CN107392114A - A kind of finger vein identification method and system based on neural network model - Google Patents
A kind of finger vein identification method and system based on neural network model Download PDFInfo
- Publication number
- CN107392114A CN107392114A CN201710515995.9A CN201710515995A CN107392114A CN 107392114 A CN107392114 A CN 107392114A CN 201710515995 A CN201710515995 A CN 201710515995A CN 107392114 A CN107392114 A CN 107392114A
- Authority
- CN
- China
- Prior art keywords
- image information
- finger vein
- neural network
- processing
- image
- 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
- 210000003462 vein Anatomy 0.000 title claims abstract description 136
- 238000003062 neural network model Methods 0.000 title claims abstract description 57
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000012545 processing Methods 0.000 claims abstract description 128
- 230000009467 reduction Effects 0.000 claims abstract description 40
- 238000000605 extraction Methods 0.000 claims abstract description 31
- 238000010606 normalization Methods 0.000 claims description 62
- 238000012549 training Methods 0.000 claims description 36
- 238000007781 pre-processing Methods 0.000 claims description 30
- 238000013528 artificial neural network Methods 0.000 claims description 27
- 239000013598 vector Substances 0.000 claims description 26
- 238000004364 calculation method Methods 0.000 claims description 15
- 238000001914 filtration Methods 0.000 claims description 15
- 238000003672 processing method Methods 0.000 claims description 7
- 230000011218 segmentation Effects 0.000 claims description 6
- 238000003709 image segmentation Methods 0.000 claims description 3
- 230000007935 neutral effect Effects 0.000 abstract 1
- 230000008569 process Effects 0.000 description 10
- 238000005516 engineering process Methods 0.000 description 9
- 230000006870 function Effects 0.000 description 8
- 230000008859 change Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 230000000007 visual effect Effects 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 230000007547 defect Effects 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 2
- 238000012800 visualization Methods 0.000 description 2
- 238000007476 Maximum Likelihood Methods 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 230000003542 behavioural effect Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000003331 infrared imaging Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000007480 spreading Effects 0.000 description 1
- 238000002054 transplantation Methods 0.000 description 1
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
- G06V40/1347—Preprocessing; Feature extraction
-
- 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
- G06N3/08—Learning methods
-
- 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)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Collating Specific Patterns (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention discloses a kind of finger vein identification method and system based on neural network model, wherein the finger vein identification method includes：Obtain the finger venous image information of user to be identified；Image information pretreatment is carried out to the finger venous image information, obtains image information after pretreatment；Dimensionality reduction part binary feature extraction process is carried out to image information after the pretreatment, obtains the finger vein features information of user to be identified；Processing is identified to the finger vein pattern information of the user to be identified using by the neural network model that depth belief network is trained, identifies the identity information of the user to be identified.In embodiments of the present invention, can largely solve the problems, such as that refer to vein recognition system changes for light, improves the redundancy to noise, and improve matching accuracy using neutral net in matching stage.
Description
Technical Field
The invention relates to the technical field of biological recognition, in particular to a finger vein recognition method and system based on a neural network model.
Background
Biometric technology (biometrics), i.e. distinguishing individuals based on the physiological and behavioral characteristics of a person itself. The problems of easy counterfeiting, easy damage and easy change of common identity identification means such as signature, voice, fingerprint and the like exist, and the problems of close range, unfriendliness and the like exist in iris, DNA, palm and the like. The biological characteristic information extraction and matching identification technology research of the finger veins can well overcome the defects, and the precision, the safety and the convenience of a biological identification system are improved. By combining the image processing technology, the machine learning technology, the embedded technology, the digital signal processing technology and the COS development technology, system integration and platform transplantation can be conveniently carried out. Therefore, the finger vein recognition technology becomes one of the first most promising important members in the current biometric feature recognition family, and is applied to the fields of building entrance guard, banks, ATM (automatic teller machines), PC (personal computer) login, automobile safety and the like.
But the vein identification technology still has a plurality of defects. Because the finger veins mostly adopt an infrared imaging mode, the illumination change amplitude of the collected images is large, and great difficulty is caused to matching identification. In addition, most of the recognition systems with high matching recognition rate need to involve a large amount of calculation for feature extraction and matching, and various key parameters need to be adjusted in experiments.
Disclosure of Invention
The invention aims to overcome the defects of the prior art, and provides a finger vein recognition method and system based on a neural network model, which can solve the problem of light change of a finger vein recognition system to a greater extent, improve the redundancy of noise, and improve the matching accuracy by using the neural network in the matching stage.
In order to solve the technical problem, the invention provides a finger vein recognition method based on a neural network model, which comprises the following steps:
acquiring finger vein image information of a user to be identified;
carrying out image information preprocessing on the finger vein image information to obtain preprocessed image information;
performing dimensionality reduction local binary feature extraction processing on the preprocessed image information to obtain finger vein feature information of a user to be identified;
and adopting a neural network model trained through a deep belief network to identify the finger vein characteristic information of the user to be identified, and identifying the identity information of the user to be identified.
Preferably, the image information preprocessing of the acquired finger vein image information includes:
carrying out graying processing on the finger vein image information to obtain grayed image information;
filtering and denoising the grayed image information to obtain denoised grayed image information;
and carrying out normalization processing on the noise-reduced gray image information to obtain a normalization processing result.
