CN106326886B - Finger vein image quality appraisal procedure based on convolutional neural networks - Google Patents
Finger vein image quality appraisal procedure based on convolutional neural networks Download PDFInfo
- Publication number
- CN106326886B CN106326886B CN201610979315.4A CN201610979315A CN106326886B CN 106326886 B CN106326886 B CN 106326886B CN 201610979315 A CN201610979315 A CN 201610979315A CN 106326886 B CN106326886 B CN 106326886B
- Authority
- CN
- China
- Prior art keywords
- image
- quality
- neural networks
- convolutional neural
- layer
- 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.)
- Active
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
- G06V40/12—Fingerprints or palmprints
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- 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
- G06V40/12—Fingerprints or palmprints
- G06V40/1365—Matching; Classification
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Human Computer Interaction (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
- Collating Specific Patterns (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of finger vein image quality appraisal procedure and system based on convolutional neural networks, this method is that opponent first refers to that the quality of vein gray level image is labeled, next establishes training sample set, is then trained using training sample set to convolutional neural networks model.Finally any one width gray level image and bianry image are input in trained model, the output for choosing the second full articulamentum in two convolutional neural networks models respectively is the depth characteristic vector of input gray level image and bianry image；It connects two depth characteristic vectors and forms Combined expression vector, and be entered into support vector machines and be trained, the quality of prediction finger venous image is calculated using probabilistic SVMs；The appraisal procedure and assessment system can largely promote the precision of finger vein image quality assessment, improve the recognition performance of Verification System.
Description
Technical field
The invention belongs to biometrics identification technology field, in particular to a kind of finger vena based on convolutional neural networks
Image quality measure method and assessment system.
Background technique
With the growth of Internet technology fast development and information security threats, how effectively to discriminate one's identification to protect individual
Become urgent problem with property safety.Compared with traditional authentication mode such as key and password, it is based on physiology and behavior
Biological characteristic be difficult to be stolen, replicate and lose.Therefore, biometrics are applied by extensive research and successfully
In personal identification.Following two: 1 outside mode such as face, fingerprint, palmmprint and rainbow can be divided into based on physiological biological mode
Film；2 Internal biological mode: finger vena, palm vein and hand back vein.System based on external biological mode is subject to attack
It hits.Such as, it is easy to a width fingerprint template is stolen and forged to attack fingerprint recognition system.It is different from external biological mode,
It is interior biological mode be located at finger epidermis under make it difficult to be stolen and forge, therefore they have higher safety
Energy.
Since the acquisition process of finger venous image is influenced by many factors, such as environment light, environment temperature, light dissipate
It penetrates, the behavior of the change of physiological characteristic and user, therefore finger vena identification is still a challenging task.Such as
Fruit cannot overcome these factors well, then including a large amount of low-quality images in the image acquired.In general, these low-qualitys
Spirogram picture eventually reduces the performance of Verification System.Finger vein image quality assessment as a kind of effective solution by
Extensive research.In existing finger vein image quality assessment algorithm, researcher assumes these factors such as picture contrast
It is related with the quality of image with vein quantity.Then, such as using some description artificially designed: Radon transformation, Gaussian Energy
These attributes of model, Gabor filter, curvature measuring.Existing method utilizes mankind's intuition or biometric image quality
Priori knowledge come determine influence quality attribute and using by hand description son these attributes are extracted, such as
CN101866486 discloses a kind of finger vein image quality judging method, and this method is by obtaining finger venous image
Contrast mass fraction, positional shift mass fraction, effective coverage mass fraction, direction fuzziness mass fraction obtain matter in turn
Amount score adds up carry out overall merit by weight, establishes the overall merit mass function of finger venous image, and then comes pair
Finger vein image quality is assessed, but these methods still have following disadvantage:
1 to be difficult the attribute for proving to choose by hand be centainly relevant to the picture quality of finger vena.For example, some be based on
Human vision or the high quality graphic of understanding are but certified system refusal.
The attribute of 2 researchers picture quality it is not possible that inquiry agency has an impact.
Even if 3 these attributes are positively correlated with the quality of image, it is also difficult to establish effective mathematical model and go to retouch them.
Summary of the invention
In order to overcome the above-mentioned deficiencies of the prior art, the present invention provides a kind of finger venas based on convolutional neural networks
Image quality measure method and assessment system.First, overcome conventional finger vein image quality appraisal procedure by intuition or
Person's priori knowledge the shortcomings that evaluating picture quality, can more objectively evaluate picture quality.Second, the invention energy
The quality tab of image is enough automatically generated, to reduce hard work and error caused by artificial mark.Third, the hair
It is bright to learn from original finger venous image automatically to feature relevant to picture quality, it avoids and artificially selects and mention
The problem of taking discrimination feature.
Specific technical solution of the present invention is as follows:
A kind of finger vein image quality appraisal procedure based on convolutional neural networks, this method comprises the following steps:
S1: being labeled the quality of finger vena gray level image in database, obtains the grayscale image for having quality tab
Picture marks out low quality gray level image and high quality gray level image, and obtains the vein of the gray level image with quality tab
Feature obtains bianry image after being encoded；
S2: the bianry image training sample set of quality tab is had in establishment step S1；
S3: the gray level image training sample set of quality tab is had in establishment step S1；
S4: the convolutional neural networks model of gray level image depth characteristic is extracted；The convolutional neural networks model includes: defeated
Enter layer, the first convolutional layer, the first pond layer, the second convolutional layer, the second pond layer, third convolutional layer, the first full articulamentum, second
Full articulamentum and output layer；
S5: the convolutional neural networks model of bianry image depth characteristic is extracted；The convolutional neural networks model includes: defeated
Enter layer, the first convolutional layer, the first pond layer, the second convolutional layer, the second pond layer, the first full articulamentum, the second full articulamentum and
Output layer；
S6: the training of convolutional neural networks model
Each layer filter is initialized using the random number of Gaussian distributed, the initial value of offset is arbitrary constant；It adopts
Convolutional neural networks are trained with stochastic gradient descent method；Bianry image training sample set that step S2 is established and
The gray level image training sample set that step S3 is established is divided into different subclass, is separately input to step S5 and step in batches
In convolutional neural networks model applied by rapid S4, when the image of all batches carries out a forward direction in convolutional neural networks model
After propagation, calculate gradient and carry out backpropagation with update filter power and offset, by iterate find filter and
The optimal solution of offset；
S7: after completing training, it will predict that finger venous image is input to the convolutional neural networks mould of step S4 and step S5
In type, the output of the second full articulamentum is one width grayscale image of input in convolutional neural networks model in selecting step S4 and S5 step
The depth characteristic vector of picture and bianry image；It connects two depth characteristic vectors and forms a width input prediction finger venous image
Combined expression vector；
S8: the Combined expression vector that step S7 is formed is input in support vector machines and is trained, is supported using probability
Vector machine predicts the quality of finger venous image to calculate.
