the parameter of multi-layer perception (MLP) can on-line training and
It automatically updates, improves certification speed and precision.
Description of the drawings
Fig. 1 is technical solution of the present invention implementing procedure figure；
Fig. 2 is the vein image schematic diagram of training sample set of the present invention；
Fig. 3 is sample of the present invention image random cropping schematic diagram；
Fig. 4 is sample of the present invention image rotation processing schematic diagram；
Fig. 5 is channel packet convolutional network part-structure figure of the present invention；
Fig. 6 is convolutional network structure convolution schematic diagram of the present invention；
Fig. 7 is multi-layer perception (MLP) structural schematic diagram of the present invention.
Specific implementation mode
In order to make the purpose , technical scheme and advantage of the present invention be clearer, below in conjunction with specific embodiment, to this
Invention is further elaborated.
Embodiment 1
- 7 are please referred to Fig.1, the present invention provides a kind of technical solution, a kind of palm vein authentication based on convolutional neural networks
Method, the palm vein authentication method based on convolutional neural networks of being somebody's turn to do are as follows：
Step S1, according to training sample set image, convolutional neural networks are trained, are obtained a kind of based on channel packet
Convolutional network model.
Specifically, in order to enable the parameter of network to be more accurately fitted the feature of palm vein, present invention employs
The method of a variety of vein image combined trainings, as shown in Fig. 2, including palm vein image, finger venous image and hand back vein
Image, these three images are the shooting, collecting acquisition under near infrared light, have similar patterned feature, can make training sample
Collection image data is expanded, wherein containing VERA-Palmvein (100 classifications), FINGER VEIN USM (FV-USM
192 classifications) etc. public databases and the vein image data that collect certainly.
Further, to make trained sample that diversity be presented, a series of data are used to training sample set image
Add lustre to strategy, including：
Greyscale transformation is carried out to training sample set image, simulates the image effect under various brightness；
Random site cutting is carried out to training sample set image；
Random angles rotation is carried out to training sample set image；
Random proportional sizes compression, amplification are carried out to training sample set image.
More specifically, as shown in figure 3, random cropping to training sample set image, in order to allow the training of convolutional network to learn
Acquistion to parameter adapt to the scene of image translation, 4 kinds are randomly selected from 8 kinds of cutting methods to training sample image, is made
Training sample expands 4 times.8 kinds of cutting methods include the upper left corner 3/4, the upper right corner 3/4, the lower left corner 3/4, the lower right corner for cutting image
3/4, top 3/4, lower part 3/4, left part 3/4, right part 3/4.
Described carries out random angles rotation to training sample set image, as shown in figure 4, in practical application scene, by
It can cause collected vein image there are certain angle offset, technical solution of the present invention in the use habit of user is different
In, training sample image is rotated to 5 angles of random selection between 45 degree in -45 degree, makes 5 times of training sample expansion.
Described is trained convolutional neural networks, and convolutional neural networks are appointed on ImageNet data sets with classification
Business training obtains training pattern, then by the model in VERA-Palmvein (100 classifications), FINGER VEIN USM (FV-
192 classifications of USM) data set on finely tune, be finally finely adjusted training using the vein image collected certainly.
Being trained to weight parameter therein by big data sample makes its penalty values be preferably minimized, and initial learning rate is set
It is set to 0.05, is changed in the form of exponential damping, stops instruction not when being obviously reduced until decaying to 1e-10 or penalty values
Practice.Batchsize is dimensioned to 56, and removes average value processing to input picture.
The convolutional network model, network structure are 10 layers intermediate as shown in figure 5, wherein first layer is full convolution
Convolutional layer is merged for the channel packet convolutional layer at interval and channel, is most followed by one layer of average down-sampled layer and full articulamentum.Channel
The convolution kernel of 5 × 5 sizes is all used in grouping convolutional layer, later immediately down-sampled and nonlinear activation layer；It merges in channel
Convolutional layer is all using the convolution kernel of 1 × 1 size.