Preferably, the normalizing the noise-reduced grayscale image information includes:
carrying out normalization processing on each pixel point of the noise-reduced grayed image information by adopting an MAX-MIN method, wherein the gray value of each pixel point is in a range of [0,255] after the normalization processing;
and normalizing the image size of the de-noised grayed image information by adopting a bilinear interpolation method, wherein the size of the processed image size is in a range of 64 x 64.
Preferably, the performing the dimension reduction local binary feature extraction processing on the preprocessed image information includes:
4 × 4 segmentation processing is carried out on the image information with the size of 64 × 64 after the preprocessing, 16 image blocks are obtained, and the pixel of each image block is 16 × 16;
comparing each pixel in each image block with 8 field pixels of the pixels, if the field pixels are larger than the pixels, setting the field pixels to be 0, otherwise, setting the field pixels to be 1, and acquiring 8-bit binary numbers as 256-dimensional features of the corresponding image blocks;
carrying out dimension reduction processing on 256-dimensional features of each image block, wherein each processed image block is 59-dimensional features;
performing histogram calculation processing on 59-dimensional features of each processed image block to obtain a dimension-reduced LBP histogram;
carrying out normalization processing on the dimension-reduced LBP histogram to obtain a normalized dimension-reduced LBP histogram;
connecting the normalized dimension-reduced LBP histogram of each image block in series to obtain serial characteristic vector data;
and carrying out normalization processing on the serial feature vector data, wherein each processed feature vector data is [0,0.5 ].
Preferably, the recognizing the finger vein feature information of the user to be recognized by using a neural network model trained through a deep belief network includes:
and inputting the acquired finger vein characteristic information of the user to be recognized into a neural network model trained through a deep belief network for matching, outputting a matching result of the user to be recognized, and recognizing the identity information of the user to be recognized.
Preferably, the training step of the neural network model trained by the deep belief network includes:
acquiring finger vein image information of a user to be trained;
image information preprocessing is carried out on the finger vein image information of the user to be trained, and the preprocessed image information to be trained is obtained;
performing dimensionality reduction local binary feature extraction processing on the preprocessed image information to be trained to obtain finger vein feature information of a user to be trained;
performing parameter initialization processing on the neural network by adopting a depth information network to obtain a parameter initialization neural network;
and inputting the finger vein characteristic information of the user to be trained into a parameter initialization neural network for training, and obtaining a neural network model for deep belief network training.
In addition, the embodiment of the present invention further provides a finger vein recognition system based on a neural network model, and the finger vein recognition system includes:
an image acquisition module: the finger vein recognition system is used for acquiring finger vein image information of a user to be recognized;
an image preprocessing module: the finger vein image processing device is used for preprocessing the image information of the finger vein image information and acquiring the preprocessed image information;
a feature extraction module: the system is used for performing dimensionality reduction local binary feature extraction processing on the preprocessed image information to acquire finger vein feature information of a user to be identified;
an identity recognition module: the neural network model is trained through a deep belief network and used for identifying the finger vein characteristic information of the user to be identified, and identifying the identity information of the user to be identified.
Preferably, the image preprocessing module comprises:
a graying processing unit: the finger vein image processing device is used for carrying out gray processing on the finger vein image information to obtain gray image information;
a noise reduction processing unit: the system is used for carrying out filtering noise reduction processing on the grayed image information to obtain the grayed image information after noise reduction;
a normalization processing unit: the normalization processing module is used for performing normalization processing on the grayed image information subjected to noise reduction to obtain a normalization processing result;
preferably, the feature extraction module includes:
an image segmentation unit: the image information which is used for carrying out 4-4 segmentation processing on the preprocessed image information with the size of 64-64 is obtained, 16 image blocks are obtained, and the pixel of each image block is 16-16;
a pixel comparison unit: the image processing method comprises the steps of comparing each pixel in each image block with 8 field pixels of the pixels, setting the field pixels to be 0 if the field pixels are larger than the pixels, and otherwise setting the field pixels to be 1, and acquiring 8-bit binary numbers as 256-dimensional features of the corresponding image blocks;
a dimension reduction processing unit: the image processing method is used for carrying out dimension reduction processing on 256-dimensional features of each image block, and each processed image block is 59-dimensional features;
a histogram calculation unit: the histogram calculation processing module is used for performing histogram calculation processing on 59-dimensional features of each processed image block, and the processed image blocks are dimension-reduced LBP histograms;
a histogram normalization unit: the histogram is used for carrying out normalization processing on the dimension-reduced LBP histogram to obtain a normalized dimension-reduced LBP histogram;
a series unit: the normalized dimension-reduced LBP histogram is used for connecting each image block in series to obtain serial feature vector data;
a data normalization unit: and the method is used for carrying out normalization processing on the serial feature vector data, and each processed feature vector data is [0,0.5 ].
Preferably, the finger vein recognition system further comprises a neural network model training module:
an image acquisition unit: the device is used for acquiring finger vein image information of a user to be trained;
an image preprocessing unit: the device is used for preprocessing the image information of the finger vein image information of the user to be trained and acquiring the preprocessed image information to be trained;
a feature extraction unit: the system is used for performing dimensionality reduction local binary feature extraction processing on the preprocessed image information to be trained to acquire finger vein feature information of a user to be trained;
an initialization unit: the device is used for performing parameter initialization processing on the neural network by adopting a depth information network to obtain a parameter initialization neural network;
a training unit: and the neural network model is used for inputting the finger vein characteristic information of the user to be trained into the parameter initialization neural network for training, and acquiring the deep belief network training neural network model.