It is further to improve, the quality of finger vena gray level image in database is labeled method particularly includes:
S11: the selection of enrollment image
Select a finger appoints piece image, and two width finger venas are extracted and matched using mature recognizer method
Image, and calculate the image and remaining width image averaging value distance；Select minimum average B configuration apart from corresponding image as the finger
Enrollment image, other images are as test image；
S12: the mark of picture quality
The distance between every width test image of same finger and its enrollment image is calculated to obtain matching score in class；
The distance between each enrollment image is calculated to obtain matching score between class；Score is matched between score and class according to matching in class,
Calculate the receptance FAR of mistake and the reject rate FRR of mistake；A threshold value is preset, when the receptance FAR of mistake is equal to the threshold value
When such as FAR=0.1%, then according to whether distinguishing low quality by image that system mistake is refused or the correctly image that receives
Image or high quality graphic.
It is further to improve, in the first convolutional layer, the second convolutional layer or third convolutional layer, l layers of characteristic imageIt presses
It is calculated according to following formula:
Wherein,It is l layers of input spectrum,It is the convolution kernel between m-th of input and n output characteristic spectrum, *
It is convolution operation, Ml-1It is the quantity of input feature vector spectrum,It is the offset of n-th output spectra.
It is further to improve, use in the first convolutional layer, the second convolutional layer or third convolutional layer correct linear unit as
Excitation function is defined as follows:
Wherein,Indicate l layers of output spectra.
It is further to improve, it is in the first pond layer, the second pond layer that the output of the first convolutional layer and the second convolutional layer is special
Sign spectrum divides the region that does not overlap, chooses the mean value of preceding p maximum value in each region as the typical value in the region to the
The output of one convolutional layer and the second convolutional layer is sampled；If IkIndicate the output spectra after k-th of convolution kernel carries out convolution,It indicates to IkAll elements in middle s × s regional areaCarry out from greatly to
The set obtained after small sequence, T=s × s indicate the number of element；IkThe output feature obtained after after samplingAccording to as follows
Formula calculates:
Further to improve, g walks filter and weighs wgUpdate rule are as follows:
wg+1=Δg+1+wg
Wherein Δ indicates momentum, and λ is learning rate,For wgGradient.
Further to improve, the probabilistic SVMs used are the matter by combined depth feature vector V and it
It measures label q ∈ { 0,1 }, probabilistic SVMs is trained, output probability value is p
ξ (v) indicates the output of traditional support vector machine, and ω and γ indicate two ginsengs that probabilistic SVMs training obtains
Number.
Another aspect of the present invention provides a kind of finger vein image quality assessment system based on convolutional neural networks, this is commented
Estimate the database that system includes assessment unit and communicates with assessment unit, the databases contain finger vena grayscale image
Picture, the assessment unit include:
Quality annotation module is labeled for the quality to finger vena gray level image, is obtained with quality tab
Gray level image, and the vein pattern of the gray level image with quality tab is obtained, bianry image is obtained after being encoded；
Training sample set builds formwork erection block jointly, bianry image with quality tab for obtaining quality annotation module and
Gray level image establishes bianry image training sample set and gray level image training sample set respectively；
Model building module, for establishing the convolutional neural networks for extracting bianry image and gray level image depth characteristic respectively
Model；
Convolution training module, for training sample set to be built jointly to the bianry image training sample set and ash of the foundation of formwork erection block
Degree image training sample set is divided into different subclass, is separately input to extract bianry image and gray level image depth in batches
In the corresponding convolutional neural networks model of feature, it is trained；
Processing module is connected, for obtaining gray level image and bianry image in the convolutional neural networks model after training
Depth characteristic vector；And Combined expression vector is formed for connecting two depth characteristic vectors；
Computing module is trained for Combined expression vector to be input in support vector machines, calculates prediction finger
The quality of vein image.
Further to improve, quality annotation module includes:
Image selection and feature extraction submodule, for optionally selecting a width grayscale image from several images of same root finger
As and carry out feature extraction and obtain bianry image；
Enrollment image selection submodule, for calculating the image of image selection and the selection of feature extraction submodule and same
A piece finger is left the average value distance of image, selects minimum average B configuration apart from corresponding image as enrollment image, other
Image is as test image；
Computational submodule, the distance between every width test image and its enrollment image for calculating same finger
Score is matched in class, the distance between each enrollment image is calculated and obtains matching score between class；And according to matching point in class
Score is matched between several and class, calculates the receptance FAR of mistake and the reject rate FRR of mistake；
Whether judging submodule, the receptance FAR for misjudgment are equal to preset threshold value, if the receptance of mistake
FAR is equal to the threshold value, sends sort instructions to classification submodule；
Classify submodule, for the image of False Rejects will to be labeled with and correctly the image that receives is classified, and to mark
Infuse the instruction that submodule sends mark；
Submodule is marked, for the image of False Rejects will to be labeled with or correctly the image that receives carries out quality annotation, and
Set corresponding quality tab.