In order to adapt to the input picture of different size resolution ratio, the present invention provides a kind of average down-sampled technologies, such as：Volume
The characteristic plane of product output is M × M sizes, and as the input of averagely down-sampled layer, and the output size of average down-sampled layer is solid
M × M sizes are equally divided into 8 × 8 equal portions by positioning 8 × 8, then the pixel being maximized to each equal portions is as final output, nothing
Resolution ratio by input picture is much, and the feature vector finally exported is 8 × 8 sizes.
Convolutional network structure convolution schematic diagram as shown in FIG. 6, M are the port number of input picture, DKIt is convolution kernel
Size, N be convolution output port number, DFIt is the size of input picture, parameter amount is DK×DK× M × N, calculation amount are
DK×DK×DF×DF×M×N。
Traditional full convolution is divided into two steps by convolution mode in the present invention：First, each channel is carried out individual
Convolution operation exports M channel, size DF×DF；Second, 1 × 1 fusion convolution is carried out to this M channel, is exported N number of logical
Road, parameter amount are DK×DK× M+1 × 1 × M × N, and calculation amount is DK×DK×DF×DF×M+1×1×DF×DF×M×
N, calculation amount, nuclear parameter amount are obtained for and substantially reduce in contrast.
Step S2, user's registration image is input to convolutional network model and generates feature vector, and as template storage to mould
Plate memory module.
Specifically, when user registers for the first time, the palm under its near infrared light is shot, collect vein image and is incited somebody to action
It is input in convolutional network model and generates feature vector, and the template characteristic as matching identification is stored to template memory module.
Step S3, images to be recognized is inputted, feature vector to be identified is generated through the extraction of the convolutional network aspect of model.
Specifically, in authentication phase, by shooting, collecting under near infrared light to palm vein image, as to be identified
Image is input in convolutional network model, carries out feature extraction, is generated feature vector to be identified and is carried out with template characteristic vector
Matching identification.
Step S4, identification is compared with the template characteristic vector in template memory module in feature vector to be identified.
Specifically, in characteristic matching network, multi-layer perception (MLP) is used, as shown in fig. 7, the network is divided into three layers, first
Layer and the second layer are characterized extract layer, last layer is classification layer.The last output data of convolutional network structure as shown in Figure 5
By the input as multi-layer perception (MLP), final output is the probability value of generic, i.e., special with the template of registration user's storage
Levy the likelihood probability value of image.
Step S5, comparison result obtain probability value be maximized, most probable value be more than certain threshold value, then certification at
Work(, otherwise authentification failure.
Specifically, through multi-layer perception (MLP) export likelihood probability value, take its maximum value, with convolutional neural networks training and adjust
Whole optimal threshold is compared, and is more than the threshold value, is then regarded the most probable value to be effectively matched, and corresponding user
Store the success of template characteristic image authentication；Otherwise, authentification failure or it is considered as non-registered users.
By improved model, significantly compact model scale, and provide the method for a variety of data set joint trainings with
The disadvantage that sample size is small in authentication application is solved, multi-layer perception (MLP) is used in the feature verification stage and calculates it and compare and know
Other likelihood probability value, the parameter of multi-layer perception (MLP) on-line training and can automatically update, and improve certification speed and precision.
The foregoing is only a preferred embodiment of the present invention, but scope of protection of the present invention is not limited thereto,
Any one skilled in the art in the technical scope disclosed by the present invention, according to the technique and scheme of the present invention and its
Inventive concept is subject to equivalent substitution or change, should be covered by the protection scope of the present invention.
Claims (5)
1. a kind of palm vein authentication method based on convolutional neural networks, which is characterized in that should be based on convolutional neural networks
Palm vein authentication method is as follows：
S1, according to training sample set image, convolutional neural networks are trained, a kind of convolution net based on channel packet is obtained
Network model；
S2, user's registration image are input to convolutional network model and generate feature vector, and store mould to template as template storage
Block；
S3, input images to be recognized generate feature vector to be identified through the extraction of the convolutional network aspect of model；
Identification is compared with the template characteristic vector in template memory module in S4, feature vector to be identified；
The probability value that S5, comparison result obtain is maximized, and most probable value is more than certain threshold value, then certification success, otherwise recognizes
Card failure.