In the embodiment of the invention, the problem of light change of the finger vein recognition system can be solved to a greater extent, the redundancy to noise is improved, the overall neural network is initialized by using the deep belief network, the convergence rate of the neural network is improved, the operation cost is reduced, the matching time of the overall finger vein recognition system is reduced, the regularization parameters are used in the matching stage to prevent overfitting, the matching accuracy is improved, and the false rejection rate and the false recognition rate are reduced.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a flow chart of a method of a finger vein recognition method based on a neural network model in an embodiment of the present invention;
FIG. 2 is a schematic flow chart of the feature extraction process in the embodiment of the present invention;
FIG. 3 is a flow chart illustrating the training steps of the neural network model in an embodiment of the present invention;
FIG. 4 is a schematic structural component diagram of a finger vein recognition system based on a neural network model in an embodiment of the present invention;
fig. 5 is a schematic structural composition diagram of another finger vein recognition system based on a neural network model in the embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Fig. 1 is a schematic flow chart of a method for identifying finger veins based on a neural network model in an embodiment of the present invention, and as shown in fig. 1, the method for identifying finger veins includes:
s11: acquiring finger vein image information of a user to be identified;
s12: carrying out image information preprocessing on the finger vein image information to obtain preprocessed image information;
s13: performing dimensionality reduction local binary feature extraction processing on the preprocessed image information to obtain finger vein feature information of a user to be identified;
s14: and adopting a neural network model trained through a deep belief network to identify the finger vein characteristic information of the user to be identified, and identifying the identity information of the user to be identified.
Further explanation of S11 is:
acquiring finger vein image information of a user to be identified, placing a finger on a vein image information acquisition device by the user to be identified, and acquiring and storing the vein image information on the finger as the finger vein image information of the user to be identified by the vein image information acquisition device.
Further explanation of S12 is:
carrying out image information preprocessing on the finger vein image information to obtain preprocessed image information; in the preprocessing process, image information graying processing, filtering noise reduction processing and normalization processing are required to be performed in sequence.
Further comprising the following steps: carrying out graying processing on the finger vein image information to obtain grayed image information; filtering and denoising the grayed image information to obtain denoised grayed image information; and carrying out normalization processing on the noise-reduced gray image information to obtain a normalization processing result.
In the process of graying the finger vein image information, a graying method conforming to the characteristics of human eyes is adopted, so that:
Gray＝0.299R+0.587G+0.114B
where R, G, B is the three-pass value of the original input color image information.
In the process of filtering and denoising the grayed image information, mean filtering is adopted to realize the processing, so that:
where M is 9, I is a region of 3 × 3 around the pixel, f (x, y) is the gray level of the pixel, and g (x, y) is the gray level of the pixel after mean filtering.
Normalizing the de-noised grayed image information by adopting an MAX-MIN method, wherein the gray value of each pixel point of the de-noised grayed image information is in a range of [0,255] after the normalization processing; and normalizing the image size of the de-noised grayed image information by adopting a bilinear interpolation method, wherein the size of the processed image size is in a range of 64 x 64.
The pixel gray value normalization is to uniformly normalize the gray value of each pixel point to be 0-255, the normalization method is a MAX-MIN method, and the normalization formula is as follows:
wherein z (x, y) is the gray value of each pixel point after normalization.
The method comprises the steps of performing scale normalization processing by using a bilinear interpolation method, traversing each pixel point on image information, performing interpolation and adjustment in the x and y directions, and performing the scale normalization processing to obtain an image with the size of 64 x 64.
Further explanation of S13 is:
performing dimensionality reduction local binary feature extraction processing on the preprocessed image information to obtain finger vein feature information of a user to be identified; extracting dimension-reducing local binary features (LBP) from each image information with the size of 64 x 64, wherein the specific extraction process comprises the following steps: fig. 2 is a schematic processing flow diagram of the feature extraction processing in the embodiment of the present invention, as shown in fig. 2:
s131: 4 × 4 segmentation processing is carried out on the image information with the size of 64 × 64 after the preprocessing, 16 image blocks are obtained, and the pixel of each image block is 16 × 16;
further, the image information of the normalized image size of 64 × 64 is divided into blocks of 4 × 4, 16 blocks are formed after the division, and the pixel value of each block is 16 × 16.
S132: comparing each pixel in each image block with 8 field pixels of the pixels, if the field pixels are larger than the pixels, setting the field pixels to be 0, otherwise, setting the field pixels to be 1, and acquiring 8-bit binary numbers as 256-dimensional features of the corresponding image blocks;
further, each pixel in the image block with the size of 4 × 4 and the pixel value of 16 × 16 after the block division is compared with eight neighborhood pixels of the pixel (upper left, middle left, lower left, upper right, etc.), and a central pixel is larger than a certain domain pixel and is set to be 1; for the central pixel smaller than a domain pixel, set to 0, an 8-bit binary number is obtained as the feature of the position, which has 256 dimensions.
S133: carrying out dimension reduction processing on 256-dimensional features of each image block, wherein each processed image block is 59-dimensional features;
furthermore, the generated feature vector is 256-dimensional, and when matching is performed, the calculation amount is large, training in a subsequent neural network is difficult, and when the dimension is high, redundant irrelevant information is contained in many cases, so that dimension reduction is performed, and the dimension is changed from 256-dimensional to 59-dimensional. Since the number of changes of the binary cyclic sequence 0 and 1 of 90% or more of the information in the image does not exceed 2 times, both are included in 58 dimensions, such as the number of changes of 10100000 is 3, the number of changes exceeding 2 is classified into 59 th dimension, and there are two cases where the number of changes is 0, namely, all 0 and all 1, 0 for 1, and 7 × 8 for 2 is 56. The total number is 58. All other values are divided into 59 th dimensions, and the number of each dimension in the 59 th dimensions is counted.