Compared with prior art, the beneficial effects of the present invention are: the finger provided by the invention based on convolutional neural networks
Vein image quality appraisal procedure and assessment system can largely promote the precision of finger vein image quality assessment,
Improve Verification System recognition performance, compared with other finger vein image quality appraisal procedures, its advantages be embodied in
Under several aspects:
1. the finger vein image quality appraisal procedure and assessment system energy provided by the invention based on convolutional neural networks
It is enough that automatically finger vena gray level image is labeled, to reduce artificial mark bring hard work and error.
2. finger vein image quality appraisal procedure and assessment system provided by the invention based on convolutional neural networks are first
The depth characteristic of secondary fusion finger vena bianry image and gray level image realizes the quality estimation of finger vein image.
3. the convolutional neural networks model difference that the present invention uses is compared with traditional convolutional neural networks model:
First, for all convolutional layers, using the maximum value of corresponding position element between calculating this layer of characteristic spectrum of input as this layer
Characteristic spectrum is simultaneously entered into activation primitive；Second, in all pond layers, calculate preceding p in the regional area of characteristic image
The mean value of a maximum value samples input feature vector image.
4. the present invention is merged to depth characteristic using probability supporting vector and is predicted the quality of finger venous image, from
And effectively improve the precision of image quality measure.
5. the finger vein image quality appraisal procedure and assessment system provided by the invention based on convolutional neural networks, no
It is only applicable to the quality evaluation of finger venous image, and can be applied in other biological characteristic image quality evaluation.
Detailed description of the invention
Fig. 1 is the flow chart of finger vein image quality appraisal procedure of the embodiment 1 based on convolutional neural networks；
Fig. 2 is the convolutional neural networks model structure schematic diagram for the depth characteristic that embodiment 4 extracts gray level image；
Fig. 3 is the convolutional neural networks model structure schematic diagram for the depth characteristic that embodiment 4 extracts bianry image；
Fig. 4 is the structural block diagram of finger vein image quality assessment system of the embodiment 6 based on convolutional neural networks；
Fig. 5 is the structural block diagram of 7 quality annotation module of embodiment.
Specific embodiment
Embodiment 1
A kind of finger vein image quality appraisal procedure based on convolutional neural networks, as shown in Figure 1, this method includes such as
Lower step:
S1: being labeled the quality of finger vena gray level image in database, obtains the grayscale image for having quality tab
Picture obtains low quality gray level image and high quality gray level image, and obtains the vein pattern of the gray level image with quality tab,
Bianry image is obtained after being encoded；
S2: the bianry image training sample set with quality tab obtained in establishment step S1；
S3: the gray level image training sample set of quality tab is had in establishment step S1；
S4: the convolutional neural networks model of gray level image depth characteristic is extracted；The convolutional neural networks model includes: defeated
Enter layer, the first convolutional layer, the first pond layer, the second convolutional layer, the second pond layer, third convolutional layer, the first full articulamentum, second
Full articulamentum and output layer；
S5: the convolutional neural networks model of bianry image depth characteristic is extracted；The convolutional neural networks model includes: defeated
Enter layer, the first convolutional layer, the first pond layer, the second convolutional layer, the second pond layer, the first full articulamentum, the second full articulamentum and
Output layer；
S6: the training of convolutional neural networks model
Each layer filter is initialized using the random number of Gaussian distributed, the initial value of offset is arbitrary constant；It adopts
Convolutional neural networks are trained with stochastic gradient descent method；Bianry image training sample set that step S2 is established and
The gray level image training sample set that step S3 is established is divided into different subclass, is separately input to step S5 and step in batches
In convolutional neural networks model applied by rapid S4, when the image of all batches carries out a forward direction in convolutional neural networks model
After propagation, calculate gradient and carry out backpropagation with update filter power and offset, by iterate find filter and
The optimal solution of offset；
S7: after completing training, it will predict that finger venous image is input to the convolutional neural networks mould of step S4 and step S5
In type, the output of the second full articulamentum is one width grayscale image of input in convolutional neural networks model in selecting step S4 and S5 step
The depth characteristic vector of picture and bianry image；It connects two depth characteristic vectors and forms a width input prediction finger venous image
Combined expression vector；
S8: the Combined expression vector that step S7 is formed is input in support vector machines and is trained, is supported using probability
Vector machine predicts the quality of finger venous image to calculate.
Finger vein image quality appraisal procedure provided by the invention based on convolutional neural networks can be largely
The precision for promoting finger vein image quality assessment, improves the recognition performance of Verification System, method provided by the invention is melted for the first time
The depth characteristic for closing finger vena bianry image and gray level image realizes that the quality of finger vein image is estimated.
Embodiment 2
A kind of finger vein image quality appraisal procedure based on convolutional neural networks, the method are different from embodiment 1
, the quality to finger vena gray level image in database is labeled method particularly includes:
S11: the selection of enrollment image
Select a finger appoints piece image, and two width finger venas are extracted and matched using mature recognizer method
Image, and calculate the image and remaining width image averaging value distance；Select minimum average B configuration apart from corresponding image as the finger
Enrollment image, other images are as test image；
S12: the mark of picture quality
The distance between every width test image of same finger and its enrollment image is calculated to obtain matching score in class；
The distance between each enrollment image is calculated to obtain matching score between class；Score is matched between score and class according to matching in class,
Calculate the receptance FAR of mistake and the reject rate FRR of mistake；Default security level of the threshold value as system, when mistake
When receptance FAR is equal to the threshold value such as FAR=0.1%, then according to whether by image that system mistake is refused or correctly receiving
Image distinguishes low-quality image or high quality graphic.