2. a kind of palm vein authentication method based on convolutional neural networks according to claim 1, which is characterized in that step
Convolutional network model in rapid S1 includes multiple convolutional layers, non-linear layer, full articulamentum, SoftMax classification layers and Multilayer Perception
Machine, it is specific as follows：
Convolutional layer in convolutional network model is divided into channel packet convolutional layer and channel fusion convolutional layer, and channel packet convolutional layer refers to
Independent convolution is grouped using the convolution kernel of 5 × 5 sizes to each characteristic plane of input feature vector, channel fusion convolutional layer refers to
To each channel of the output feature of the channel packet convolutional layer using the convolution kernel of 1 × 1 size into row of channels integration；
Convolutional network model has used average down-sampled technology, allows the image for inputting arbitrary size, is finally exported in convolutional layer
Characteristic plane be equally divided into 8 × 8 equal portions, to it is every it is a in pixel be maximized the output valve as this part, by average
The input picture of the arbitrary size vector that finally all output is tieed up for 8 × 8 after down-sampled；
Multi-layer perception (MLP) is divided into three layers, and first layer and the second layer are characterized extract layer, last layer is classification layer, the ginseng of the network
Number needs to train in user's registration, and the neuron number of last layer changes also with the number of users of registration.
3. a kind of palm vein authentication method based on convolutional neural networks according to claim 1, which is characterized in that step
The method that training sample set in rapid S1 uses a variety of vein image combined trainings, including palm vein image, finger vena figure
Picture and hand back vein image, these three images are to have similar patterned feature under near infrared light obtained by shooting, collecting, can
So that training sample set image data is expanded, wherein containing 100 classifications of VERA-Palmvein, FINGER VEIN
The public databases such as 192 classifications of USM and the vein image data collected certainly.
4. a kind of palm vein authentication method based on convolutional neural networks according to claim 2, which is characterized in that step
Training sample set in rapid S1 additionally uses a series of data augmentation strategy, makes training sample that diversity be presented, specific as follows：
Greyscale transformation is carried out to training sample set image, simulates the image effect under various brightness；
Random site cutting is carried out to training sample set image；
Random angles rotation is carried out to training sample set image；
Random proportional sizes compression, amplification are carried out to training sample set image.
5. a kind of palm vein authentication method based on convolutional neural networks according to claim 1, which is characterized in that step
Being trained to convolutional neural networks in rapid S1, includes the following steps：
When training, full articulamentum and SoftMax classification layers are added in convolutional neural networks, training convolutional layer is come with classification task
Weight parameter；
Convolutional network is trained on ImageNet data sets in advance, then pre-training parameter is utilized to be concentrated in training sample
It is trained in palm vein database, finger vena database, hand back vein database and is finely tuned with weight parameter；
Pass through the training classification task on different data sets so that convolutional layer learns to obtain the energy of extraction veinprint feature
Power.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810363951.3A CN108615002A (en) | 2018-04-22 | 2018-04-22 | A kind of palm vein authentication method based on convolutional neural networks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810363951.3A CN108615002A (en) | 2018-04-22 | 2018-04-22 | A kind of palm vein authentication method based on convolutional neural networks |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108615002A true CN108615002A (en) | 2018-10-02 |
Family
ID=63660671