S134: performing histogram calculation processing on 59-dimensional features of each processed image block to obtain a dimension-reduced LBP histogram;
further, histogram calculation is carried out on each block, and after histogram accumulation is carried out on all the blocks, a dimension-reduced LBP histogram is finally extracted.
S135: carrying out normalization processing on the dimension-reduced LBP histogram to obtain a normalized dimension-reduced LBP histogram;
further, the MAX-MIN method is adopted to carry out normalization processing on the dimension-reduced LBP histogram, and the dimension-reduced LBP histogram is normalized to be between [0 and 1 ].
S136: connecting the normalized dimension-reduced LBP histogram of each image block in series to obtain serial characteristic vector data;
further, the histograms of all blocks are concatenated, which results in the feature vector of the current detection window, with a vector length of 944(59 × 16).
S137: and carrying out normalization processing on the serial feature vector data, wherein each processed feature vector data is [0,0.5 ].
Furthermore, normalization processing method is adopted to normalize the feature vector data obtained by splicing to [0,0.5 ].
Further explanation of S14 is:
adopting a neural network model trained through a deep belief network to identify the finger vein characteristic information of the user to be identified, and identifying the identity information of the user to be identified; further, the acquired finger vein characteristic information of the user to be recognized is input into a trained neural network model for matching processing, a matching result of the user to be recognized is output, and the identity information of the user to be recognized is recognized.
Further, after the neural network model is subjected to deep belief network initialization parameters, continuously performing supervision training on training samples to obtain proper weights and biases of nerves, and storing the weights and the biases, wherein training labels are user serial numbers corresponding to the training samples, and the training samples are dimension-reduced LBP vein feature vectors extracted from finger vein images and spliced; during matching, the finger vein images collected by each person are preprocessed and feature extracted to obtain feature vectors, the feature vectors are input into a neural network with trained node weights and bias, and the output result is the matched user serial number.
Fig. 3 is a schematic flowchart of a training step of a neural network model in an embodiment of the present invention, and as shown in fig. 3, the training step of the neural network model includes:
s21: acquiring finger vein image information of a user to be trained;
further, finger vein image information of the user to be identified is obtained, the user to be identified places a finger on the vein image information acquisition device, and the vein image information acquisition device acquires and stores the vein image information on the finger as the finger vein image information of the user to be identified.
S22: image information preprocessing is carried out on the finger vein image information of the user to be trained, and the preprocessed image information to be trained is obtained;
further, image information preprocessing is carried out on the finger vein image information, and preprocessed image information is obtained; in the preprocessing process, image information graying processing, filtering noise reduction processing and normalization processing are required to be performed in sequence.
Further comprising the following steps: carrying out graying processing on the finger vein image information to obtain grayed image information; filtering and denoising the grayed image information to obtain denoised grayed image information; and carrying out normalization processing on the noise-reduced gray image information to obtain a normalization processing result.
In the process of graying the finger vein image information, a graying method conforming to the characteristics of human eyes is adopted, so that:
Gray＝0.299R+0.587G+0.114B
where R, G, B is the three-pass value of the original input color image information.
In the process of filtering and denoising the grayed image information, mean filtering is adopted to realize the processing, so that:
where M is 9, I is a region of 3 × 3 around the pixel, f (x, y) is the gray level of the pixel, and g (x, y) is the gray level of the pixel after mean filtering.
Normalizing the noise-reduced grayed image information by adopting an MAX-MIN method, wherein the pixel of each pixel of the noise-reduced grayed image information is within a range of [0,1] after the normalization processing; and normalizing the image size of the de-noised grayed image information by adopting a bilinear interpolation method, wherein the size of the processed image size is in a range of 64 x 64.
The pixel gray value normalization is to uniformly normalize the gray value of each pixel point to be 0-255, the normalization method is a MAX-MIN method, and the normalization formula is as follows:
wherein z (x, y) is the gray value of each pixel point after normalization.
The method comprises the steps of performing scale normalization processing by using a bilinear interpolation method, traversing each pixel point on image information, performing interpolation and adjustment in the x and y directions, and performing the scale normalization processing to obtain an image with the size of 64 x 64.
S23: performing dimensionality reduction local binary feature extraction processing on the preprocessed image information to be trained to obtain finger vein feature information of a user to be trained;
further, performing dimensionality reduction local binary feature extraction processing on the preprocessed image information to obtain finger vein feature information of the user to be identified; extracting a Local Binary feature (LBP) from each image information with the size of 64 x 64; reference is made specifically to S13.
S24: performing parameter initialization processing on the neural network by adopting a depth information network to obtain a parameter initialization neural network;
further, in a Restricted Boltzmann Machine (RBM), where v is a visible layer unit, h is a hidden layer unit, each node is a random binary variable node (which can only take a value of 0 or 1), and meanwhile, the total probability distribution satisfies Boltzmann distribution.