Finger venous image in sample set provided by the invention derives from the database http of The Hong Kong Polytechnic University: //
Www4.comp.polyu.edu.hk/~csajaykr/fvdatabase.htm.The database includes the 3132 width hands of 156 people
Refer to vein image；In two stages, each finger provides 6 width image patterns in each acquisition phase, often for the acquisition of data
Individual provides two fingers, so everyone provides 24 width images two stages；Wherein, preceding 105 people is in two acquisition ranks
Section provides 2520 images；Remaining 51 people are only involved in the Image Acquisition of second stage, a total of 612 width image.Due to
Two phase acquisition images more tally with the actual situation, so the present invention only only used 2520 images acquired from 105 people
105 people × 2 piece fingers × for 6 width image × 2 stages are introduced.
The quality for illustrating the present invention finger venous image is labeled that the specific method is as follows:
The extraction and matching of first step vein pattern
The extraction of 1.1 vein patterns: mainly image is enhanced using following Gabor wavelet
In formula, pn=[x, y]TIt indicates in reference axis both horizontally and vertically, p0=[x0,y0]TIt is the distance to origin,
ωmIt is flat rate, C is 2 × 2 positive definite covariance matrixes, | | indicate dot-product operation.Pass through coordinate transformWithThe Gabor filter of not Tongfang can be arrived, whereinθnBe rotation angle and by from
The direction dispersion K is such asWherein q=1,2 ..., K (K=8).
Enhance finger vein features by following equation；
WhereinIndicate GθThe mean value of (x, y)) .* expression volume
Product, f (x, y) is finger venous image.
Vein pattern is further enhanced using morphological operation；
WithIt indicates to carry out gray scale expansion and corrosion to image by structural element b.
Then, characteristic image Z is encoded to obtain bianry image using following equation；
What the matching of vein pattern was mainly realized by the following method:
The matching of 1.2 vein patterns:
Assuming that R and T respectively indicate the two-value registered images and test image of m × n；By being extended to obtain template to R
ImageSuch as template is obtained by the way that its length and width are expanded to 2w+m and 2h+n and is expressed as follows:
Matching score between R and T calculates as follows:
W and h indicates the distance moved in the horizontal and vertical directions in formula；
Φ is defined as follows:
The selection of second step enrollment image
There are 210 fingers in database, every finger there are 12 width finger venous images.All images by front method
Extract vein pattern bianry image.Then optionally take a width bianry image and using matching algorithm above-mentioned calculate the image with
It is left all image averaging value distances.The operation is repeated, the average distance of other images is calculated.Finally select minimum average B configuration distance
Template image of the corresponding image as the finger, other images are as test image.Therefore, 210 width register mould in data
Plate and 2310 width test images.
The mark of third step picture quality
The distance between the same every width test image of finger and its template image, common property raw 2320 are calculated according to matching algorithm
Score is matched in class.Correspondingly, by matching score between available 21945 class of the distance between calculation template.According between class
The receptance FAR of mistake and the reject rate (FRR) of mistake are calculated with score is matched in class；Have at one compared with high safety grade
Under the conditions of FAR be equal to 0.1%, 0.1% is preset value, and marking image refuse by system mistake is low-quality image, is marked
Label are set as 0；The image labeling correctly received by system is high quality graphic, and label is set as 1.
Finger vein image quality appraisal procedure provided by the invention based on convolutional neural networks and other finger venas
Image quality measure method is compared, and can automatically finger vein image be labeled, to reduce artificial mark band
Hard work and error.
Embodiment 3
A kind of finger vein image quality appraisal procedure based on convolutional neural networks, the method are different from embodiment 1
, the bianry image training sample set with quality tab that obtains in step S2 establishment step S1 method particularly includes:
After marking all test images according to the mask method of step S1, the 1155 width images of 105 fingers are chosen as training image,
Remaining image is as test image；In training set, a total of 101 width low-quality image and 1054 panel height quality images.
In test set, high quality and low-quality image are respectively 110 width and 1045 width.Due to low-mass ratio in training set
The sample of high quality is few, leads to all kinds of imbalances；In order to overcome this problem, benefit generates low-quality figure in the following method
Picture；For example, in order to generate the synthesis sample of low-quality image x, the optional two width low-quality image x first from training set1With
x2.Then, equation y is utilizedl=x1+rand(0,1)(x2-x1) one interim image pattern of (l=1,2 ..., L) generation.Most
Afterwards, pass through equation pl=x1+rand(0,1)(yl- x) (l=1,2 ..., K) calculate synthesis new samples.It, can according to this method
To generate the image of 953 width synthesis, so that low-quality image total in training set is 1054 width.
The gray level image training sample set of quality tab is had in step S3 in establishment step S1 method particularly includes: by
In the corresponding width gray level image of every width bianry image, therefore it is based on according to the bianry image training sample set of front is available
The gray level image training sample set of gray level image.
Embodiment 4
A kind of finger vein image quality appraisal procedure based on convolutional neural networks, the method are different from embodiment 1
, the convolutional neural networks model of gray level image depth characteristic is extracted in step S4, as shown in Fig. 2, first convolutional layer,
In second convolutional layer or third convolutional layer, l layers of characteristic imageIn each element be equal to upper one layer of all characteristic pattern in it is right
Answer the maximum value of position, characteristic imageIt is calculated according to following formula:
Wherein,It is l layers of input spectrum,It is the convolution kernel between m-th of input and n output characteristic spectrum, *
It is convolution operation, Ml-1It is the quantity of input feature vector spectrum,It is the offset of n-th output spectra.
Use amendment linear unit as excitation function in first convolutional layer, the second convolutional layer or third convolutional layer,
It is defined as follows:
In formula,Indicate l layers of output spectra.
The output characteristic spectrum of the first convolutional layer and the second convolutional layer is divided mutually in first pond layer, the second pond layer
Nonoverlapping region, choose the mean value of preceding p maximum value in each region as the typical value in the region to the first convolutional layer or
The output of second convolutional layer is sampled；If IkIndicate the output spectra after k-th of convolution kernel carries out convolution,It indicates to IkAll elements in middle s × s regional areaCarry out from greatly to
The set obtained after small sequence, T=s × s indicate the number of element；To IkThe output feature obtained after samplingAccording to as follows
Formula calculates:
The neuron of half is randomly discharged in the first full articulamentum and the second full articulamentum using discarding method.