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810363951.3A Pending CN108615002A (en) | 2018-04-22 | 2018-04-22 | A kind of palm vein authentication method based on convolutional neural networks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108615002A (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109815869A (en) * | 2019-01-16 | 2019-05-28 | 浙江理工大学 | A kind of finger vein identification method based on the full convolutional network of FCN |
| CN110163182A (en) * | 2019-05-30 | 2019-08-23 | 辽宁工业大学 | A kind of hand back vein identification method based on KAZE feature |
| CN110211069A (en) * | 2019-06-03 | 2019-09-06 | 广东工业大学 | A kind of image denoising model and training method, system, equipment, computer media |
| CN110321871A (en) * | 2019-07-11 | 2019-10-11 | 电子科技大学成都学院 | A kind of palm vein identification system and method based on LSTM |
| CN110414373A (en) * | 2019-07-08 | 2019-11-05 | 武汉大学 | A kind of deep learning palm vein identification system and method based on cloud side end cooperated computing |
| CN110717372A (en) * | 2019-08-13 | 2020-01-21 | 平安科技（深圳）有限公司 | Identity verification method and device based on finger vein recognition |
| CN111178221A (en) * | 2019-12-24 | 2020-05-19 | 珠海格力电器股份有限公司 | Identity recognition method and device |
| CN112861743A (en) * | 2021-02-20 | 2021-05-28 | 厦门熵基科技有限公司 | Palm vein image anti-counterfeiting method, device and equipment |
| CN113139627A (en) * | 2021-06-22 | 2021-07-20 | 北京小白世纪网络科技有限公司 | Mediastinal lump identification method, system and device |
| CN113486758A (en) * | 2021-06-30 | 2021-10-08 | 浙江大学 | Automatic hand acupuncture point positioning method |
| CN113657498A (en) * | 2021-08-17 | 2021-11-16 | 展讯通信（上海）有限公司 | Biological feature extraction method, training method, authentication method, device and equipment |
| JP2021533493A (en) * | 2019-04-03 | 2021-12-02 | 平安科技(深▲せん▼)有限公司Ping An Technology (Shenzhen) Co., Ltd. | Finger vein matching method, device, computer equipment, and storage medium |
| WO2023160048A1 (en) * | 2022-02-28 | 2023-08-31 | 腾讯科技（深圳）有限公司 | Palmprint sample generation method and apparatus, and device, medium and program product |
| CN117809341A (en) * | 2024-02-29 | 2024-04-02 | 广州市仪美医用家具科技股份有限公司 | Palm vein-based identity recognition method, system, equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020082815A1 (en) * | 2000-12-22 | 2002-06-27 | Isabelle Rey-Fabret | Method for forming an optimized neural network module intended to simulate the flow mode of a multiphase fluid stream |
| CN106529468A (en) * | 2016-11-07 | 2017-03-22 | 重庆工商大学 | Finger vein identification method and system based on convolutional neural network |
| CN106650721A (en) * | 2016-12-28 | 2017-05-10 | 吴晓军 | Industrial character identification method based on convolution neural network |
-
2018
- 2018-04-22 CN CN201810363951.3A patent/CN108615002A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020082815A1 (en) * | 2000-12-22 | 2002-06-27 | Isabelle Rey-Fabret | Method for forming an optimized neural network module intended to simulate the flow mode of a multiphase fluid stream |
| CN106529468A (en) * | 2016-11-07 | 2017-03-22 | 重庆工商大学 | Finger vein identification method and system based on convolutional neural network |
| CN106650721A (en) * | 2016-12-28 | 2017-05-10 | 吴晓军 | Industrial character identification method based on convolution neural network |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109815869A (en) * | 2019-01-16 | 2019-05-28 | 浙江理工大学 | A kind of finger vein identification method based on the full convolutional network of FCN |
| JP2021533493A (en) * | 2019-04-03 | 2021-12-02 | 平安科技(深▲せん▼)有限公司Ping An Technology (Shenzhen) Co., Ltd. | Finger vein matching method, device, computer equipment, and storage medium |
| JP7141518B2 (en) | 2019-04-03 | 2022-09-22 | 平安科技（深▲せん▼）有限公司 | Finger vein matching method, device, computer equipment, and storage medium |
| US11893773B2 (en) | 2019-04-03 | 2024-02-06 | Ping An Technology (Shenzhen) Co., Ltd. | Finger vein comparison method, computer equipment, and storage medium |
| CN110163182A (en) * | 2019-05-30 | 2019-08-23 | 辽宁工业大学 | A kind of hand back vein identification method based on KAZE feature |
| CN110211069B (en) * | 2019-06-03 | 2021-09-03 | 广东工业大学 | Image denoising model training method, system, equipment and computer medium |
| CN110211069A (en) * | 2019-06-03 | 2019-09-06 | 广东工业大学 | A kind of image denoising model and training method, system, equipment, computer media |