The energy function of the RBM is defined as:
E(v,h)＝-b'v-c'h-h'Wv
wherein W represents the weight connecting the hidden and visible units; b, c are the offsets of the visible layer and the hidden layer, respectively; meanwhile, because the hidden nodes of the binary graph are directly independent from each other:
by adjusting the parameters, the visual layer v1 obtained from the hidden layer is the same as the original visual layer v, and the hidden layer can be used as the characteristic of the input data of the visual layer; the joint probability density for a certain configuration can be determined by the Boltzmann distribution and the energy of this configuration:
wherein,
then, the maximum likelihood function learning algorithm trains the neural network by updating the hidden layer unit and the visualization unit at the same time, and the cost function is shown as a formula:
where m is the number of samples for the learning rate. The cost function is optimized by a gradient descent method. Finally, obtaining the proper weight values of v and h for constructing the deep belief network.
The DBN is obtained by the RBM through layer-by-layer training and stacking, and is therefore a deep-level graphical model for learning and extracting training data. Wherein the joint probability distribution of the initial vector x and the l hidden layer h ^ k is shown in equation 8:
wherein x is h0,P(hk-1|hk) Is the conditional probability of a visualization unit on a k-layer RBM on a hidden layer unit, P (h)l-1|hl) Is the joint distribution on the top-most RBM.
In this trial experiment, the training process of the deep belief network is as follows:
let input data x be h(0)Starting training as a visual layer of the first layer RBM;
a representation of the input data is obtained with the first layer for use on the second layer. Indicates the mean activation p (h) is selected(1)|h(0))；
The second tier RBM is trained using the average activation as a training sample.
And iterating 2 and 3 steps according to the number of layers, and upwards spreading the average value.
Fine-tuning all parameters of the deep network architecture.
S25: and inputting the finger vein characteristic information of the user to be trained into a parameter initialization neural network for training, and obtaining a neural network model for deep belief network training.
Further, the structure of the entire neural network is shown as follows, the network is composed of four layers, one input layer, two hidden layers and one output layer, since the number of images used for training is 64 × 12, the number of nodes of the input layer is 768 (including no bias node), the number of nodes of the two hidden layers is 50, 50, and the output layer node is 64, where θ(1)，θ(2)，θ(3)Initialized by the DBN parameter.
In the network:
an input layer: a is(1)＝x；
Hidden layer 1: z is a radical of(2)＝θ(1)a(1)，a(2)＝g(z(2))；
Hidden layer 2: z is a radical of(3)＝θ(2)a(2)，a(3)＝g(z(3))；
An output layer: z is a radical of(4)＝θ(3)a(3)，a(4)＝g(z(4))＝hθ(x)。
Where a denotes the value of the input node, z denotes the value of the output node, g denotes the activation function, and h denotes the prediction function.
The training process of the whole network is divided into three steps:
1. firstly, initializing hidden layer parameters through a DBN;
2. the output layer is initialized randomly;
3. and calculating an initial error function and gradient, and continuously iterating until the times are reached or the convergence is reached.
Fig. 4 is a schematic structural composition diagram of a finger vein recognition system based on a neural network model in an embodiment of the present invention, and as shown in fig. 4, the finger vein recognition system includes:
an image acquisition module: the finger vein recognition system is used for acquiring finger vein image information of a user to be recognized;
an image preprocessing module: the finger vein image processing device is used for preprocessing the image information of the finger vein image information and acquiring the preprocessed image information;
a feature extraction module: the system is used for performing dimensionality reduction local binary feature extraction processing on the preprocessed image information to acquire finger vein feature information of a user to be identified;
an identity recognition module: the neural network model is trained through a deep belief network and used for identifying the finger vein characteristic information of the user to be identified, and identifying the identity information of the user to be identified.
Preferably, the image preprocessing module comprises:
a graying processing unit: the finger vein image processing device is used for carrying out gray processing on the finger vein image information to obtain gray image information;
a noise reduction processing unit: the system is used for carrying out filtering noise reduction processing on the grayed image information to obtain the grayed image information after noise reduction;
a normalization processing unit: the normalization processing module is used for performing normalization processing on the grayed image information subjected to noise reduction to obtain a normalization processing result;
preferably, the normalizing the noise-reduced grayscale image information includes:
carrying out normalization processing on each pixel point of the noise-reduced grayed image information by adopting an MAX-MIN method, wherein the gray value of each pixel point is in a range of [0,255] after the normalization processing;
and normalizing the image size of the de-noised grayed image information by adopting a bilinear interpolation method, wherein the size of the processed image size is in a range of 64 x 64.
Preferably, the feature extraction module includes:
an image segmentation unit: the image information which is used for carrying out 4-4 segmentation processing on the preprocessed image information with the size of 64-64 is obtained, 16 image blocks are obtained, and the pixel of each image block is 16-16;
a pixel comparison unit: the image processing method comprises the steps of comparing each pixel in each image block with 8 field pixels of the pixels, setting the field pixels to be 0 if the field pixels are larger than the pixels, and otherwise setting the field pixels to be 1, and acquiring 8-bit binary numbers as 256-dimensional features of the corresponding image blocks;
a dimension reduction processing unit: the image processing method is used for carrying out dimension reduction processing on 256-dimensional features of each image block, and each processed image block is 59-dimensional features;
a histogram calculation unit: the histogram calculation processing module is used for performing histogram calculation processing on 59-dimensional features of each processed image block, and the processed image blocks are dimension-reduced LBP histograms;
a histogram normalization unit: the histogram is used for carrying out normalization processing on the dimension-reduced LBP histogram to obtain a normalized dimension-reduced LBP histogram;
a series unit: the normalized dimension-reduced LBP histogram is used for connecting each image block in series to obtain serial feature vector data;
a data normalization unit: and the method is used for carrying out normalization processing on the serial feature vector data, and each processed feature vector data is [0,0.5 ].