In the output layer, the probability of N=2 class is predicted using softmax function；
Wherein,The output x of the last one hidden layermLinear combination.
The convolutional neural networks model that bianry image depth characteristic is extracted in step S5 is as shown in Figure 3.
Compared with traditional convolutional neural networks model, the convolutional neural networks model difference that the present invention uses is: the
One, the spy for all convolutional layers, using the maximum value of corresponding position element between calculating this layer of characteristic spectrum of input as this layer
Sign is composed and is entered into activation primitive；Second, to calculating in all pond layers, preceding p in the regional area of characteristic image is a
The mean value of maximum value samples input feature vector image.
Embodiment 5
A kind of finger vein image quality appraisal procedure based on convolutional neural networks, the method are different from embodiment 4
, the training of convolutional neural networks model method particularly includes:
1. initializing each layer filter using the random number of Gaussian distributed, the initial value of offset is arbitrary constant；
Using stochastic gradient descent method come convolutional neural networks model shown in training step S4 and step S5.
1. its quality tab is q ∈ { 0,1 } 2. for piece image F, wherein 0 indicates low-quality image, 1 indicates high
Quality image；Training set is expressed as { (F1,q1),(F2,q2),…,(FN,qN)}；Training dataset is divided into different subsets
It closes, is input in convolutional neural networks model applied by step S4 and step S5 in batches；When the image of all batches is in net
After network carries out a propagated forward, calculates gradient and carry out backpropagation to update filter power wkWith offset bk；Such as: g
It walks filter and weighs wgUpdate rule are as follows:
wg+1=Δg+1+wg
Wherein Δ indicates momentum, and λ is learning rate,For the gradient of wg.
3. finding the optimal solution of filter and offset by iterating.When precision is met the requirements, stop iteration, thus
Complete the training of this deep neural network model.
Step S7's method particularly includes: after completing training, remove the output layer of convolutional neural networks, when one width gray scale of input
For image into the convolutional neural networks model of step S4, the second full articulamentum will export a depth characteristic vector；The vector is
It is expressed for the depth of input gray level image；When input a width bianry image into the convolutional neural networks model of step S5, second
Full articulamentum exports the depth characteristic vector of a two straight images；Assuming that v1And v2A respectively width gray level image and corresponding two
It is worth the depth characteristic vector of image；The Combined expression vector to a width input picture is formed by connecting two depth characteristic vectors
V=[v1v2], then, which is input in support vector machines and is trained；
Step S8's method particularly includes: the Evaluation Model on Quality based on support vector machines:
Image quality measure based on support vector machines carrys out the quality of forecast image using probabilistic SVMs in closing.
Its probabilistic SVMs used is defined as follows: by combined depth feature vector v and its quality tab q ∈ { 0,1 },
Probabilistic SVMs are trained, output probability value is p
ξ (v) indicates the output of traditional support vector machine, and ω and γ indicate two ginsengs that probabilistic SVMs training obtains
Number.After training, probabilistic SVMs can calculate the quality of any input feature value v corresponding image.
The present invention merges to depth characteristic using probability supporting vector and predicts the quality of finger venous image, thus
Effectively improve the precision of image quality measure.
Embodiment 6
A kind of finger vein image quality assessment system based on convolutional neural networks, as shown in figure 4, the assessment system packet
Assessment unit 1 and the database to communicate with assessment unit 12 are included, is stored with finger vena gray level image in the database 2,
The assessment unit 1 includes:
Quality annotation module 11 is labeled for the quality to finger vena gray level image, is obtained and is had quality tab
Gray level image, and obtain with quality tab gray level image vein pattern, obtain bianry image after being encoded；
Training sample set builds formwork erection block 12 jointly, the binary map with quality tab for obtaining quality annotation module 11
Picture and gray level image establish bianry image training sample set and gray level image training sample set respectively；
Model building module 13, for establishing the convolutional Neural net for extracting bianry image and gray level image depth characteristic respectively
Network model；
Convolution training module 14, for building training sample set jointly bianry image training sample set that formwork erection block 12 is established
It is divided into different subclass with gray level image training sample set, is separately input to extract bianry image and gray level image in batches
In the corresponding convolutional neural networks model of depth characteristic, it is trained；
Processing module 15 is connected, for obtaining gray level image and binary map in the convolutional neural networks model after training
The depth characteristic vector of picture；And Combined expression vector is formed for connecting two depth characteristic vectors；
Computing module 16 is trained for Combined expression vector to be input in support vector machines, calculates prediction hand
Refer to the quality of vein image.
Finger vein image quality assessment system energy provided by the invention based on convolutional neural networks
Enough precision for largely promoting finger vein image quality assessment, improve the recognition performance of Verification System, with other fingers
Vein image quality appraisal procedure realizes opponent compared to the depth characteristic for merging finger vena bianry image and gray level image for the first time
Refer to the quality estimation of vein image.
Embodiment 7
A kind of finger vein image quality assessment system based on convolutional neural networks, the assessment system and embodiment 6 are not
With as shown in figure 5, quality annotation module 11 includes:
Image selection and feature extraction submodule 110, for optionally selecting width ash from several images of same root finger
It spends image and carries out feature extraction and obtain bianry image；
Enrollment image selection submodule 111, for calculating the figure of image selection and the selection of feature extraction submodule 110
Average value distance as being left image with same root finger, selects minimum average B configuration apart from corresponding image as enrollment figure
Picture, other images are as test image；
Computational submodule 112, for calculate between every width test image of same finger and its enrollment image away from
From obtaining matching score in class, calculates the distance between each enrollment image and obtain matching score between class；And according in class
With score is matched between score and class, the receptance FAR of mistake and the reject rate FRR of mistake are calculated；
Whether judging submodule 113, the receptance FAR for misjudgment are equal to preset threshold value, if mistake connects
It is equal to the threshold value by rate FAR, sends sort instructions to classification submodule 114；
Classify submodule 114, for the image of False Rejects will to be labeled with and correctly the image that receives is classified, and to
Mark the instruction that submodule 115 sends mark；
Submodule 115 is marked, for carrying out quality annotation to the image for being labeled with False Rejects or the image correctly received,
And set corresponding quality tab.