| CN110414373A (en) * | 2019-07-08 | 2019-11-05 | 武汉大学 | A kind of deep learning palm vein identification system and method based on cloud side end cooperated computing |
| CN110414373B (en) * | 2019-07-08 | 2021-09-24 | 武汉大学 | Deep learning palm vein recognition system and method based on cloud edge-side cooperative computing |
| CN110321871A (en) * | 2019-07-11 | 2019-10-11 | 电子科技大学成都学院 | A kind of palm vein identification system and method based on LSTM |
| CN110717372A (en) * | 2019-08-13 | 2020-01-21 | 平安科技（深圳）有限公司 | Identity verification method and device based on finger vein recognition |
| CN111178221A (en) * | 2019-12-24 | 2020-05-19 | 珠海格力电器股份有限公司 | Identity recognition method and device |
| CN112861743A (en) * | 2021-02-20 | 2021-05-28 | 厦门熵基科技有限公司 | Palm vein image anti-counterfeiting method, device and equipment |
| CN113139627A (en) * | 2021-06-22 | 2021-07-20 | 北京小白世纪网络科技有限公司 | Mediastinal lump identification method, system and device |
| CN113486758A (en) * | 2021-06-30 | 2021-10-08 | 浙江大学 | Automatic hand acupuncture point positioning method |
| CN113486758B (en) * | 2021-06-30 | 2024-03-08 | 浙江大学 | Automatic hand acupoint positioning method |
| CN113657498A (en) * | 2021-08-17 | 2021-11-16 | 展讯通信（上海）有限公司 | Biological feature extraction method, training method, authentication method, device and equipment |
| WO2023160048A1 (en) * | 2022-02-28 | 2023-08-31 | 腾讯科技（深圳）有限公司 | Palmprint sample generation method and apparatus, and device, medium and program product |
| CN117809341A (en) * | 2024-02-29 | 2024-04-02 | 广州市仪美医用家具科技股份有限公司 | Palm vein-based identity recognition method, system, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108615002A (en) | A kind of palm vein authentication method based on convolutional neural networks | |
| CN107886064B (en) | Face recognition scene adaptation method based on convolutional neural network | |
| CN106529468B (en) | A kind of finger vein identification method and system based on convolutional neural networks | |
| Xin et al. | Multimodal feature-level fusion for biometrics identification system on IoMT platform | |
| CN106096535A (en) | A kind of face verification method based on bilinearity associating CNN | |
| CN106709418B (en) | Face identification method and identification device based on scene photograph and certificate photo | |
| CN106599994A (en) | Sight line estimation method based on depth regression network | |
| CN107392082A (en) | A kind of small area fingerprint comparison method based on deep learning | |
| CN107958217A (en) | A kind of fingerprint classification identifying system and method based on deep learning | |
| CN107169479A (en) | Intelligent mobile equipment sensitive data means of defence based on fingerprint authentication | |
| Liu et al. | Fingerprint pore matching using deep features | |
| CN109740578A (en) | It is a kind of suitable for illumination, the face identification method of posture, expression shape change | |
| CN106909909A (en) | A kind of Face datection and alignment schemes based on shared convolution feature | |
| CN108875907A (en) | A kind of fingerprint identification method and device based on deep learning | |
| Xu et al. | An effective recognition approach for contactless palmprint | |
| Bakhshi et al. | End to end fingerprint verification based on convolutional neural network | |
| Huang et al. | Robust finger vein recognition based on deep CNN with spatial attention and bias field correction | |
| Wu et al. | Parallel multi-path age distinguish network for cross-age face recognition | |
| CN110929243A (en) | Pedestrian identity recognition method based on mobile phone inertial sensor | |
| Ramakrishnan et al. | An efficient automatic attendance system using fingerprint reconstruction technique | |
| NAWWAR et al. | Improvement of confusion matrix for hand vein recognition based on deep-learning multi-classifier decisions | |
| Liu et al. | A novel high-resolution fingerprint representation method | |
| Rao et al. | Amalgamation Biometric Deep Features in Smart City-ITS Authentication | |
| Hashemi et al. | Biometric identification through hand geometry | |
| Liu et al. | An improved finger vein recognition model with a residual attention mechanism |
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: 20181002 |
|
| RJ01 | Rejection of invention patent application after publication |