Preferably, the recognizing the finger vein feature information of the user to be recognized by using a neural network model trained through a deep belief network includes:
and inputting the acquired finger vein characteristic information of the user to be recognized into a neural network model trained through a deep belief network for matching, outputting a matching result of the user to be recognized, and recognizing the identity information of the user to be recognized.
Fig. 5 is a schematic structural component diagram of another finger vein recognition system based on a neural network model in an embodiment of the present invention, as shown in fig. 5, in this embodiment, a part of module implementation processes are processed with reference to the implementation process of fig. 4, which needs to be further described with reference to fig. 5 as follows:
the finger vein recognition system further comprises a neural network model training module:
an image acquisition unit: the device is used for acquiring finger vein image information of a user to be trained;
an image preprocessing unit: the device is used for preprocessing the image information of the finger vein image information of the user to be trained and acquiring the preprocessed image information to be trained;
a feature extraction unit: the system is used for performing dimensionality reduction local binary feature extraction processing on the preprocessed image information to be trained to acquire finger vein feature information of a user to be trained;
an initialization unit: the device is used for performing parameter initialization processing on the neural network by adopting a depth information network to obtain a parameter initialization neural network;
a training unit: and the neural network model is used for inputting the finger vein characteristic information of the user to be trained into the parameter initialization neural network for training, and acquiring the deep belief network training neural network model.
Specifically, the working principle of the system related function module according to the embodiment of the present invention may refer to the related description of the method embodiment, and is not described herein again.
In the embodiment of the invention, the problem of light change of the finger vein recognition system can be solved to a greater extent, the redundancy to noise is improved, the overall neural network is initialized by using the deep belief network, the convergence rate of the neural network is improved, the operation cost is reduced, the matching time of the overall finger vein recognition system is reduced, the regularization parameters are used in the matching stage to prevent overfitting, the matching accuracy is improved, and the false rejection rate and the false recognition rate are reduced.
Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by associated hardware instructed by a program, which may be stored in a computer-readable storage medium, and the storage medium may include: a Read Only Memory (ROM), a Random Access Memory (RAM), a magnetic or optical disk, or the like.
In addition, the finger vein recognition method and system based on the neural network model provided by the embodiment of the present invention are described in detail above, and a specific example should be adopted herein to explain the principle and the implementation of the present invention, and the description of the above embodiment is only used to help understanding the method and the core idea of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.
Claims (10)
1. A finger vein recognition method based on a neural network model is characterized by comprising the following steps:
acquiring finger vein image information of a user to be identified;
carrying out image information preprocessing on the finger vein image information to obtain preprocessed image information;
performing dimensionality reduction local binary feature extraction processing on the preprocessed image information to obtain finger vein feature information of a user to be identified;
and adopting a neural network model trained through a deep belief network to identify the finger vein characteristic information of the user to be identified, and identifying the identity information of the user to be identified.
2. The finger vein recognition method based on the neural network model according to claim 1, wherein the image information preprocessing of the acquired finger vein image information comprises:
carrying out graying processing on the finger vein image information to obtain grayed image information;
filtering and denoising the grayed image information to obtain denoised grayed image information;
and carrying out normalization processing on the noise-reduced gray image information to obtain a normalization processing result.
3. The method for recognizing the finger vein based on the neural network model according to claim 2, wherein the normalizing the noise-reduced grayed image information comprises:
carrying out normalization processing on each pixel point of the noise-reduced grayed image information by adopting an MAX-MIN method, wherein the gray value of each pixel point is in a range of [0,255] after the normalization processing;
and normalizing the image size of the de-noised grayed image information by adopting a bilinear interpolation method, wherein the size of the processed image size is in a range of 64 x 64.
4. The method for recognizing the finger vein based on the neural network model according to claim 1, wherein the performing the dimensionality reduction local binary feature extraction processing on the preprocessed image information includes:
4 × 4 segmentation processing is carried out on the image information with the size of 64 × 64 after the preprocessing, 16 image blocks are obtained, and the pixel of each image block is 16 × 16;
comparing each pixel in each image block with 8 field pixels of the pixels, if the field pixels are larger than the pixels, setting the field pixels to be 0, otherwise, setting the field pixels to be 1, and acquiring 8-bit binary numbers as 256-dimensional features of the corresponding image blocks;
carrying out dimension reduction processing on 256-dimensional features of each image block, wherein each processed image block is 59-dimensional features;
performing histogram calculation processing on 59-dimensional features of each processed image block to obtain a dimension-reduced LBP histogram;
carrying out normalization processing on the dimension-reduced LBP histogram to obtain a normalized dimension-reduced LBP histogram;
connecting the normalized dimension-reduced LBP histogram of each image block in series to obtain serial characteristic vector data;
and carrying out normalization processing on the serial feature vector data, wherein each processed feature vector data is [0,0.5 ].
5. The finger vein recognition method based on the neural network model according to claim 1, wherein the recognition processing of the finger vein feature information of the user to be recognized by using the neural network model trained by the deep belief network comprises:
and inputting the acquired finger vein characteristic information of the user to be recognized into a neural network model trained through a deep belief network for matching, outputting a matching result of the user to be recognized, and recognizing the identity information of the user to be recognized.
6. The neural network model-based finger vein recognition method according to claim 1 or 5, wherein the training step of the neural network model trained by the deep belief network comprises:
acquiring finger vein image information of a user to be trained;
image information preprocessing is carried out on the finger vein image information of the user to be trained, and the preprocessed image information to be trained is obtained;
performing dimensionality reduction local binary feature extraction processing on the preprocessed image information to be trained to obtain finger vein feature information of a user to be trained;
performing parameter initialization processing on the neural network by adopting a depth information network to obtain a parameter initialization neural network;
and inputting the finger vein characteristic information of the user to be trained into a parameter initialization neural network for training, and obtaining a neural network model for deep belief network training.