Finger vein image quality assessment system provided by the invention based on convolutional neural networks and other finger venas
Image quality measure system is compared, and can automatically finger vein image be labeled, to reduce artificial mark band
Hard work and error.And the present invention provides assessment system and can more accurately finger vein image be labeled, and mentions
Height mark quality.
Finally it should be noted that the above examples are only used to illustrate the technical scheme of the present invention and are not limiting, by this
The technical solution of invention is modified or replaced equivalently, without departing from the spirit and scope of the technical solution of the present invention, should all
It is included within the scope of the claims of the present invention.
Claims (5)
1. a kind of finger vein image quality appraisal procedure based on convolutional neural networks, which is characterized in that the method includes
Following steps:
S1: being labeled the quality of finger vena gray level image in database, obtains the gray level image for having quality tab, and
The vein pattern for obtaining the gray level image with quality tab, obtains bianry image after being encoded；
S2: the bianry image training sample set with quality tab obtained in establishment step S1；
S3: the gray level image training sample set of quality tab is had in establishment step S1；
S4: the convolutional neural networks model of gray level image depth characteristic is extracted；The convolutional neural networks model includes: input
Layer, the first convolutional layer, the first pond layer, the second convolutional layer, the second pond layer, third convolutional layer, the first full articulamentum, second are entirely
Articulamentum and output layer；
S5: the convolutional neural networks model of bianry image depth characteristic is extracted；The convolutional neural networks model includes: input
Layer, the first convolutional layer, the first pond layer, the second convolutional layer, the second pond layer, the first full articulamentum, the second full articulamentum and defeated
Layer out；
In first convolutional layer, the second convolutional layer or third convolutional layer, l layers of characteristic imageAccording to following formula meter
It calculates:
Wherein,It is l layers of input spectrum,It is the convolution kernel between m-th of input and n output characteristic spectrum, * is convolution
Operation, Ml-1It is the quantity of input feature vector spectrum,It is the offset of n-th of output spectra；
The division of the output characteristic spectrum of the first convolutional layer and the second convolutional layer is not weighed mutually in first pond layer, the second pond layer
The mean value of preceding p maximum value in each region is chosen as the typical value in the region to the first convolutional layer or second in folded region
The output of convolutional layer is sampled；If IkIndicate the output spectra after k-th of convolution kernel carries out convolution,It indicates to IkMiddle s × s
All elements in regional areaThe set obtained after being sorted from large to small, wherein 0≤t < T-1,0
≤ m, n < s, T=s × s indicate the number of element；To IkThe output feature obtained after samplingIt is calculated according to following formula:
Wherein, p≤T；
S6: the training of convolutional neural networks model
Each layer filter is initialized using the random number of Gaussian distributed, the initial value of offset is arbitrary constant；Using with
Machine gradient descent method is trained convolutional neural networks；By the step S2 bianry image training sample set established and step
The gray level image training sample set that S3 is established is divided into different subclass, is separately input to step S5 and step S4 in batches
In applied convolutional neural networks model, when the image of all batches carries out a propagated forward in convolutional neural networks model
Afterwards, it calculates gradient and carries out backpropagation to update filter power and offset, find filter and offset by iterating
Optimal solution；
S7: after completing training, prediction finger venous image being input in the convolutional neural networks model of step S4 and step S5,
In selecting step S4 and S5 step in convolutional neural networks model the second full articulamentum output be input one width gray level image and
The depth characteristic vector of bianry image；Connect the joint that two depth characteristic vectors form a width input prediction finger venous image
Express vector；
S8: the Combined expression vector that step S7 is formed is input in support vector machines and is trained, probability supporting vector is used
Machine predicts the quality of finger venous image to calculate.
2. appraisal procedure according to claim 1, which is characterized in that described to finger vena gray level image in database
Quality is labeled method particularly includes:
S11: the selection of enrollment image
Select a finger appoints piece image, is extracted using recognizer method and matches two width finger venous images, and counted
Calculate the image and remaining width image averaging value distance；Select enrollment of the minimum average B configuration apart from corresponding image as the finger
Image, other images are as test image；
S12: the mark of picture quality
The distance between every width test image of same finger and its enrollment image is calculated to obtain matching score in class；It calculates
The distance between each enrollment image obtains matching score between class；Score is matched between score and class according to matching in class, is calculated
The receptance FAR of the mistake and reject rate FRR of mistake；A threshold value is preset, when receptance FAR is equal to preset threshold value, then
According to whether being labeled as the image of False Rejects by system or correctly the image that receives distinguishes low quality gray level image or height
Quality gray image.
3. appraisal procedure according to claim 1, which is characterized in that first convolutional layer, the second convolutional layer or third
It uses amendment linear unit as excitation function in convolutional layer, is defined as follows:
Wherein,Indicate l layers of output spectra.
4. appraisal procedure according to claim 1, which is characterized in that g walks filter and weighs wgUpdate rule are as follows:
wg+1=Δg+1+wg
Wherein Δ indicates momentum, and λ is learning rate,For wgGradient.
5. appraisal procedure according to claim 1, which is characterized in that the probabilistic SVMs used are to pass through connection
The quality tab q ∈ { 0,1 } for closing depth characteristic vector V and it, is trained probabilistic SVMs, output probability value is
p
ξ (v) indicates the output of traditional support vector machine, and ω and γ indicate two parameters that probabilistic SVMs training obtains.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610979315.4A CN106326886B (en) | 2016-11-07 | 2016-11-07 | Finger vein image quality appraisal procedure based on convolutional neural networks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610979315.4A CN106326886B (en) | 2016-11-07 | 2016-11-07 | Finger vein image quality appraisal procedure based on convolutional neural networks |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106326886A CN106326886A (en) | 2017-01-11 |
| CN106326886B true CN106326886B (en) | 2019-05-10 |
Family
ID=57816696
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610979315.4A Active CN106326886B (en) | 2016-11-07 | 2016-11-07 | Finger vein image quality appraisal procedure based on convolutional neural networks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106326886B (en) |
Families Citing this family (43)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106295475A (en) * | 2015-05-29 | 2017-01-04 | 北京东方金指科技有限公司 | A kind of based on sum of ranks than the registered fingerprint replacement method of method |
| CN106910192B (en) * | 2017-03-06 | 2020-09-22 | 长沙全度影像科技有限公司 | Image fusion effect evaluation method based on convolutional neural network |
| CN106920224B (en) * | 2017-03-06 | 2019-11-05 | 长沙全度影像科技有限公司 | A method of assessment stitching image clarity |
| CN106920215B (en) * | 2017-03-06 | 2020-03-27 | 长沙全度影像科技有限公司 | Method for detecting registration effect of panoramic image |
| WO2018187953A1 (en) * | 2017-04-12 | 2018-10-18 | 邹霞 | Facial recognition method based on neural network |
| CN108305240B (en) * | 2017-05-22 | 2020-04-28 | 腾讯科技（深圳）有限公司 | Image quality detection method and device |
| CN107273927B (en) * | 2017-06-13 | 2020-09-22 | 西北工业大学 | Unsupervised field adaptive classification method based on inter-class matching |
| CN107657209B (en) * | 2017-07-07 | 2020-07-28 | 杭州电子科技大学 | Template image registration mechanism based on finger vein image quality |
| CN107644415B (en) * | 2017-09-08 | 2019-02-22 | 众安信息技术服务有限公司 | A kind of text image method for evaluating quality and equipment |
| CN107705299B (en) * | 2017-09-25 | 2021-05-14 | 安徽睿极智能科技有限公司 | Image quality classification method based on multi-attribute features |
| CN107967442A (en) * | 2017-09-30 | 2018-04-27 | 广州智慧城市发展研究院 | A kind of finger vein identification method and system based on unsupervised learning and deep layer network |
| CN107832684B (en) * | 2017-10-26 | 2021-08-03 | 通华科技（大连）有限公司 | Intelligent vein authentication method and system with autonomous learning capability |
| CN107895144A (en) * | 2017-10-27 | 2018-04-10 | 重庆工商大学 | A kind of finger vein image anti-counterfeiting discrimination method and device |
| CN108010015A (en) * | 2017-11-07 | 2018-05-08 | 深圳市金城保密技术有限公司 | One kind refers to vein video quality evaluation method and its system |
| CN108053407B (en) * | 2017-12-22 | 2021-04-13 | 联想(北京)有限公司 | Data processing method and data processing system |
| JP2019121054A (en) * | 2017-12-28 | 2019-07-22 | 株式会社東海理化電機製作所 | Fingerprint authentication device |
| CN108288027B (en) * | 2017-12-28 | 2020-10-27 | 新智数字科技有限公司 | Image quality detection method, device and equipment |
| CN108764292B (en) * | 2018-04-27 | 2022-03-18 | 北京大学 | Deep learning image target mapping and positioning method based on weak supervision information |
| CN109360183B (en) * | 2018-08-20 | 2021-05-11 | 中国电子进出口有限公司 | Face image quality evaluation method and system based on convolutional neural network |
| CN109360633B (en) * | 2018-09-04 | 2022-08-30 | 北京市商汤科技开发有限公司 | Medical image processing method and device, processing equipment and storage medium |
| CN109409226B (en) * | 2018-09-25 | 2022-04-08 | 五邑大学 | Finger vein image quality evaluation method and device based on cascade optimization CNN |
| CN109272499B (en) * | 2018-09-25 | 2020-10-09 | 西安电子科技大学 | Non-reference image quality evaluation method based on convolution self-coding network |
| CN109409227A (en) * | 2018-09-25 | 2019-03-01 | 五邑大学 | A kind of finger vena plot quality appraisal procedure and its device based on multichannel CNN |
| CN109523514A (en) * | 2018-10-18 | 2019-03-26 | 西安电子科技大学 | To the batch imaging quality assessment method of Inverse Synthetic Aperture Radar ISAR |
| CN109409314A (en) * | 2018-11-07 | 2019-03-01 | 济南浪潮高新科技投资发展有限公司 | A kind of finger vein identification method and system based on enhancing network |
| CN109270384B (en) * | 2018-11-13 | 2019-06-11 | 中南民族大学 | A kind of method and system of the electric arc of electrical equipment for identification |
| CN109214376A (en) * | 2018-11-22 | 2019-01-15 | 济南浪潮高新科技投资发展有限公司 | A kind of fingerprint identification method and device based on depth stratification |
| CN109829887B (en) * | 2018-12-26 | 2022-07-01 | 南瑞集团有限公司 | Image quality evaluation method based on deep neural network |
| CN109934114B (en) * | 2019-02-15 | 2023-05-12 | 重庆工商大学 | Finger vein template generation and updating algorithm and system |
| CN109978840A (en) * | 2019-03-11 | 2019-07-05 | 太原理工大学 | A kind of method of discrimination of the quality containing texture image based on convolutional neural networks |
| CN110058943B (en) * | 2019-04-12 | 2021-09-21 | 三星（中国）半导体有限公司 | Memory optimization method and device for electronic device |
| CN110400335B (en) * | 2019-07-25 | 2022-05-24 | 广西科技大学 | Texture image quality estimation method based on deep learning |
| CN110532908B (en) * | 2019-08-16 | 2023-01-17 | 中国民航大学 | Finger vein image scattering removal method based on convolutional neural network |