7. A finger vein recognition system based on a neural network model, the finger vein recognition system comprising:
an image acquisition module: the finger vein recognition system is used for acquiring finger vein image information of a user to be recognized;
an image preprocessing module: the finger vein image processing device is used for preprocessing the image information of the finger vein image information and acquiring the preprocessed image information;
a feature extraction module: the system is used for performing dimensionality reduction local binary feature extraction processing on the preprocessed image information to acquire finger vein feature information of a user to be identified;
an identity recognition module: the neural network model is trained through a deep belief network and used for identifying the finger vein characteristic information of the user to be identified, and identifying the identity information of the user to be identified.
8. The neural network model-based finger vein recognition system of claim 1, wherein the image preprocessing module comprises:
a graying processing unit: the finger vein image processing device is used for carrying out gray processing on the finger vein image information to obtain gray image information;
a noise reduction processing unit: the system is used for carrying out filtering noise reduction processing on the grayed image information to obtain the grayed image information after noise reduction;
a normalization processing unit: and the normalization processing module is used for performing normalization processing on the noise-reduced grayed image information to obtain a normalization processing result.
9. The neural network model-based finger vein recognition system of claim 1, wherein the feature extraction module comprises:
an image segmentation unit: the image information which is used for carrying out 4-4 segmentation processing on the preprocessed image information with the size of 64-64 is obtained, 16 image blocks are obtained, and the pixel of each image block is 16-16;
a pixel comparison unit: the image processing method comprises the steps of comparing each pixel in each image block with 8 field pixels of the pixels, setting the field pixels to be 0 if the field pixels are larger than the pixels, and otherwise setting the field pixels to be 1, and acquiring 8-bit binary numbers as 256-dimensional features of the corresponding image blocks;
a dimension reduction processing unit: the image processing method is used for carrying out dimension reduction processing on 256-dimensional features of each image block, and each processed image block is 59-dimensional features;
a histogram calculation unit: the histogram calculation processing module is used for performing histogram calculation processing on 59-dimensional features of each processed image block, and the processed image blocks are dimension-reduced LBP histograms;
a histogram normalization unit: the histogram is used for carrying out normalization processing on the dimension-reduced LBP histogram to obtain a normalized dimension-reduced LBP histogram;
a series unit: the normalized dimension-reduced LBP histogram is used for connecting each image block in series to obtain serial feature vector data;
a data normalization unit: and the method is used for carrying out normalization processing on the serial feature vector data, and each processed feature vector data is [0,0.5 ].
10. The neural network model-based finger vein recognition system of claim 7, further comprising a neural network model training module:
an image acquisition unit: the device is used for acquiring finger vein image information of a user to be trained;
an image preprocessing unit: the device is used for preprocessing the image information of the finger vein image information of the user to be trained and acquiring the preprocessed image information to be trained;
a feature extraction unit: the system is used for performing dimensionality reduction local binary feature extraction processing on the preprocessed image information to be trained to acquire finger vein feature information of a user to be trained;
an initialization unit: the device is used for performing parameter initialization processing on the neural network by adopting a depth information network to obtain a parameter initialization neural network;
a training unit: and the neural network model is used for inputting the finger vein characteristic information of the user to be trained into the parameter initialization neural network for training, and acquiring the deep belief network training neural network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710515995.9A CN107392114A (en) | 2017-06-29 | 2017-06-29 | A kind of finger vein identification method and system based on neural network model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710515995.9A CN107392114A (en) | 2017-06-29 | 2017-06-29 | A kind of finger vein identification method and system based on neural network model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107392114A true CN107392114A (en) | 2017-11-24 |
Family
ID=60334518
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710515995.9A Pending CN107392114A (en) | 2017-06-29 | 2017-06-29 | A kind of finger vein identification method and system based on neural network model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107392114A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108009520A (en) * | 2017-12-21 | 2018-05-08 | 东南大学 | A kind of finger vein identification method and system based on convolution variation self-encoding encoder neutral net |
| CN108171243A (en) * | 2017-12-18 | 2018-06-15 | 广州七乐康药业连锁有限公司 | A kind of medical image information recognition methods and system based on deep neural network |
| CN110084238A (en) * | 2019-04-09 | 2019-08-02 | 五邑大学 | Finger vena image segmentation method, device and storage medium based on LadderNet network |
| CN110147732A (en) * | 2019-04-16 | 2019-08-20 | 平安科技（深圳）有限公司 | Refer to vein identification method, device, computer equipment and storage medium |