| CN110674824A (en) * | 2019-09-26 | 2020-01-10 | 五邑大学 | Finger vein segmentation method and device based on R2U-Net and storage medium |
| CN110751105B (en) * | 2019-10-22 | 2022-04-08 | 珠海格力电器股份有限公司 | Finger image acquisition method and device and storage medium |
| CN111223082B (en) * | 2020-01-03 | 2023-06-20 | 圣点世纪科技股份有限公司 | Quantitative evaluation method for finger vein image quality |
| CN111210426B (en) * | 2020-01-15 | 2021-03-02 | 浙江大学 | Image quality scoring method based on non-limiting standard template |
| CN111709906A (en) * | 2020-04-13 | 2020-09-25 | 北京深睿博联科技有限责任公司 | Medical image quality evaluation method and device |
| CN111539306B (en) * | 2020-04-21 | 2021-07-06 | 中南大学 | Remote sensing image building identification method based on activation expression replaceability |
| CN111612083B (en) * | 2020-05-26 | 2023-05-12 | 济南博观智能科技有限公司 | Finger vein recognition method, device and equipment |
| CN112288010B (en) * | 2020-10-30 | 2022-05-13 | 黑龙江大学 | Finger vein image quality evaluation method based on network learning |
| CN112200156B (en) * | 2020-11-30 | 2021-04-30 | 四川圣点世纪科技有限公司 | Vein recognition model training method and device based on clustering assistance |
| CN114539586B (en) * | 2022-04-27 | 2022-07-19 | 河南银金达新材料股份有限公司 | Surface treatment production and detection process of polymer film |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101866486A (en) * | 2010-06-11 | 2010-10-20 | 哈尔滨工程大学 | Finger vein image quality judging method |
| CN103544705A (en) * | 2013-10-25 | 2014-01-29 | 华南理工大学 | Image quality testing method based on deep convolutional neural network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016054779A1 (en) * | 2014-10-09 | 2016-04-14 | Microsoft Technology Licensing, Llc | Spatial pyramid pooling networks for image processing |
-
2016
- 2016-11-07 CN CN201610979315.4A patent/CN106326886B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101866486A (en) * | 2010-06-11 | 2010-10-20 | 哈尔滨工程大学 | Finger vein image quality judging method |
| CN103544705A (en) * | 2013-10-25 | 2014-01-29 | 华南理工大学 | Image quality testing method based on deep convolutional neural network |
Non-Patent Citations (3)
| Title |
|---|
| Finger-Vein Quality Assessment by Representation Learning from Binary Images;Huafeng Qin et al.;《International Conference on Neural Information Processing 》;20151112;第31-32页第3.1节，第38页第3.3节，第41页第3.3.4节 |
| ImageNet Classification with Deep Convolutional Neural Networks;Krizhevsky A et al.;《Advances in neural information processing systems. 2012》;20121208;第6页第5节 |
| 手指静脉图像质量评估与特征识别算法研究;秦华锋;《中国博士学位论文全文数据库 信息科技辑》;20130515;第423-427页第2-3节，表2-3，图1-2 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106326886A (en) | 2017-01-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106326886B (en) | Finger vein image quality appraisal procedure based on convolutional neural networks | |
| CN106529468B (en) | A kind of finger vein identification method and system based on convolutional neural networks | |
| CN107194341B (en) | Face recognition method and system based on fusion of Maxout multi-convolution neural network | |
| CN106372581B (en) | Method for constructing and training face recognition feature extraction network | |
| CN107392082B (en) | Small-area fingerprint comparison method based on deep learning | |
| CN108427921A (en) | A kind of face identification method based on convolutional neural networks | |
| CN107886064A (en) | A kind of method that recognition of face scene based on convolutional neural networks adapts to | |
| Feng et al. | Robust and efficient algorithms for separating latent overlapped fingerprints | |
| CN101630364A (en) | Method for gait information processing and identity identification based on fusion feature | |
| CN106548159A (en) | Reticulate pattern facial image recognition method and device based on full convolutional neural networks | |
| CN107403084A (en) | A kind of personal identification method based on gait data | |
| CN109934114A (en) | A kind of finger vena template generation and more new algorithm and system | |
| CN107392187A (en) | A kind of human face in-vivo detection method based on gradient orientation histogram | |
| CN102214299A (en) | Method for positioning facial features based on improved ASM (Active Shape Model) algorithm | |
| Liu et al. | Fingerprint pore matching using deep features | |
| CN106778489A (en) | The method for building up and equipment of face 3D characteristic identity information banks | |
| AL-Allaf et al. | Pattern recognition neural network for improving the performance of iris recognition system | |
| CN109255339B (en) | Classification method based on self-adaptive deep forest human gait energy map | |
| CN106529395A (en) | Signature image recognition method based on deep brief network and k-means clustering | |
| CN107895144A (en) | A kind of finger vein image anti-counterfeiting discrimination method and device | |
| CN104361339B (en) | Slap shape Graph Extraction and recognition methods | |
| CN103714340A (en) | Self-adaptation feature extracting method based on image partitioning | |
| CN110263726A (en) | A kind of finger vein identification method and device based on depth correlation feature learning | |
| CN103942545A (en) | Method and device for identifying faces based on bidirectional compressed data space dimension reduction | |
| Qin et al. | Multi-scaling detection of singular points based on fully convolutional networks in fingerprint images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20230328 Address after: 401329 2F, Building 18, Section I, Science Valley, Hangu Town, Jiulongpo District, Chongqing Patentee after: Chongqing Financial Technology Research Institute Patentee after: Qin Huafeng Address before: 400067 No. 19, Xuefu Avenue, Nan'an District, Chongqing Patentee before: CHONGQING TECHNOLOGY AND BUSINESS University |
|
| TR01 | Transfer of patent right |