| CN110298343A (en) * | 2019-07-02 | 2019-10-01 | 哈尔滨理工大学 | A kind of hand-written blackboard writing on the blackboard recognition methods |
| 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 |
| CN111166070A (en) * | 2019-12-17 | 2020-05-19 | 五邑大学 | Medical storage cabinet based on finger vein authentication and management method thereof |
| CN111639550A (en) * | 2020-05-13 | 2020-09-08 | 燕山大学 | Finger vein identification method based on multi-scale fusion LBP and improved depth confidence network |
| CN112560710A (en) * | 2020-12-18 | 2021-03-26 | 北京曙光易通技术有限公司 | Method for constructing finger vein recognition system and finger vein recognition system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102411711A (en) * | 2012-01-04 | 2012-04-11 | 山东大学 | Finger vein recognition method based on individualized weight |
| CN104102913A (en) * | 2014-07-15 | 2014-10-15 | 无锡优辰电子信息科技有限公司 | Wrist vein certification system |
| CN106548160A (en) * | 2016-11-09 | 2017-03-29 | 浙江博天科技有限公司 | A kind of face smile detection method |
-
2017
- 2017-06-29 CN CN201710515995.9A patent/CN107392114A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102411711A (en) * | 2012-01-04 | 2012-04-11 | 山东大学 | Finger vein recognition method based on individualized weight |
| CN104102913A (en) * | 2014-07-15 | 2014-10-15 | 无锡优辰电子信息科技有限公司 | Wrist vein certification system |
| CN106548160A (en) * | 2016-11-09 | 2017-03-29 | 浙江博天科技有限公司 | A kind of face smile detection method |
Non-Patent Citations (1)
| Title |
|---|
| CHENG CHEN,ETC.: "A Finger Vein Recognition Algorithm Using Feature Block Fusion and Depth Neural Network", 《INTERNATIONAL SYMPOSIUM ON COMPUTATIONAL INTELLIGENCE AND INTELLIGENT SYSTEMS》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108171243A (en) * | 2017-12-18 | 2018-06-15 | 广州七乐康药业连锁有限公司 | A kind of medical image information recognition methods and system based on deep neural network |
| CN108009520A (en) * | 2017-12-21 | 2018-05-08 | 东南大学 | A kind of finger vein identification method and system based on convolution variation self-encoding encoder neutral net |
| CN110084238A (en) * | 2019-04-09 | 2019-08-02 | 五邑大学 | Finger vena image segmentation method, device and storage medium based on LadderNet network |
| CN110084238B (en) * | 2019-04-09 | 2023-01-03 | 五邑大学 | Finger vein image segmentation method and device based on LadderNet network and storage medium |
| CN110147732A (en) * | 2019-04-16 | 2019-08-20 | 平安科技（深圳）有限公司 | Refer to vein identification method, device, computer equipment and storage medium |
| CN110298343A (en) * | 2019-07-02 | 2019-10-01 | 哈尔滨理工大学 | A kind of hand-written blackboard writing on the blackboard recognition methods |
| CN111166070A (en) * | 2019-12-17 | 2020-05-19 | 五邑大学 | Medical storage cabinet based on finger vein authentication and management method thereof |
| 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 |
| CN111639550A (en) * | 2020-05-13 | 2020-09-08 | 燕山大学 | Finger vein identification method based on multi-scale fusion LBP and improved depth confidence network |
| CN112560710A (en) * | 2020-12-18 | 2021-03-26 | 北京曙光易通技术有限公司 | Method for constructing finger vein recognition system and finger vein recognition system |
| CN112560710B (en) * | 2020-12-18 | 2024-03-01 | 北京曙光易通技术有限公司 | Method for constructing finger vein recognition system and finger vein recognition system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107392114A (en) | A kind of finger vein identification method and system based on neural network model | |
| CN108009520B (en) | Finger vein identification method and system based on convolution variational self-encoder network | |
| US10726260B2 (en) | Feature extraction and matching for biometric authentication | |
| Das et al. | A new efficient and adaptive sclera recognition system | |
| JP2008123521A (en) | Face recognition method and device using extended gabor wavelet feature | |
| Salve et al. | Iris recognition using SVM and ANN | |
| CN114612968A (en) | Convolutional neural network-based lip print identification method | |
| Lazimul et al. | Fingerprint liveness detection using convolutional neural network and fingerprint image enhancement | |
| CN112560710A (en) | Method for constructing finger vein recognition system and finger vein recognition system | |
| CN112116012A (en) | Finger vein instant registration and identification method and system based on deep learning | |
| Deepamalar et al. | An enhanced palm vein recognition system using multi-level fusion of multimodal features and adaptive resonance theory | |
| Sallam et al. | Iris recognition system using deep learning techniques | |
| Tayade et al. | Sclera feature extraction using DWT co-efficients | |
| Mazumdar et al. | Biometric authentication system based on texture features of retinal images | |
| Harakannanavar et al. | Performance evaluation of face recognition based on multiple feature descriptors using Euclidean distance classifier | |
| Alhanjouri | Improved HMM by deep learning for ear classification | |
| Abdullahi et al. | Performance analysis of particle swarm optimization algorithm-based parameter tuning for fingerprint image enhancement | |
| Trofimov et al. | Palm vein identification based on vein segmentation and triplet loss function | |
| Rao et al. | Fusion Biometric Deep Features Blended in Its Authentication | |
| CN107844735A (en) | Authentication method and device for biological characteristics | |
| Jeremiah | Enhanced multimodal biometric access control system using chicken swarm optimization and self-organizing feature maps: a study on ear and iris recognition | |
| Sharma et al. | Experimental Analysis of a Multimodal biometric System using Preprocessing and Feature Extraction Techniques and Their Impact on Analytical Results | |
| UDROIU et al. | USABLE SECURITY IN BIOMETRIC AUTHENTICATION SYSTEMS | |
| Sathiya et al. | AN ENHANCED ENSEMBLE HYBRID DEEP LEARNING ALGORITHM FOR IMPROVING THE ACCURACY IN IRIS SEGMENTATION. | |
| Aithal | An Alternative Approach to Fingerprint Hash Code Generation Based on Modified Filtering Techniques |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20171124 |